{ "edition": "PE-2026.10-v4", "releaseSha256": "136612b7e4aeaf2c18e303c81a5d860e91680f65b8fba78e62adedf85bbb2dae", "reports": [ { "slug": "ai-native-enterprise", "number": 1, "title": "What makes an AI-native enterprise?", "shortTitle": "What makes an AI-native enterprise?", "dek": "A research synthesis on work, organizational capability and the evidence needed to measure them.", "sector": null, "kind": "Research synthesis", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "59 of 80 selected entity records contain qualifying AI operating evidence in this review. 21 remain unranked because no qualifying claim was located; this does not establish absent capability.", "The corpus contains 272 recorded claims, of which 224 meet the declared eligibility rules. Specific commitments, deployed practices and operating detail are coded separately.", "Decision Lineage appears in 4 of these 80 profiles, compared with 44 for AI Core. These are counts within the selected research corpus, not enterprise adoption rates." ], "executiveSources": [], "evidenceAgeContext": "Dated sources run from 2019-05-14 to 2026-10-01. 79 of 224 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [ { "id": "N01", "displayNumber": 3, "title": "Nine operating questions for the enterprise", "kind": "narrative" } ], "exhibitRegistry": [ { "id": "E01", "displayNumber": 1, "title": "Where evidence is visible across the eight sector cohorts", "kind": "generated" }, { "id": "E02", "displayNumber": 2, "title": "The evidence record includes what did not qualify", "kind": "generated" }, { "id": "N01", "displayNumber": 3, "title": "Nine operating questions for the enterprise", "kind": "narrative" }, { "id": "E03", "displayNumber": 4, "title": "Operating detail deserves a separate test", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "Build a benchmark from evidence to a decision", "kind": "generated" } ], "exhibitOrder": [ "E01", "E02", "N01", "E03", "E04" ], "metrics": [ { "value": 80, "label": "selected entity records" }, { "value": 272, "label": "recorded claims" }, { "value": 224, "label": "eligible claims" }, { "value": 21, "label": "unranked records" } ], "exhibits": [ { "id": "E01", "type": "coverage", "title": "Where evidence is visible across the eight sector cohorts", "subtitle": "Number of selected entity records with a qualifying claim in each theme; cohort size shown per row", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "label": "Aviation & Logistics", "n": 10, "values": [ 7, 3, 5, 3, 1, 2, 0, 2, 2 ] }, { "label": "Banking", "n": 10, "values": [ 7, 8, 4, 7, 2, 3, 2, 4, 7 ] }, { "label": "Energy & Utilities", "n": 10, "values": [ 7, 6, 5, 3, 0, 3, 1, 3, 1 ] }, { "label": "Healthcare", "n": 10, "values": [ 6, 3, 3, 4, 0, 3, 0, 1, 4 ] }, { "label": "Manufacturing & Materials", "n": 10, "values": [ 5, 2, 2, 1, 0, 3, 0, 2, 2 ] }, { "label": "Real Estate & Construction", "n": 10, "values": [ 4, 1, 4, 3, 0, 3, 1, 1, 0 ] }, { "label": "Retail & Consumer", "n": 10, "values": [ 3, 1, 3, 5, 0, 0, 1, 1, 2 ] }, { "label": "Telecommunications", "n": 10, "values": [ 5, 5, 3, 4, 1, 2, 3, 3, 3 ] } ], "note": "Source: frozen Mayden public-evidence corpus. Counts describe located evidence, not comparable sector maturity or prevalence. Cohorts are purposeful, heterogeneous and sometimes overlap.", "displayNumber": 1 }, { "id": "E02", "type": "funnel", "title": "The evidence record includes what did not qualify", "subtitle": "Recorded claims pass explicit access, scope, deployment and review gates", "maximum": 272, "rows": [ { "label": "All recorded claims", "value": 272 }, { "label": "Eligible for scoring", "value": 224 }, { "label": "Excluded / held", "value": 48 } ], "note": "Source: release claim ledger. These counts do not measure research recall, factual accuracy or company capability. All exclusions remain inspectable in the public dataset.", "displayNumber": 2 }, { "id": "E03", "type": "stages", "title": "Operating detail deserves a separate test", "subtitle": "Eligible claim records, n = 224; disclosure anchors are not maturity stages", "maximum": 97, "rows": [ { "label": "Specific commitment / pilot", "value": 53 }, { "label": "Implemented practice", "value": 74 }, { "label": "Implemented + operating detail", "value": 97 } ], "note": "Source: reviewed coding. Detail is not independent assurance, and multiple claim records may describe a shared underlying practice.", "displayNumber": 4 }, { "id": "E04", "type": "process", "title": "Build a benchmark from evidence to a decision", "subtitle": "Mayden research architecture; conceptual framework, not an observed causal model", "rows": [ { "label": "Define", "detail": "Entity, workflow, population and period." }, { "label": "Observe", "detail": "Sources, tasks, usage and operating boundaries." }, { "label": "Challenge", "detail": "Alternative explanations, controls and exclusions." }, { "label": "Calculate", "detail": "Declared rules, sensitivity and reproducibility." }, { "label": "Decide", "detail": "Specific action, evidence gap and next test." } ], "note": "Source: Mayden methodological synthesis. Confidential company research remains a separate instrument requiring permission, respondent testing and validation.", "displayNumber": 5 } ], "cases": [], "agenda": [ { "question": "Choose one recurring workflow.", "evidence": "Its users, decisions, frequency, business importance and operating boundary." }, { "question": "Make authority and evidence inspectable.", "evidence": "Input provenance, evaluations, interventions, accountable owner and recovery path." }, { "question": "Track value with a credible comparator.", "evidence": "Quality, cost, risk and adoption denominators over a stated period." } ], "sources": [ { "title": "Generative AI at Work", "publisher": "Quarterly Journal of Economics", "date": "2025-02-04", "url": "https://doi.org/10.1093/qje/qjae044", "locator": "Staggered workplace rollout; observational causal design", "relationship": "peer-reviewed empirical", "id": "S01" }, { "title": "Experimental evidence on the productivity effects of generative artificial intelligence", "publisher": "Science", "date": "2023-07-13", "url": "https://doi.org/10.1126/science.adh2586", "locator": "Preregistered randomized online experiment", "relationship": "peer-reviewed experiment", "id": "S02" }, { "title": "Navigating the Jagged Technological Frontier", "publisher": "Harvard Business School working paper 24-013", "date": "2023-09", "url": "https://www.hbs.edu/ris/download.aspx?name=24-013.pdf", "locator": "Randomized field experiment", "relationship": "working-paper experiment", "id": "S03" }, { "title": "The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise", "publisher": "NBER working paper 33641; published Organization Science June 2026", "date": "2025-04", "url": "https://www.nber.org/papers/w33641", "locator": "Preregistered randomized field experiment", "relationship": "original field experiment", "id": "S04" }, { "title": "Shifting Work Patterns with Generative AI", "publisher": "NBER working paper 33795, revision 1", "date": "2025-05-06", "url": "https://www.nber.org/system/files/working_papers/w33795/revisions/w33795.rev1.pdf", "locator": "Six-month randomized field experiment", "relationship": "working-paper experiment", "id": "S05" }, { "title": "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity", "publisher": "METR", "date": "2025-07-10", "url": "https://metr.org/Early_2025_AI_Experienced_OS_Devs_Study-paper.pdf", "locator": "Task-randomized controlled trial", "relationship": "original experiment", "id": "S06" }, { "title": "We are Changing our Developer Productivity Experiment Design", "publisher": "METR", "date": "2026-02-24", "url": "https://metr.org/blog/2026-02-24-uplift-update/", "locator": "Follow-up randomized study; methodological disclosure", "relationship": "original follow-up methods report", "id": "S07" }, { "title": "Firm Data on AI", "publisher": "NBER working paper 34836; author-hosted February version", "date": "2026-02", "url": "https://wfhresearch.com/wp-content/uploads/2026/03/Firm-Data-on-AI-16-February-2026.pdf", "locator": "Stratified firm surveys", "relationship": "original survey", "id": "S08" }, { "title": "When combinations of humans and AI are useful: A systematic review and meta-analysis", "publisher": "Nature Human Behaviour", "date": "2024-10-28", "url": "https://doi.org/10.1038/s41562-024-02024-1", "locator": "Preregistered systematic review/meta-analysis", "relationship": "peer-reviewed evidence synthesis", "id": "S09" }, { "title": "Artificial intelligence capability: Conceptualization, measurement calibration, and empirical study on its impact on organizational creativity and firm performance", "publisher": "Information & Management 58(3)", "date": "2021-04-01", "url": "https://doi.org/10.1016/j.im.2021.103434", "locator": "Instrument development and cross-sectional survey", "relationship": "peer-reviewed survey and measurement", "id": "S10" }, { "title": "The Productivity J-Curve: How Intangibles Complement General Purpose Technologies", "publisher": "NBER; American Economic Journal: Macroeconomics 13(1), 2021", "date": "2021-01", "url": "https://www.nber.org/papers/w25148", "locator": "Economic model and historical empirical analysis", "relationship": "peer-reviewed theory and empirical analysis", "id": "S11" }, { "title": "Measuring and Explaining Management Practices Across Firms and Countries", "publisher": "NBER; Quarterly Journal of Economics 122(4), 2007", "date": "2007-11", "url": "https://www.nber.org/papers/w12216", "locator": "Structured management-practice interviews and linked performance analysis", "relationship": "peer-reviewed organizational measurement", "id": "S12" }, { "title": "Artificial Intelligence Risk Management Framework (AI RMF 1.0)", "publisher": "NIST", "date": "2023-01", "url": "https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf", "locator": "Voluntary risk-management framework", "relationship": "official framework", "id": "S13" }, { "title": "Handbook on Constructing Composite Indicators: Methodology and User Guide", "publisher": "OECD / European Union / EC-JRC", "date": "2008-08-22", "url": "https://doi.org/10.1787/9789264043466-en", "locator": "Methodological handbook", "relationship": "measurement guidance", "id": "S14" }, { "title": "Best Practices for Developing and Validating Scales for Health, Social, and Behavioral Research: A Primer", "publisher": "Frontiers in Public Health", "date": "2018-06-11", "url": "https://doi.org/10.3389/fpubh.2018.00149", "locator": "Methodological primer", "relationship": "peer-reviewed measurement guidance", "id": "S15" } ], "body": "# What makes an AI-native enterprise?\n## The operating model matters more than the announcement\n\nMayden AI Native Index · Research foundation · 10 October 2026\n\n**An AI-native enterprise repeatedly converts AI capability into useful, accountable work across a defined organizational scope.** This is Mayden's proposed working definition, to be tested through field research. It asks what changes in decisions, workflows, responsibilities and learning when AI becomes part of normal operations.\n\nA company can have thousands of AI licences and little operating change. Another can use a small number of models inside essential workflows, with clear ownership, sound evidence and effective fallback arrangements. A benchmark that rewards purchasing activity will miss this difference. A benchmark that rewards autonomy without considering consequence may reward poor judgment.\n\nThe research offers a strong foundation for measuring these distinctions. It does not yet provide a universally validated definition of an AI-native enterprise. Mayden's opportunity is to build that measurement carefully, beginning with Saudi Arabia, the UAE, Qatar, Kuwait and Bahrain.\n\n### Productivity gains are real, conditional and uneven\n\nThe published customer-support study by Brynjolfsson, Li and Raymond follows 5,172 agents. AI assistance increased resolved issues per hour by 15% on average, with larger gains among less experienced workers. This is evidence about a particular deployed workflow; it is not a forecast for every company's profit margin. The published figures differ from earlier working-paper versions, a reminder that a research database needs version control. [Published study](https://doi.org/10.1093/qje/qjae044).\n\nIn a randomized experiment with 453 professionals, Noy and Zhang found faster completion and higher assessed quality on writing tasks. That demonstrates potential at the task level. It does not incorporate every approval, integration, legal review or downstream correction in an enterprise process. The Index should therefore ask whether faster drafting improved the accepted final output and the total process. [Writing experiment](https://doi.org/10.1126/science.adh2586).\n\nThe consulting experiment involving 758 BCG professionals reinforces the need to evaluate task suitability. Its findings describe a jagged boundary: apparently similar tasks can sit on different sides of what the tool handles reliably. A serious company must know where its deployed systems succeed and where users need another approach. Counting deployments cannot reveal that knowledge. [HBS working paper](https://www.hbs.edu/ris/download.aspx?name=24-013.pdf).\n\nThe executive question is precise: **Which material workflow performs better, against which baseline, after counting the work needed to make its outputs usable?**\n\n### AI can change expertise and collaboration\n\nIn the Procter & Gamble experiment, 776 professionals worked on product innovation challenges. Individuals equipped with AI matched the assessed performance of teams without AI. The study opens an important question about how expertise can travel across functional boundaries. It does not establish that companies should remove teams or that temporary experimental performance will persist in ordinary operations. [Cybernetic Teammate study](https://www.nber.org/papers/w33641).\n\nA separate six-month experiment with 7,137 knowledge workers found changes in email and document work without a significant change in meeting time in its May 2025 revision. The contrast points toward a useful hypothesis: individual changes can be easier than coordinated organizational changes. Giving people tools does not itself settle decision rights, redesign handoffs or change shared routines. [Pinned working-paper revision](https://www.nber.org/system/files/working_papers/w33795/revisions/w33795.rev1.pdf).\n\nOur next company study will test that hypothesis through actual operating examples. Who now prepares a recommendation? Who verifies it? Who can act on it? Which handoff has disappeared, and which new review step has appeared? A before-and-after workflow can reveal more than a general claim that employees are empowered.\n\n### Perceived improvement is not measured improvement\n\nMETR's early-2025 experiment found that 16 experienced open-source developers took longer on AI-allowed tasks, despite believing that the tools accelerated them. The result is bounded by its selected developers, mature repositories and tools of that period. It is useful evidence against treating users' perceived time savings as an objective productivity measure. [Original experiment](https://metr.org/Early_2025_AI_Experienced_OS_Devs_Study-paper.pdf).\n\nThe same research organization later explained why its follow-up estimates were unreliable. Developers' willingness to participate, their task selection and time accounting changed. Its February 2026 report therefore cannot support a clean claim that the original effect has reversed by a precise amount. This is a lesson for any longitudinal index: the people and work entering the measurement may change alongside the technology. [Follow-up methods report](https://metr.org/blog/2026-02-24-uplift-update/).\n\nA 2026 survey of almost 6,000 executives across four non-GCC economies also separates widespread reported AI use from limited reported realized effects. Executive expectations were more optimistic than descriptions of past results. These are survey reports, not audited financial effects. For Mayden, the implication is to record use, measured outcomes and future expectations in separate fields. [February 2026 author version](https://wfhresearch.com/wp-content/uploads/2026/03/Firm-Data-on-AI-16-February-2026.pdf).\n\n### Better combinations must be demonstrated\n\nAn existing meta-analysis of 106 experiments found that human–AI combinations, on average in that corpus, underperformed the better of the human or AI working alone. Results differed by task type. This is not the claim that AI assistance never helps people; the comparison is with the better standalone performer. Its study window ended in June 2023, so it cannot establish the performance of today's systems. [Published meta-analysis](https://doi.org/10.1038/s41562-024-02024-1).\n\nThe practical lesson is to test the combination. Human review needs enough information, time and authority to improve the result. A nominal approval click should not be treated as effective oversight. Equally, the correct design may deliberately retain human judgment even where automation is faster, because accountability and consequence matter.\n\nMayden's score should reward an appropriate operating design. It should not award automatic bonus points for replacing people, maximizing autonomy or adopting a particular model.\n\n### Capability includes organizational resources\n\nMikalef and Gupta developed a multidimensional AI capability instrument covering technical, human and organizational resources. Their survey provides useful construct-development precedent, but its associations with performance are not causal proof and its sample predates the current generative-AI environment. This work informs our question design; local validation remains necessary. [AI capability study](https://doi.org/10.1016/j.im.2021.103434).\n\nThe productivity J-curve literature explains why complementary investments in processes and skills can complicate the timing of measured gains. It provides a plausible mechanism, not an excuse for an unproductive project to continue indefinitely. Assessment should record implementation costs, expected timing, milestones and explicit stopping criteria. [J-curve research](https://www.nber.org/papers/w25148).\n\nOrganizational measurement also has an established precedent in the management-practices research of Bloom and Van Reenen. Structured evidence about how firms operate can be related to performance. That supports asking for operational examples and training assessors. It does not make a newly authored AI score valid by association. [Management measurement study](https://www.nber.org/papers/w12216).\n\n### Nine questions that executives should be able to answer\n\nMayden's existing nine themes can form a coherent research agenda. They remain candidate dimensions, with overlap and weighting to be tested.\n\n## Exhibit 03 — Nine operating questions for the enterprise\n\n| Theme | The executive question |\n|---|---|\n| AI Core | Where does AI support material recurring work, using reliable data and a workable fallback? |\n| Workforce | Can people use, challenge and improve AI in the roles they actually perform? |\n| Ecosystem | Can we evaluate, maintain and, when necessary, change critical AI dependencies? |\n| Interfaces | Is AI part of the workflow, and what share of eligible work actually uses it? |\n| Decision lineage | Can we reconstruct an important output, its sources and the actions that followed? |\n| Experimentation | What evidence causes us to scale, revise or stop a deployment? |\n| Operating model | Who owns the outcome and can authorize, halt or reverse consequential actions? |\n| Knowledge | Who keeps the organization's usable knowledge current, accessible and trustworthy? |\n| Governance | Can we discover uses, evaluate risks and respond to incidents across the lifecycle? |\n\nThese questions are deliberately about behavior. A policy can contribute evidence, but it does not prove implementation. A training course can contribute evidence, but it does not prove competence. A partnership can provide resources, but it does not prove a deployed operating practice.\n\nNIST's risk-management framework offers useful structure for lifecycle controls and responsibilities. It is guidance, not a certificate of compliance or a validated productivity score. [AI RMF 1.0](https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1.pdf).\n\n### A credible benchmark separates three records\n\nFirst is **operating capability**: what practices exist, over what scope and with what continuity. Second is **outcome evidence**: changes in quality, speed, cost, revenue, risk and work experience. Third is **evidence strength**: the basis for believing each statement.\n\nKeeping these records separate prevents a circular argument in which high-performing companies receive high capability scores because their outcomes were included in the score, and the same outcomes are then used to claim the score predicts performance.\n\nIt also prevents public visibility from masquerading as capability. A company with a detailed annual report may be easier to assess than a private firm. Lack of published evidence is an information gap. It is not proof that a practice is absent. The initial public-document product must therefore identify itself as a **Public Evidence Index**, with explicit limits on what its ranking means.\n\nNormalization cannot repair a weak construct. The OECD/JRC handbook makes weighting, aggregation and sensitivity part of the design problem. A fixed scale is useful for repeat measurement, but the anchors still need justification. [Composite-indicator handbook](https://doi.org/10.1787/9789264043466-en).\n\n### Build a regional benchmark that can travel\n\nThe five-market starting scope is a deliberate launch cohort, not a claim to cover the entire GCC. Country, sector, company size, ownership and operating scope should be visible. A regional subsidiary must not inherit a global parent's capabilities without evidence that those practices apply locally.\n\nArabic and English instruments require more than literal translation. Participants should explain what they understood each question to mean; researchers should investigate differences by language and interview mode. Voice answers need confirmation before being transformed into scores. Scale-development guidance supports an explicit progression from item design to testing and validation. [Measurement primer](https://doi.org/10.3389/fpubh.2018.00149).\n\nThe valuable longitudinal question is not simply who ranks first this year. It is **which operating changes predict sustained improvements, under which conditions, and which apparently promising practices fail to deliver**. Answering it requires repeat participants, preserved instrument versions, comparable outcomes and honest treatment of firms that stop responding.\n\nA useful confidential report can help sustain that participation. It should show an organization's supported strengths, unresolved questions and a few concrete priorities. The public research can then explain patterns without exposing entrusted company evidence.\n\n### Research note\n\nThis report is a scoped rapid evidence review and Mayden design synthesis. It is not a systematic review or a new statistical meta-analysis. Fifteen research and methodological records were selected through targeted primary-source searches on 10 October 2026. Selection prioritized identifiable original studies, peer-reviewed measurement work and authoritative methodological guidance; it was not an exhaustive database search.\n\nExperiments, observational studies, executive surveys, an existing meta-analysis and frameworks are kept distinct. No pooled productivity percentage is calculated. Source access ranged from publisher/author abstracts and indexed passages to relevant report sections; this is not a claim that every full paper and appendix was independently audited. Publication versions are pinned where findings changed. None of these sources establishes GCC company rankings, validates Mayden's nine dimensions or replaces proprietary fieldwork.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/ai-native-enterprise.pdf", "cover": "/research/index/2026-10/ai-native-enterprise.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 74511, "pdfSha256": "13860418dd82b9d8fc68a5a6fe41b14836e22410efe4d13f8a6d427dbd7c05ea", "pdfPages": 13 }, { "slug": "banking-public-evidence", "number": 2, "title": "AI in Gulf banking: inspect the operating evidence", "shortTitle": "Banking on operating evidence", "dek": "A purposeful review of ten banking entities across five headquarters markets.", "sector": "Banking", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Trace the decision behind the answer. KFH reports employee guidance linked to supervisory-board fatwas. Test whether answers preserve source accuracy and currency through actual use.", "Inspect shared controls in operation. Emirates NBD describes agent logging, evaluation and accountable owners. Request a completed case showing how these controls affected deployment or intervention.", "Separate availability from adoption. NBK reports broad Copilot availability alongside a narrower consistent-user population. Establish who uses the tool, for which work, and how accepted outputs are measured." ], "executiveSources": [ { "claimIds": [ "bank-kfh-01", "bank-kfh-03" ], "sources": [ { "claimId": "bank-kfh-01", "url": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Reinforces-Banking-Leadership-with-Forward-Looking-AI-Initiatives.html", "sourceTitle": "KFH Reinforces Banking Leadership with Forward-Looking AI Initiatives" }, { "claimId": "bank-kfh-03", "url": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Reinforces-Banking-Leadership-with-Forward-Looking-AI-Initiatives.html", "sourceTitle": "KFH Reinforces Banking Leadership with Forward-Looking AI Initiatives" } ] }, { "claimIds": [ "bank-emirates-nbd-03", "bank-emirates-nbd-04", "bank-emirates-nbd-07" ], "sources": [ { "claimId": "bank-emirates-nbd-03", "url": "https://www.emiratesnbd.com/en/innovation/insights/why-banks-should-build-their-own-agentic-ai-framework", "sourceTitle": "Why banks should build their own Agentic AI Framework" }, { "claimId": "bank-emirates-nbd-04", "url": "https://www.emiratesnbd.com/en/innovation/insights/why-banks-should-build-their-own-agentic-ai-framework", "sourceTitle": "Why banks should build their own Agentic AI Framework" }, { "claimId": "bank-emirates-nbd-07", "url": "https://www.emiratesnbd.com/en/innovation/innovation-in-action", "sourceTitle": "Innovation in Action" } ] }, { "claimIds": [ "bank-nbk-01", "bank-nbk-08", "bank-nbk-09" ], "sources": [ { "claimId": "bank-nbk-01", "url": "https://www.nbk.com/kuwait/news-and-insights/media-relations/news.html?news=nbk-launches-microsoft-copilot-across-all-divisions--pioneering-ai-integration-and-empowering-employees-", "sourceTitle": "NBK Launches Microsoft Copilot Across All Divisions" }, { "claimId": "bank-nbk-08", "url": "https://argaamplus.s3.amazonaws.com/4b4e6fe7-09dd-4311-be33-c97735b392e7.pdf", "sourceTitle": "NBK Sustainability Report 2025" }, { "claimId": "bank-nbk-09", "url": "https://argaamplus.s3.amazonaws.com/4b4e6fe7-09dd-4311-be33-c97735b392e7.pdf", "sourceTitle": "NBK Sustainability Report 2025" } ] } ], "evidenceAgeContext": "Dated sources run from 2025 to 2026-07-22. 5 of 52 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "Three banking evidence trails, three different tests", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 63, "label": "recorded claims" }, { "value": 52, "label": "eligible claims" }, { "value": 1, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "Three banking evidence trails, three different tests", "subtitle": "Operating object → information used → decision evidence requested. These are separate cases, not a linked system.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "Customer conversation", "unit": "Service-improvement issue", "finding": "Emirates NBD says PCA turns calls into recurring themes, intent and operational issues shared across teams.", "boundary": "Named platform; decision-use quality not demonstrated.", "implication": "Which issue was acted on, and did the service outcome change?", "claimIds": [ "bank-emirates-nbd-01", "bank-emirates-nbd-02" ], "sources": [ { "claimId": "bank-emirates-nbd-01", "url": "https://www.emiratesnbd.com/en/innovation/insights/turning-millions-of-customer-conversations-into-better-banking-experiences-with-ai", "sourceTitle": "Turning millions of customer conversations into better banking experiences with AI" }, { "claimId": "bank-emirates-nbd-02", "url": "https://www.emiratesnbd.com/en/innovation/insights/turning-millions-of-customer-conversations-into-better-banking-experiences-with-ai", "sourceTitle": "Turning millions of customer conversations into better banking experiences with AI" } ] }, { "label": "Policy answer", "unit": "Employee guidance", "finding": "KFH says ShariaBot references specific supervisory-board fatwas when answering staff questions.", "boundary": "KFH Kuwait; source links do not reconstruct the full decision.", "implication": "Can a reviewer recover the source version and check a disputed answer?", "claimIds": [ "bank-kfh-01", "bank-kfh-03" ], "sources": [ { "claimId": "bank-kfh-01", "url": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Reinforces-Banking-Leadership-with-Forward-Looking-AI-Initiatives.html", "sourceTitle": "KFH Reinforces Banking Leadership with Forward-Looking AI Initiatives" }, { "claimId": "bank-kfh-03", "url": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Reinforces-Banking-Leadership-with-Forward-Looking-AI-Initiatives.html", "sourceTitle": "KFH Reinforces Banking Leadership with Forward-Looking AI Initiatives" } ] }, { "label": "Permitted data", "unit": "Employee workspace", "finding": "NBK says Copilot Web excludes confidential data and disables document upload and download.", "boundary": "Specified access controls; broader framework remains on the roadmap.", "implication": "Can an attempted prohibited action demonstrate that the boundary is enforced?", "claimIds": [ "bank-nbk-07" ], "sources": [ { "claimId": "bank-nbk-07", "url": "https://argaamplus.s3.amazonaws.com/4b4e6fe7-09dd-4311-be33-c97735b392e7.pdf", "sourceTitle": "NBK Sustainability Report 2025" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "emirates-nbd", "label": "Emirates NBD", "values": [ 3, 3, 3, 2, 3, 3, 2, 3, 3 ], "score": 93, "rank": 1 }, { "id": "fab", "label": "First Abu Dhabi Bank", "values": [ 2, 2, null, 2, null, 2, null, 3, 2 ], "score": 48, "rank": 2 }, { "id": "kfh", "label": "Kuwait Finance House", "values": [ 3, null, null, 2, 2, null, null, 3, 3 ], "score": 48, "rank": 2 }, { "id": "cbq", "label": "Commercial Bank of Qatar", "values": [ 3, 2, null, 3, null, null, 2, 2, null ], "score": 44, "rank": 4 }, { "id": "nbk", "label": "National Bank of Kuwait", "values": [ 2, 3, null, 3, null, null, null, null, 3 ], "score": 41, "rank": 5 }, { "id": "al-rajhi", "label": "Al Rajhi Bank", "values": [ 3, 3, 3, null, null, null, null, null, 1 ], "score": 37, "rank": 6 }, { "id": "qnb", "label": "QNB", "values": [ null, 3, 1, 2, null, 1, null, null, 3 ], "score": 37, "rank": 6 }, { "id": "bbk", "label": "Bank of Bahrain and Kuwait", "values": [ null, 2, 1, 2, null, null, null, null, 3 ], "score": 30, "rank": 8 }, { "id": "snb", "label": "Saudi National Bank", "values": [ 3, 1, null, null, null, null, null, null, null ], "score": 15, "rank": 9 }, { "id": "nbb", "label": "National Bank of Bahrain", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "Emirates NBD", "value": 93, "secondary": 9, "rank": 1 }, { "label": "First Abu Dhabi Bank", "value": 48, "secondary": 6, "rank": 2 }, { "label": "Kuwait Finance House", "value": 48, "secondary": 5, "rank": 2 }, { "label": "Commercial Bank of Qatar", "value": 44, "secondary": 5, "rank": 4 }, { "label": "National Bank of Kuwait", "value": 41, "secondary": 4, "rank": 5 }, { "label": "Al Rajhi Bank", "value": 37, "secondary": 4, "rank": 6 }, { "label": "QNB", "value": 37, "secondary": 5, "rank": 6 }, { "label": "Bank of Bahrain and Kuwait", "value": 30, "secondary": 4, "rank": 8 }, { "label": "Saudi National Bank", "value": 15, "secondary": 2, "rank": 9 }, { "label": "National Bank of Bahrain", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 52; one practice may support multiple distinct themes", "maximum": 27, "rows": [ { "label": "Specific commitment / pilot", "value": 5 }, { "label": "Implemented practice", "value": 20 }, { "label": "Implemented + operating detail", "value": 27 } ], "note": "Source: claim ledger. 11 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "Emirates NBD", "value": 92.6, "after": 59.26 }, { "label": "First Abu Dhabi Bank", "value": 48.1, "after": 7.41 }, { "label": "Kuwait Finance House", "value": 48.1, "after": 22.22 }, { "label": "Commercial Bank of Qatar", "value": 44.4, "after": 14.81 }, { "label": "National Bank of Kuwait", "value": 40.7, "after": 25.93 }, { "label": "Al Rajhi Bank", "value": 37.0, "after": 25.93 }, { "label": "QNB", "value": 37.0, "after": 14.81 }, { "label": "Bank of Bahrain and Kuwait", "value": 29.6, "after": 3.7 }, { "label": "Saudi National Bank", "value": 14.8, "after": 3.7 }, { "label": "National Bank of Bahrain", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "emirates-nbd", "title": "Emirates NBD", "scope": "Emirates NBD Group; named platform/process scope", "claims": [ { "id": "bank-emirates-nbd-01", "claim": "Emirates NBD describes PCA analysing customer calls into structured service-improvement intelligence.", "url": "https://www.emiratesnbd.com/en/innovation/insights/turning-millions-of-customer-conversations-into-better-banking-experiences-with-ai", "locator": "PCA operating description; English language scope", "level": 3, "eligible": true, "limitation": "Claim scope: Emirates NBD Group; named platform/process scope. English support currently described; Arabic is future. Six million annual calls is addressable volume, not verified processed count. Application: internal-operating-practice.", "rationale": "PCA page explicitly describes implemented call-analysis workflow and structured themes. English currently supported; Arabic described as future." }, { "id": "bank-emirates-nbd-02", "claim": "PCA turns unstructured conversations into recurring themes, customer intent and operational issues shared across teams.", "url": "https://www.emiratesnbd.com/en/innovation/insights/turning-millions-of-customer-conversations-into-better-banking-experiences-with-ai", "locator": "PCA operating description; English language scope", "level": 3, "eligible": true, "limitation": "Claim scope: Emirates NBD Group; named platform/process scope. No retrieval quality or decision trace sample supplied. Application: internal-operating-practice.", "rationale": "PCA converts unstructured calls into reusable themes and intent for operational teams; distinguish shared intelligence from independently validated knowledge quality." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Reconstruct one consequential decision." }, { "companyId": "fab", "title": "First Abu Dhabi Bank", "scope": "FAB Group; individual use-case scope not presumed universal", "claims": [ { "id": "bank-fab-03", "claim": "FAB reports integrating over 90% of structured data into an enterprise platform with an agentic layer.", "url": "https://www.bankfab.com/en-ae/about-fab/group/in-the-media/fab-accelerates-enterprise-ai-adoption", "locator": "AI Innovation Hub; structured data; responsible AI framework", "level": 3, "eligible": true, "limitation": "Claim scope: FAB Group; individual use-case scope not presumed universal. Denominator definition and data quality not disclosed; integration does not prove provenance. Application: internal-operating-practice.", "rationale": "17 February 2026 announcement reports data coverage above 90% and agentic layer. Structured-data denominator and quality not independently checked." }, { "id": "bank-fab-01", "claim": "FAB reports AI agents embedded in departmental routines and over 30 agentic cases progressing in several business functions.", "url": "https://www.bankfab.com/en-ae/about-fab/group/in-the-media/fab-accelerates-enterprise-ai-adoption", "locator": "AI Innovation Hub; structured data; responsible AI framework", "level": 2, "eligible": true, "limitation": "Claim scope: FAB Group; individual use-case scope not presumed universal. Thirty cases are progressing, not thirty verified live cases. Application: internal-operating-practice.", "rationale": "The primary page explicitly places agents in routine departmental operations. Retain level 2; the separate count above 30 denotes progressing cases, not confirmed live deployments." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Reconstruct one consequential decision." }, { "companyId": "nbb", "title": "National Bank of Bahrain", "scope": "NBB; NBB Group including BisB where source explicitly says group", "claims": [ { "id": "bank-nbb-01", "claim": "NBB's ARRAY agreement specifies assessment of its technology footprint and planned NLP/GenAI integration.", "url": "https://nbbonline.com/news/17533/", "locator": "Planned assessment and NLP/GenAI collaboration", "level": null, "eligible": false, "limitation": "Claim scope: NBB; NBB Group including BisB where source explicitly says group. No live AI workflow established by agreement announcement. Application: internal-operating-practice. Exclusion: No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass. Complete readable source support not established; indexed-only evidence cannot score.", "rationale": "Original official URL inaccessible during review; indexed-only passage cannot score under release protocol. A separate directly readable annual report may support another claim. Exclusion: No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass. Complete readable source support not established; indexed-only evidence cannot score." }, { "id": "bank-nbb-02", "claim": "NBB Group launched a seven-day AI and digital academy with four phases and a planned banking-project hackathon.", "url": "https://nbbonline.com/news/21947/", "locator": "Seven-day, four-phase academy description", "level": null, "eligible": false, "limitation": "Claim scope: NBB Group including Bahrain Islamic Bank. Training delivery and final hackathon outcomes not independently evidenced. Application: internal-operating-practice. Exclusion: No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass. Complete readable source support not established; indexed-only evidence cannot score.", "rationale": "Original official URL returns403/access failure; indexed-only academy passage cannot score under release protocol. Replace with independently readable exact claim if found. Exclusion: No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass. Complete readable source support not established; indexed-only evidence cannot score." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Reconstruct one consequential decision." } ], "agenda": [ { "question": "Reconstruct one consequential decision.", "evidence": "Input provenance, model version, accountable decision owner, override and customer recourse." }, { "question": "Separate service speed from decision quality.", "evidence": "Task-specific baseline, exception rates and downstream customer outcomes." }, { "question": "Map the boundary of delegated authority.", "evidence": "Approved actions, access limits, escalation triggers and incident records." } ], "sources": [ { "title": "Turning millions of customer conversations into better banking experiences with AI", "publisher": "Emirates NBD", "date": "2026-07", "url": "https://www.emiratesnbd.com/en/innovation/insights/turning-millions-of-customer-conversations-into-better-banking-experiences-with-ai", "locator": "PCA operating description; English language scope", "originalUrl": "https://www.emiratesnbd.com/en/innovation/insights/turning-millions-of-customer-conversations-into-better-banking-experiences-with-ai", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-operating-case", "id": "S01" }, { "title": "Why banks should build their own Agentic AI Framework", "publisher": "Emirates NBD", "date": "2026-07", "url": "https://www.emiratesnbd.com/en/innovation/insights/why-banks-should-build-their-own-agentic-ai-framework", "locator": "Leap platform controls; enterprise abstraction", "originalUrl": "https://www.emiratesnbd.com/en/innovation/insights/why-banks-should-build-their-own-agentic-ai-framework", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-operating-case", "id": "S02" }, { "title": "People & Capability", "publisher": "Emirates NBD", "date": "Undated", "url": "https://www.emiratesnbd.com/en/innovation/people", "locator": "Learning built into everyday work; learning pathways", "originalUrl": "https://www.emiratesnbd.com/en/innovation/people", "access": "Indexed full page excerpt | Review: full-primary-page-or-document-read", "relationship": "company-program-page", "id": "S03" }, { "title": "Innovation in Action", "publisher": "Emirates NBD", "date": "Undated", "url": "https://www.emiratesnbd.com/en/innovation/innovation-in-action", "locator": "Our AI principles; Our AI journey", "originalUrl": "https://www.emiratesnbd.com/en/innovation/innovation-in-action", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-policy-and-program-page", "id": "S04" }, { "title": "Annual Report 2025", "publisher": "Emirates NBD", "date": "2026", "url": "https://cdn.emiratesnbd.com/en/assets/file/ir/annual_report_2025.pdf", "locator": "PDF spread p.32 (zero-based31), printed pp.58–59, Project Trinity", "originalUrl": "https://cdn.emiratesnbd.com/en/assets/file/ir/annual_report_2025.pdf", "access": "Full PDF text retrieved in second pass; 132 PDF pages/spreads. Technology, HR, compliance and Project Trinity AI passages reviewed. | Review: full-primary-page-or-document-read", "relationship": "company-annual-report", "id": "S05" }, { "title": "FAB accelerates enterprise AI adoption through group-wide innovation and talent enablement", "publisher": "First Abu Dhabi Bank", "date": "2026-02-17", "url": "https://www.bankfab.com/en-ae/about-fab/group/in-the-media/fab-accelerates-enterprise-ai-adoption", "locator": "AI Innovation Hub; structured data; responsible AI framework", "originalUrl": "https://www.bankfab.com/en-ae/about-fab/group/in-the-media/fab-accelerates-enterprise-ai-adoption", "access": "Full indexed page text | Review: full-source-read", "relationship": "company-announcement", "id": "S06" }, { "title": "AI in Banking: Driving Innovation & Growth", "publisher": "First Abu Dhabi Bank", "date": "Undated", "url": "https://www.bankfab.com/en-ae/about-fab/innovation-and-ai", "locator": "Practical applications; 6 October 2026 DANA case summary", "originalUrl": "https://www.bankfab.com/en-ae/about-fab/innovation-and-ai", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-program-page", "id": "S07" }, { "title": "KFH Reinforces Banking Leadership with Forward-Looking AI Initiatives", "publisher": "Kuwait Finance House", "date": "2025-08-30", "url": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Reinforces-Banking-Leadership-with-Forward-Looking-AI-Initiatives.html", "locator": "Specialized virtual assistant; integrating human expertise", "originalUrl": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Reinforces-Banking-Leadership-with-Forward-Looking-AI-Initiatives.html", "access": "Full indexed article; direct rendered retrieval omitted body | Review: full-primary-page-or-document-read", "relationship": "company-announcement", "id": "S08" }, { "title": "KFH Adopts Advanced AI Technologies to Enhance Customer Experience", "publisher": "Kuwait Finance House", "date": "2025-07-27", "url": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Adopts-Advanced-AI-Technologies-to-Enhance-Customer-Experience.html", "locator": "Speech Analytics; Automated Quality Management", "originalUrl": "https://www.kfh.com/en/home/Personal/news/2025/KFH-Adopts-Advanced-AI-Technologies-to-Enhance-Customer-Experience.html", "access": "Full indexed article | Review: full-primary-page-or-document-read", "relationship": "company-operating-case", "id": "S09" }, { "title": "KFH Sustainability Report 2025", "publisher": "Kuwait Finance House", "date": "2026", "url": "https://www.kfh.com/dam/jcr:6d21db7a-84cb-4487-b3f2-7bcf64d4c382/KFH%2BSustainability%2BReport%2B2025.pdf", "locator": "Printed p.95; PDF p.94 (zero-based93)", "originalUrl": "https://www.kfh.com/dam/jcr:6d21db7a-84cb-4487-b3f2-7bcf64d4c382/KFH%2BSustainability%2BReport%2B2025.pdf", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-sustainability-report", "id": "S10" }, { "title": "Commercial Bank Annual Report 2025", "publisher": "Commercial Bank of Qatar", "date": "2026", "url": "https://www.cbq.qa/-/media/project/cbq/cbqwebsite/documents/corporate-affairs/annual-reports-en/cbq-annual-report-2025--en.pdf", "locator": "PDF p.37, Data & Artificial Intelligence", "originalUrl": "https://www.cbq.qa/-/media/project/cbq/cbqwebsite/documents/corporate-affairs/annual-reports-en/cbq-annual-report-2025--en.pdf", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-annual-report", "id": "S11" }, { "title": "Investor Presentation FY 2025", "publisher": "Commercial Bank of Qatar", "date": "2026", "url": "https://www.cbq.qa/-/media/project/cbq/cbqwebsite/documents/financial-reports-en/2025/q4/investor-presentation---fy-2025.pdf", "locator": "Strategy slide 9, Accelerating our AI Journey", "originalUrl": "https://cbq.qa/-/media/project/cbq/cbqwebsite/documents/financial-reports-en/2025/q4/investor-presentation---fy-2025.pdf", "access": "Full indexed slide text | Review: full-primary-page-or-document-read | Readable corroboration: https://www.cbq.qa/-/media/project/cbq/cbqwebsite/documents/financial-reports-en/2025/q4/investor-presentation---fy-2025.pdf", "relationship": "company-investor-presentation", "id": "S12" }, { "title": "Commercial Bank Awarded Most Innovative Use of AI Technology by International Finance", "publisher": "Commercial Bank of Qatar", "date": "2025-07-03", "url": "https://www.cbq.com.qa/en/about-us/news/list/commercial-bank-awarded-most-innovative-use-of-ai-technology-by-international-finance", "locator": "Data Strategy and internal teams paragraphs", "originalUrl": "https://www.cbq.com.qa/en/about-us/news/list/commercial-bank-awarded-most-innovative-use-of-ai-technology-by-international-finance", "access": "Full indexed article; award itself excluded from coding | Review: full-primary-page-or-document-read", "relationship": "company-announcement", "id": "S13" }, { "title": "NBK Launches Microsoft Copilot Across All Divisions", "publisher": "National Bank of Kuwait", "date": "2025-07-27", "url": "https://www.nbk.com/kuwait/news-and-insights/media-relations/news.html?news=nbk-launches-microsoft-copilot-across-all-divisions--pioneering-ai-integration-and-empowering-employees-", "locator": "Rollout and employee workspace paragraphs", "originalUrl": "https://www.nbk.com/kuwait/news-and-insights/media-relations/news.html?news=nbk-launches-microsoft-copilot-across-all-divisions--pioneering-ai-integration-and-empowering-employees-", "access": "Full indexed article | Review: full-primary-page-or-document-read", "relationship": "company-announcement", "id": "S14" }, { "title": "NBK Organizes a Series of AI Mindset and Awareness Masterclasses", "publisher": "National Bank of Kuwait", "date": "2025-09-14", "url": "https://www.nbk.com/kuwait/news-and-insights/media-relations/news.html?month=09&year=2025", "locator": "Named article under September 2025 archive", "originalUrl": "https://www.nbk.com/kuwait/news-and-insights/media-relations/news.html?month=09&year=2025", "access": "Full indexed article; archive URL contains multiple articles", "relationship": "company-announcement", "id": "S15" }, { "title": "NBK Closes Exceptional 2025 Marked by Unprecedented Achievements", "publisher": "National Bank of Kuwait", "date": "2025-12-25", "url": "https://www.nbk.com/kuwait/news-and-insights/media-relations/news.html?news=nbk-closes-exceptional-2025-marked-by-unprecedented-achievements", "locator": "AI Mindset Curriculum paragraph", "originalUrl": "https://www.nbk.com/kuwait/news-and-insights/media-relations/news.html?news=nbk-closes-exceptional-2025-marked-by-unprecedented-achievements", "access": "Full indexed article", "relationship": "company-announcement", "id": "S16" }, { "title": "NBK Annual Report 2025", "publisher": "National Bank of Kuwait", "date": "2026-02-22", "url": "https://www.nbk.com/dam/jcr:73b1509e-5280-4ef3-a29a-392c1b565299/nbk-annual-report-2025-e.pdf", "locator": "Printed p.37; PDF p.38 (zero-based37)", "originalUrl": "https://www.nbk.com/dam/jcr:73b1509e-5280-4ef3-a29a-392c1b565299/nbk-annual-report-2025-e.pdf", "access": "Full page or document text retrieved", "relationship": "company-annual-report", "id": "S17" }, { "title": "NBK Sustainability Report 2025", "publisher": "National Bank of Kuwait", "date": "2026-04-30", "url": "https://argaamplus.s3.amazonaws.com/4b4e6fe7-09dd-4311-be33-c97735b392e7.pdf", "locator": "Printed p.192, Responsible AI & Data Governance; mirror PDF zero-based96", "originalUrl": "https://www.nbk.com/dam/jcr%3Ae9a25c2e-890b-4ce2-b3ee-0f9fa4c3999f/nbk-sustainability-report-2025.pdf", "access": "Official indexed report and issuer-authored full PDF mirror read; official full PDF retrieval failed (size). Mirror https://argaamplus.s3.amazonaws.com/4b4e6fe7-09dd-4311-be33-c97735b392e7.pdf | Review: full-company-authored-mirror-read-not-hashmatched | Readable corroboration: https://argaamplus.s3.amazonaws.com/4b4e6fe7-09dd-4311-be33-c97735b392e7.pdf", "relationship": "company-sustainability-report", "id": "S18" }, { "title": "Integrated Annual Report 2025: Innovation and product development", "publisher": "Al Rajhi Bank", "date": "2026", "url": "https://www.alrajhibank.com.sa/ir25/customer_engagement_and_responsible_banking/innovation_and_product_development.html", "locator": "Group-wide insights and real-time marketing", "originalUrl": "https://www.alrajhibank.com.sa/ir25/customer_engagement_and_responsible_banking/innovation_and_product_development.html", "access": "Full indexed page text | Review: full-primary-page-or-document-read", "relationship": "company-annual-report-html", "id": "S19" }, { "title": "Integrated Annual Report 2024: Learning and development at alrajhi group", "publisher": "Al Rajhi Bank", "date": "2025", "url": "https://www.alrajhibank.com.sa/ir24/cultivating_a_future_forward_workforce/learning_and_development_at_alrajhi_group.html", "locator": "METAKKED paragraph", "originalUrl": "https://www.alrajhibank.com.sa/ir24/cultivating_a_future_forward_workforce/learning_and_development_at_alrajhi_group.html", "access": "Indexed relevant passage | Review: full-primary-page-or-document-read", "relationship": "company-annual-report-html", "id": "S20" }, { "title": "Integrated Annual Report 2025: Risk management", "publisher": "Al Rajhi Bank", "date": "2026", "url": "https://www.alrajhibank.com.sa/ir25/corporate_governance/risk_management.html", "locator": "Cybersecurity risk; Information Technology Risk", "originalUrl": "https://www.alrajhibank.com.sa/ir25/corporate_governance/risk_management.html", "access": "Full indexed page text", "relationship": "company-annual-report-html", "id": "S21" }, { "title": "Al Rajhi Integrated Annual Report 2025 — Learning and Development", "publisher": "Al Rajhi Bank", "date": "2026", "url": "https://www.alrajhibank.com.sa/ir25/cultivating_a_future_forward_workforce/learning_and_development.html", "locator": "Digital learning platforms, METAKKED", "originalUrl": "https://www.alrajhibank.com.sa/ir25/cultivating_a_future_forward_workforce/learning_and_development.html", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-annual-report", "id": "S22" }, { "title": "Al Rajhi Integrated Annual Report 2025 — Data privacy and information management", "publisher": "Al Rajhi Bank", "date": "2026", "url": "https://www.alrajhibank.com.sa/ir25/corporate_governance/data_privacy_and_information_management.html", "locator": "Opening section, paragraph beginning Both internal and external privacy policies", "originalUrl": "https://www.alrajhibank.com.sa/ir25/corporate_governance/data_privacy_and_information_management.html", "access": "Full page or document text retrieved", "relationship": "company-annual-report", "id": "S23" }, { "title": "Al Rajhi Integrated Annual Report 2025 — Shared services group", "publisher": "Al Rajhi Bank", "date": "2026", "url": "https://www.alrajhibank.com.sa/ir25/business_in_perspective/shared_services_group.html", "locator": "Strategic initiative 04: Building a world-class data platform", "originalUrl": "https://www.alrajhibank.com.sa/ir25/business_in_perspective/shared_services_group.html", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-annual-report", "id": "S24" }, { "title": "QNB Group AI & Data Governance Statement", "publisher": "QNB", "date": "2026", "url": "https://www.qnb.com/sites/qnb/qnbqatar/document/en/DataAIGovernanceStatement-", "locator": "PDF p.4 (printed p.3), Governance Approach", "originalUrl": "https://www.qnb.com/sites/qnb/qnbqatar/document/en/DataAIGovernanceStatement-", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-policy", "id": "S25" }, { "title": "QNB Annual Report 2025", "publisher": "QNB", "date": "2026", "url": "https://vpr.hkma.gov.hk/statics/assets/doc/100336/ar_25/ar_25_eng.pdf", "locator": "PDF p.31 (zero-based30), Falcon", "originalUrl": "https://vpr.hkma.gov.hk/statics/assets/doc/100336/ar_25/ar_25_eng.pdf", "access": "Bank-authored annual report on Hong Kong Monetary Authority register", "relationship": "company-annual-report", "id": "S26" }, { "title": "QNB Group Maintains Disciplined Execution Across Key Strategic Priorities in Q2", "publisher": "QNB", "date": "2026-07-22", "url": "https://www.qnb.com/sites/qnb/qnbglobal/en/ennew22072026news", "locator": "QNB Center for Business Innovation", "originalUrl": "https://qnb.com/sites/qnb/qnbglobal/en/ennew22072026news", "access": "Full indexed page text | Review: full-primary-page-or-document-read | Readable corroboration: https://www.qnb.com/sites/qnb/qnbglobal/en/ennew22072026news", "relationship": "company-announcement", "id": "S27" }, { "title": "BBK partners with ARRAY to launch Innovation Center", "publisher": "Bank of Bahrain and Kuwait", "date": "2025-11-05", "url": "https://www.bbkonline.com/bbk-partners-with-array-to-launch-innovation-center-hiring-bahraini-software-engineers-to-accelerate-digital-transformation/", "locator": "Center launch; 30 engineers; planned AI solutions", "originalUrl": "https://www.bbkonline.com/bbk-partners-with-array-to-launch-innovation-center-hiring-bahraini-software-engineers-to-accelerate-digital-transformation/", "access": "Full page or document text retrieved | Review: full-primary-page-or-document-read", "relationship": "company-announcement", "id": "S28" }, { "title": "BBK Annual Report 2025", "publisher": "Bank of Bahrain and Kuwait", "date": "2026-04", "url": "https://www.bbkonline.com/wp-content/uploads/2026/04/BBK-2025-Annual-Report-English-Spread.pdf", "locator": "Printed p.29; PDF spread p.16 (zero-based15)", "originalUrl": "https://www.bbkonline.com/wp-content/uploads/2026/04/BBK-2025-Annual-Report-English-Spread.pdf", "access": "Full page or document text retrieved", "relationship": "company-annual-report", "id": "S29" }, { "title": "BBK launches its Digital Management Trainee Program in FinTech", "publisher": "Bank of Bahrain and Kuwait", "date": "2022-07", "url": "https://www.bbkonline.com/bbk-launches-its-digital-management-trainee-program-in-fintech/", "locator": "Two-phase programme and AI curriculum", "originalUrl": "https://www.bbkonline.com/bbk-launches-its-digital-management-trainee-program-in-fintech/", "access": "Indexed article; historic context only", "relationship": "company-program-page", "id": "S30" }, { "title": "Saudi National Bank Annual Report 2025", "publisher": "Saudi National Bank", "date": "2026", "url": "https://argaamplus.s3.amazonaws.com/07ce1455-a38a-41b1-ad0e-78b4c0e01090.pdf", "locator": "PDF spread p.19 (zero-based18), printed pp.36–37; Retail Banking", "originalUrl": "https://argaamplus.s3.amazonaws.com/07ce1455-a38a-41b1-ad0e-78b4c0e01090.pdf", "access": "Bank-authored report on Argaam mirror; issuer PDF timed out. Mirror provenance retained, not independently hash-matched. | Review: full-company-authored-mirror-read-not-hashmatched", "relationship": "company-annual-report", "id": "S31" }, { "title": "SNB Technology Program", "publisher": "Saudi National Bank", "date": "Undated", "url": "https://www.alahli.com/en/pages/about-us/careers/snb-technology-program", "locator": "About the Program; Program Duration", "originalUrl": "https://www.alahli.com/en/pages/about-us/careers/snb-technology-program", "access": "Full page or document text retrieved", "relationship": "company-program-page", "id": "S32" }, { "title": "NBB Partners with ARRAY Innovation to Drive AI-Driven Efficiency and Customer Service Excellence", "publisher": "National Bank of Bahrain", "date": "2024-11-03", "url": "https://nbbonline.com/news/17533/", "locator": "Planned assessment and NLP/GenAI collaboration", "originalUrl": "https://nbbonline.com/news/17533/", "access": "Full indexed article; direct retrieval 403 | Review: inaccessible-indexed-only", "relationship": "company-announcement", "id": "S33" }, { "title": "With Support from Tamkeen NBB Group Unveils New AI and Digital Banking Academy", "publisher": "National Bank of Bahrain", "date": "2025-10-25", "url": "https://nbbonline.com/news/21947/", "locator": "Seven-day, four-phase academy description", "originalUrl": "https://nbbonline.com/news/21947/", "access": "Full indexed article; direct domain retrieval restricted | Review: inaccessible-indexed-only", "relationship": "company-announcement", "id": "S34" }, { "title": "NBB Group Hosts Gartner Training Session on AI for Group Board of Directors and Executive Management", "publisher": "National Bank of Bahrain", "date": "2024-07-06", "url": "https://nbbonline.com/news/14234/", "locator": "Board/executive training paragraph", "originalUrl": "https://nbbonline.com/news/14234/", "access": "Full indexed article | Review: inaccessible-indexed-only", "relationship": "company-announcement", "id": "S35" }, { "title": "NBB Annual Financial and Sustainability Report 2025", "publisher": "National Bank of Bahrain", "date": "2026-03-03", "url": "https://nbbonline.com/wp-content/uploads/2026/03/Annual-Financial-and-Sustainability-Report-2025-3.pdf", "locator": "Printed pp.68–69, NBB Group AI & Digital Banking Academy", "originalUrl": "https://nbbonline.com/wp-content/uploads/2026/03/Annual-Financial-and-Sustainability-Report-2025-3.pdf", "access": "Official indexed PDF passages retrieved; direct full PDF failed. Report date corroborated by official investor index. | Review: inaccessible-indexed-only", "relationship": "company-annual-report", "id": "S36" }, { "title": "NBB records a 4% growth in net profit attributable to shareholders for year ended 31 December 2025", "publisher": "National Bank of Bahrain", "date": "2026-02-17", "url": "https://nbbonline.com/news/27971/", "locator": "Group CEO commentary", "originalUrl": "https://nbbonline.com/news/27971/", "access": "Full official indexed release read; direct page fetch failed. | Review: inaccessible-indexed-only", "relationship": "company-results-announcement", "id": "S37" }, { "title": "Emirates NBD announcement", "publisher": "Primary source as linked", "date": "See original", "url": "https://www.emiratesnbd.com/en/media-center/emirates-nbd-ranked-1-in-inaugural-evident-ai-index-for-banks", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S38" }, { "title": "Evident regional index", "publisher": "Primary source as linked", "date": "See original", "url": "https://evidentinsights.com/banking-ai-index-mea", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S39" } ], "body": "# The AI-native bank: follow the decisions, not the announcements\n## Mayden AI Native Index — banking evidence briefing\nResearch cut-off: 10 October 2026. Public-evidence research edition, not a verified maturity assessment.\n\n### A bank becomes more AI-native when its operating system changes\n\nThe most useful question for a bank's board is no longer whether the institution uses artificial intelligence. It is whether AI is changing how the bank gathers knowledge, reaches decisions, delivers work and corrects mistakes. A collection of disconnected assistants can coexist with unchanged approval chains, inaccessible knowledge and weak measurement. Conversely, a tightly governed application in a critical workflow may reveal more organisational change than a large collection of announced experiments.\n\nThis briefing examines a deliberately selected cohort of ten banks: Saudi National Bank and Al Rajhi Bank in Saudi Arabia; Emirates NBD and First Abu Dhabi Bank in the UAE; QNB and Commercial Bank in Qatar; National Bank of Kuwait and Kuwait Finance House in Kuwait; and National Bank of Bahrain and BBK in Bahrain. Two organisations per market provide a starting comparison, not a representative sample. Oman is outside this edition. Group-level disclosures describe international operations as well as home markets and cannot be interpreted as country-level adoption rates.\n\nThe research separates reported implementation, future plans and unknowns. The associated register contains sources and bounded claims against nine Mayden themes. It measures what public evidence supports. It does not establish which bank is objectively the most AI-native.\n\n### The clearest operating signals come from specific workflows\n\nCommercial Bank of Qatar's 2025 report names deployed document recognition in onboarding, trade finance and lending, together with a retrieval-supported email assistant. It also describes its own AI computing environment. These disclosures connect technical capability to work rather than simply announcing a partner. The same report presents agentic capabilities as a future direction; those must remain separate from live tools. [Commercial Bank Annual Report 2025, p.37](https://www.cbq.qa/-/media/project/cbq/cbqwebsite/documents/corporate-affairs/annual-reports-en/cbq-annual-report-2025--en.pdf).\n\nSaudi National Bank describes AI-assisted segmentation and retail cross-selling. Its subsidiary SNB Capital also reports AI deployment, but that evidence belongs to the subsidiary's asset-management functions and must not silently become a bank-wide claim. [SNB Annual Report 2025, printed pp.36–39; company-authored report on Argaam mirror](https://argaamplus.s3.amazonaws.com/07ce1455-a38a-41b1-ad0e-78b4c0e01090.pdf).\n\nAl Rajhi's report describes predictive models inside credit, fraud and customer-lifecycle decisions. It also explains the assembly of consented information from banking and external channels. The methodological distinction matters: the relevant signal is the named decision mechanism, not the accompanying digital-banking awards. [Al Rajhi 2025 innovation review](https://www.alrajhibank.com.sa/ir25/customer_engagement_and_responsible_banking/innovation_and_product_development.html).\n\nFor an executive, the follow-up is straightforward: select one material workflow and identify precisely which decisions AI supports, what evidence it uses, who can override it, and how success is measured.\n\n### Reusable infrastructure may be the organisational turning point\n\nEmirates NBD describes Leap as a shared agent engineering framework supplying controls such as authentication, audit records and evaluation. This is a potentially important organisational pattern: common infrastructure can reduce repeated work and make rules portable across teams. The disclosure is an architectural claim by the bank, not an independent control audit. [Emirates NBD's Leap account](https://www.emiratesnbd.com/en/innovation/insights/why-banks-should-build-their-own-agentic-ai-framework).\n\nFAB describes an AI Innovation Hub, a common data platform and a group responsible-AI framework. Its February 2026 announcement distinguishes agents already embedded in operations from a larger portfolio of cases still progressing. That difference must survive any chart, headline or ranking. [FAB enterprise AI announcement](https://www.bankfab.com/en-ae/about-fab/group/in-the-media/fab-accelerates-enterprise-ai-adoption).\n\nThese cases suggest a proposition for field research: reusable organisational capabilities may explain sustainable deployment better than the number of pilots. This proposition remains a question for field research. Relevant evidence would include time to release a second use case, reuse of controls, duplicated engineering effort, incident patterns and the cost of changing an underlying model.\n\n### Knowledge must become usable without losing its authority\n\nKuwait Finance House describes employee assistants including ShariaBot, which refers to specific fatwas issued by its supervisory board, and ComplianceBot. This provides a concrete example of institutional knowledge entering daily work. It raises a deeper assessment question: can the institution demonstrate that an answer uses the current authorised source and that later changes propagate correctly? A source reference is useful, but it does not establish a complete audit trail. [KFH's internal AI initiatives](https://www.kfh.com/en/home/Personal/news/2025/KFH-Reinforces-Banking-Leadership-with-Forward-Looking-AI-Initiatives.html).\n\nCustomer conversations are another knowledge source. Emirates NBD's PCA account describes turning calls into structured signals for service improvement. Its disclosed language boundary is especially useful for regional research: English is supported, while Arabic is described as a future extension. The Index should retain language coverage and avoid translating an English deployment into an implied bilingual capability. [Emirates NBD PCA case](https://www.emiratesnbd.com/en/innovation/insights/turning-millions-of-customer-conversations-into-better-banking-experiences-with-ai).\n\nKFH separately describes automated call analysis and quality assessment. Mayden's diagnostic should ask how such evaluation affects employees, how incorrect assessments are challenged, and how quality scores are validated across language and customer groups. [KFH contact-centre case](https://www.kfh.com/en/home/Personal/news/2025/KFH-Adopts-Advanced-AI-Technologies-to-Enhance-Customer-Experience.html).\n\n### Workforce access is an input; changed work is the outcome to investigate\n\nNBK's sustainability report makes a useful distinction between tool availability and actual use: about one-third of staff were consistent Copilot Web users in 2025. The bank also discloses restrictions on confidential information and disables document transfer. Its broader AI risk and accountability framework remains a roadmap item. These specifics are more informative than an access announcement, while still leaving productivity and behavioural change untested. The report's default scope is Kuwait operations. [NBK Sustainability Report 2025, printed p.192; company-authored report on Argaam mirror](https://argaamplus.s3.amazonaws.com/4b4e6fe7-09dd-4311-be33-c97735b392e7.pdf).\n\nIndexed passages from NBB's annual report and results release point to completed academy cohorts and a service chatbot. Full direct retrieval could not be completed under this edition's verification protocol, so these remain unresolved research leads and are excluded from release scoring. They should be revisited before drawing conclusions about NBB's operating practices. [NBB Annual Report 2025, pp.68–69](https://nbbonline.com/wp-content/uploads/2026/03/Annual-Financial-and-Sustainability-Report-2025-3.pdf), [NBB 2025 results](https://nbbonline.com/news/27971/).\n\nBBK offers a different combination: an innovation centre with a stated engineering team and annual-report disclosures about AI in board documentation and management oversight. A narrow board tool may be operationally meaningful even when enterprise coverage is unknown. [BBK innovation centre](https://www.bbkonline.com/bbk-partners-with-array-to-launch-innovation-center-hiring-bahraini-software-engineers-to-accelerate-digital-transformation/), [BBK Annual Report 2025](https://www.bbkonline.com/wp-content/uploads/2026/04/BBK-2025-Annual-Report-English-Spread.pdf).\n\n### Governance should be visible in decisions\n\nQNB's published AI and Data Governance Statement names an executive AI committee and a cross-functional review forum for material use cases. This gives researchers a concrete place to examine responsibility. The next evidential step is to inspect what those bodies approve, reject, monitor and stop. Their existence should not be equated with their effectiveness. [QNB governance statement, printed p.3](https://www.qnb.com/sites/qnb/qnbqatar/document/en/DataAIGovernanceStatement-).\n\nAl Rajhi explicitly says that its formal AI governance framework is still being developed. That candour should remain visible alongside evidence of deployed predictive models. QNB, meanwhile, reports beginning to use Copilot within internal audit and including AI ethics in mandatory risk training. A useful profile preserves these different stages rather than forcing every institution into a single transformation story. [Al Rajhi data privacy review](https://www.alrajhibank.com.sa/ir25/corporate_governance/data_privacy_and_information_management.html), [QNB Annual Report 2025, PDF pp.83 and120](https://vpr.hkma.gov.hk/statics/assets/doc/100336/ar_25/ar_25_eng.pdf).\n\nAcross the cohort, the useful diagnostic is a decision reconstruction. Choose a consequential AI-supported decision and trace its information, model, approval, challenge and correction. The exercise can reveal whether governance is part of operations or remains an overlay. Greater autonomy should earn credit only where delegated authority, limits, escalation and recovery are demonstrable. More autonomy is not automatically better banking.\n\n### The evidence edition must be honest about what its ranking means\n\nA normalized public-evidence score can help readers compare documented practices against the same rubric. It cannot remove disclosure bias. A bank that publishes a detailed technology account may receive more evidence credit than an equally capable bank that publishes little. Missing material should be labelled unknown, not operational failure. Multiple announcements about the same programme should not accumulate extra points.\n\nThis review is predominantly English, uses company-authored disclosures and was conducted within a bounded search. FAB's annual PDF could not be retrieved; NBB's indexed-only annual and results leads are excluded from scoring. SNB and NBK sustainability text required company-authored reports on third-party mirrors; those copies were not hash-matched to issuer files. These unequal access conditions limit comparison. The database records those limits. No respondent interviews, system tests or independent operating audits were performed. Reported business improvements also need their own ledger: baseline, scope, period, denominator and attribution. They should remain outside capability scoring so that Mayden can later investigate the relationship between capability and results.\n\n### A benchmark worth returning to\n\nBank AI benchmarking already has established competitors. Emirates NBD's June 2026 announcement links to the Evident Middle East and Africa bank index. Regional bank AI measurement already exists; this edition neither claims priority nor imports another provider's rank into its own score. [Emirates NBD announcement](https://www.emiratesnbd.com/en/media-center/emirates-nbd-ranked-1-in-inaugural-evident-ai-index-for-banks), [Evident regional index](https://evidentinsights.com/banking-ai-index-mea).\n\nMayden's opportunity is to connect public evidence to a repeatable company diagnostic and longitudinal research. Executives should leave with a sharper picture of where their operating model works, where responsibility is unclear, and which intervention to test next. The benchmark becomes durable when the same organisations return, their evidence improves, and the study can distinguish enduring organisational change from an attractive announcement.\n\n**Executive reflection:** Could your bank reconstruct one important AI-supported decision, show who remains accountable, and demonstrate an improvement after the cost of checking and correcting the output?\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/banking-public-evidence.pdf", "cover": "/research/index/2026-10/banking-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 103258, "pdfSha256": "4f8394ee2773aea744c00963567b4914d0f8ed6818eed0c54e0a25d9375fe9a7", "pdfPages": 18 }, { "slug": "telecom-public-evidence", "number": 3, "title": "AI in Gulf telecoms: the next test is how the business operates", "shortTitle": "The AI-enabled telecom operator", "dek": "Separate internal AI practice, customer offerings and future infrastructure across ten telecom entities.", "sector": "Telecommunications", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Test what the customer can complete. Batelco describes Basma connecting billing, device searches and fraud-reporting support to backend systems. Examine completed cases, unresolved requests and escalation to people.", "Check how far governance reaches. e& reports central review, owners and risk thresholds while wider platform rollout remains a priority. Request operating-unit coverage and examples of challenged or stopped use cases.", "Map autonomy to permission. Zain reports closed-loop orchestration; e& UAE describes a roadmap toward greater network autonomy. Establish what each deployed system may change and when an operator intervenes." ], "executiveSources": [ { "claimIds": [ "TEL-031" ], "sources": [ { "claimId": "TEL-031", "url": "https://beyon.com/wp-content/uploads/2026/03/Beyon_ESG2025_English.pdf", "sourceTitle": "Beyon Sustainability Report 2025" } ] }, { "claimIds": [ "TEL-009", "TEL-010" ], "sources": [ { "claimId": "TEL-009", "url": "https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf", "sourceTitle": "e& Integrated Annual Report 2025" }, { "claimId": "TEL-010", "url": "https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf", "sourceTitle": "e& Integrated Annual Report 2025" } ] }, { "claimIds": [ "TEL-025", "TEL-012" ], "sources": [ { "claimId": "TEL-025", "url": "https://zain.com/AR2025/technology/", "sourceTitle": "Zain Annual Report 2025: Technology" }, { "claimId": "TEL-012", "url": "https://www.eand.com/en/news/22-06-26-eand-tm-strategic-blueprint.html", "sourceTitle": "e& and TM Forum announce strategic blueprint for Autonomous Networks" } ] } ], "evidenceAgeContext": "Dated sources run from 2025-02-10 to 2026-06-22. 22 of 29 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "Locate the operator’s control boundary", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 42, "label": "recorded claims" }, { "value": 29, "label": "eligible claims" }, { "value": 1, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "Locate the operator’s control boundary", "subtitle": "Network, customer-service and security cases have different permissions and fallback requirements.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "Network assurance", "unit": "Fixed-wireless network performance", "finding": "stc Bahrain reports AI assurance deployed particularly for fixed wireless access.", "boundary": "Bahrain operation; award account is not an operating audit.", "implication": "Which faults can it correct, and which require an engineer?", "claimIds": [ "TEL-034" ], "sources": [ { "claimId": "TEL-034", "url": "https://www.stc.com.bh/media-center/template?id=4346", "sourceTitle": "stc Bahrain wins awards for AI-powered CX assurance and convergent billing" } ] }, { "label": "Customer interaction", "unit": "Billing or service request", "finding": "Batelco says Basma connects to backend systems for billing, device searches and fraud-reporting support.", "boundary": "Batelco Bahrain; completion and escalation rates undisclosed.", "implication": "Does the interaction finish the request or transfer unfinished work?", "claimIds": [ "TEL-031" ], "sources": [ { "claimId": "TEL-031", "url": "https://beyon.com/wp-content/uploads/2026/03/Beyon_ESG2025_English.pdf", "sourceTitle": "Beyon Sustainability Report 2025" } ] }, { "label": "Security investigation", "unit": "Managed-security alert", "finding": "Beyon Cyber describes Oryx AI investigating alerts through live queries with human oversight.", "boundary": "Managed-security service; not proof of Batelco-wide adoption.", "implication": "What action can follow an investigation, and who authorises it?", "claimIds": [ "TEL-033" ], "sources": [ { "claimId": "TEL-033", "url": "https://beyon.com/wp-content/uploads/2026/03/Beyon-AR2025-ENGLISH.pdf", "sourceTitle": "Beyon Annual Report 2025" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "eand-group", "label": "e& group", "values": [ 3, 3, null, 2, 2, null, 1, null, 3 ], "score": 52, "rank": 1 }, { "id": "beyon", "label": "Beyon / Batelco", "values": [ 2, null, null, 3, null, null, 3, null, null ], "score": 30, "rank": 2 }, { "id": "ooredoo-group", "label": "Ooredoo group", "values": [ 2, 2, null, 2, null, null, null, 2, null ], "score": 30, "rank": 2 }, { "id": "stc-group", "label": "stc group", "values": [ 2, 2, null, null, null, null, null, 1, 3 ], "score": 30, "rank": 2 }, { "id": "stc-kuwait", "label": "stc Kuwait", "values": [ null, 2, null, null, null, 1, null, null, 2 ], "score": 19, "rank": 5 }, { "id": "stc-bahrain", "label": "stc Bahrain", "values": [ 3, null, 1, null, null, null, null, null, null ], "score": 15, "rank": 6 }, { "id": "zain-group", "label": "Zain group", "values": [ null, null, null, null, null, 1, 2, 1, null ], "score": 15, "rank": 6 }, { "id": "du", "label": "du", "values": [ null, 1, 1, 1, null, null, null, null, null ], "score": 11, "rank": 8 }, { "id": "mobily", "label": "Mobily", "values": [ null, null, 1, null, null, null, null, null, null ], "score": 4, "rank": 9 }, { "id": "vodafone-qatar", "label": "Vodafone Qatar", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "e& group", "value": 52, "secondary": 6, "rank": 1 }, { "label": "Beyon / Batelco", "value": 30, "secondary": 3, "rank": 2 }, { "label": "Ooredoo group", "value": 30, "secondary": 4, "rank": 2 }, { "label": "stc group", "value": 30, "secondary": 4, "rank": 2 }, { "label": "stc Kuwait", "value": 19, "secondary": 3, "rank": 5 }, { "label": "stc Bahrain", "value": 15, "secondary": 2, "rank": 6 }, { "label": "Zain group", "value": 15, "secondary": 3, "rank": 6 }, { "label": "du", "value": 11, "secondary": 3, "rank": 8 }, { "label": "Mobily", "value": 4, "secondary": 1, "rank": 9 }, { "label": "Vodafone Qatar", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 29; one practice may support multiple distinct themes", "maximum": 12, "rows": [ { "label": "Specific commitment / pilot", "value": 10 }, { "label": "Implemented practice", "value": 12 }, { "label": "Implemented + operating detail", "value": 7 } ], "note": "Source: claim ledger. 13 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "e& group", "value": 51.9, "after": 3.7 }, { "label": "Beyon / Batelco", "value": 29.6, "after": 11.11 }, { "label": "Ooredoo group", "value": 29.6, "after": 0.0 }, { "label": "stc group", "value": 29.6, "after": 7.41 }, { "label": "stc Kuwait", "value": 18.5, "after": 0.0 }, { "label": "stc Bahrain", "value": 14.8, "after": 3.7 }, { "label": "Zain group", "value": 14.8, "after": 0.0 }, { "label": "du", "value": 11.1, "after": 3.7 }, { "label": "Mobily", "value": 3.7, "after": 0.0 }, { "label": "Vodafone Qatar", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "eand-group", "title": "e& group", "scope": "Emirates Telecommunications Group Company PJSC (e&), consolidated group", "claims": [ { "id": "TEL-009", "claim": "e& describes central AI use-case review by a steering committee with use-case owners and risk thresholds.", "url": "https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf", "locator": "Printed p.128 / PDF page index129, Responsible AI", "level": 3, "eligible": true, "limitation": "Claim scope: Emirates Telecommunications Group Company PJSC (e&), consolidated group. Report also lists full governance-platform rollout across all units as a 2026 priority; not proof of universal implementation. Application: internal.", "rationale": "PDF p129 specifies operational committee review and risk criteria. Group claim; incomplete rollout remains possible." }, { "id": "TEL-011", "claim": "e& launched mandatory responsible-AI training in September 2025 and reports Citizen X employee model-building activity.", "url": "https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf", "locator": "Printed p.128 / PDF page index129, Embedding responsible AI across workforce", "level": 3, "eligible": true, "limitation": "Claim scope: Emirates Telecommunications Group Company PJSC (e&), consolidated group. Training mandate is not completion; employee-built models are not automatically production deployments. Application: internal.", "rationale": "p129 names launched training and employee model-building coverage. Not a tested competence result." } ], "boundary": "Evidence may concern e& UAE or other subsidiaries; group evidence cannot describe UAE enterprises generally.", "question": "Locate AI inside the operator." }, { "companyId": "beyon", "title": "Beyon / Batelco", "scope": "Beyon B.S.C., consolidated group including Batelco by Beyon", "claims": [ { "id": "TEL-031", "claim": "Batelco expanded Basma to handle billing, device searches and fraud-reporting support through backend integration.", "url": "https://beyon.com/wp-content/uploads/2026/03/Beyon_ESG2025_English.pdf", "locator": "Printed pp.58-59 / PDF page index30, Batelco by Beyon", "level": 3, "eligible": true, "limitation": "Claim scope: Batelco by Beyon, Bahrain. No accuracy, escalation or eligible-interaction denominator. Application: internal.", "rationale": "p30 describes Basma's operating functions and backend integration. Scope is Batelco, not every Beyon subsidiary." }, { "id": "TEL-033", "claim": "Beyon Cyber launched Oryx AI for alert investigation with live queries and human-in-loop oversight.", "url": "https://beyon.com/wp-content/uploads/2026/03/Beyon-AR2025-ENGLISH.pdf", "locator": "Printed pp.30-31 / PDF page index16, Beyon Cyber", "level": 3, "eligible": true, "limitation": "Claim scope: Beyon Cyber managed security operations. Classed service-delivery operating practice, not proof of use in Batelco or group functions; deployment volume unreported. Application: internal-service-delivery.", "rationale": "p16 describes AI integrated into Beyon Cyber managed-service delivery, with query orchestration and oversight. Correct name to Orryx AI; do not imply whole-group autonomy." } ], "boundary": "Batelco local operations and Beyon Cyber subsidiary claims must not imply whole-group adoption.", "question": "Locate AI inside the operator." }, { "companyId": "vodafone-qatar", "title": "Vodafone Qatar", "scope": "Vodafone Qatar P.Q.S.C.", "claims": [ { "id": "TEL-020", "claim": "Vodafone Qatar reports expanded Microsoft collaboration for cloud, AI and digital services.", "url": "https://www.vodafone.qa/en/investor-relations/financials/Annual-Report-2025", "locator": "Telco to TechCo section", "level": null, "eligible": false, "limitation": "Claim scope: Vodafone Qatar P.Q.S.C.. No attributable internal deployment coverage or mechanism. Application: customer-product. Exclusion: Customer product alone does not establish internal operating practice. No qualifying AI-specific operating anchor retained.", "rationale": "AI service partnership description does not establish internal operating practice. Exclusion: Customer product alone does not establish internal operating practice. No qualifying AI-specific operating anchor retained." }, { "id": "TEL-021", "claim": "Vodafone Qatar and KAIC intend to co-develop FlowTeam based on TENN.ai for client needs.", "url": "https://www.vodafone.qa/en/about-us/vodafone-qatar-and-kaic-sign-stategic-mou-to-advance-ai", "locator": "Main release paragraphs 1-3", "level": null, "eligible": false, "limitation": "Claim scope: Vodafone Qatar P.Q.S.C.. Indexed excerpt; full-page fetch failed. MoU is not deployed system. Application: customer-product. Exclusion: Customer product alone does not establish internal operating practice. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score.", "rationale": "Customer offering, not qualifying internal transformation. Exclusion: Customer product alone does not establish internal operating practice. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score." } ], "boundary": "Do not attribute Vodafone Group Plc worldwide practices to Vodafone Qatar.", "question": "Locate AI inside the operator." } ], "agenda": [ { "question": "Locate AI inside the operator.", "evidence": "A named network, service or employee workflow, distinct from products sold to customers." }, { "question": "Identify the operational fallback.", "evidence": "Human escalation, degraded-service procedure and evidence from live incidents." }, { "question": "Measure the service actually changed.", "evidence": "Workflow volume, resolution quality and an appropriate before/after comparator." } ], "sources": [ { "title": "e& Integrated Annual Report 2025", "publisher": "e&", "date": "Undated", "url": "https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf", "locator": "Printed p.128 / PDF page index129, Responsible AI", "originalUrl": "https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf", "access": "full_page | Review: full_pdf", "relationship": "company_annual_report", "id": "S01" }, { "title": "e& and TM Forum announce strategic blueprint for Autonomous Networks", "publisher": "e&", "date": "2026-06-22", "url": "https://www.eand.com/en/news/22-06-26-eand-tm-strategic-blueprint.html", "locator": "Main release, roadmap paragraphs", "originalUrl": "https://www.eand.com/en/news/22-06-26-eand-tm-strategic-blueprint.html", "access": "full_page", "relationship": "company_press_release", "id": "S02" }, { "title": "Beyon Sustainability Report 2025", "publisher": "Beyon", "date": "Undated", "url": "https://beyon.com/wp-content/uploads/2026/03/Beyon_ESG2025_English.pdf", "locator": "Printed pp.58-59 / PDF page index30, Batelco by Beyon", "originalUrl": "https://beyon.com/wp-content/uploads/2026/03/Beyon_ESG2025_English.pdf", "access": "full_page | Review: full_pdf", "relationship": "company_sustainability_report", "id": "S03" }, { "title": "Beyon Annual Report 2025", "publisher": "Beyon", "date": "Undated", "url": "https://beyon.com/wp-content/uploads/2026/03/Beyon-AR2025-ENGLISH.pdf", "locator": "Printed pp.26-27 / PDF page index14, Batelco by Beyon", "originalUrl": "https://beyon.com/wp-content/uploads/2026/03/Beyon-AR2025-ENGLISH.pdf", "access": "full_page | Review: full_pdf", "relationship": "company_annual_report", "id": "S04" }, { "title": "Ooredoo Annual Report 2025", "publisher": "Ooredoo", "date": "Undated", "url": "https://www.ooredoo.com/wp-content/uploads/2026/03/2025-Ooredoo-Annual-Report_English_FINAL.pdf", "locator": "Printed pp.4-5 / PDF page index3, Customer Experience", "originalUrl": "https://www.ooredoo.com/wp-content/uploads/2026/03/2025-Ooredoo-Annual-Report_English_FINAL.pdf", "access": "full_page | Review: full_pdf", "relationship": "company_annual_report", "id": "S05" }, { "title": "Ooredoo Group FY 2025", "publisher": "Ooredoo", "date": "Undated", "url": "https://www.ooredoo.com/en/media/news_view/ooredoo-group-fy-2025/", "locator": "Syntys / sovereign AI cloud paragraph", "originalUrl": "https://www.ooredoo.com/en/media/news_view/ooredoo-group-fy-2025/", "access": "full_page", "relationship": "company_results_release", "id": "S06" }, { "title": "stc Sustainability Report 2025: Corporate governance", "publisher": "stc group", "date": "Undated", "url": "https://www.stc.com/content/dam/groupsites/stc-sustainability-report-2025/corporate-governance/", "locator": "Responsible AI, paragraphs 1-4", "originalUrl": "https://www.stc.com/content/dam/groupsites/stc-sustainability-report-2025/corporate-governance/", "access": "full_page | Review: full_page", "relationship": "company_sustainability_report", "id": "S07" }, { "title": "stc Sustainability Report 2025: Our customers", "publisher": "stc group", "date": "Undated", "url": "https://www.stc.com/content/dam/groupsites/stc-sustainability-report-2025/our-customers/", "locator": "Reliable and resilient infrastructure and services", "originalUrl": "https://www.stc.com/content/dam/groupsites/stc-sustainability-report-2025/our-customers/", "access": "indexed_excerpt | Review: full_page", "relationship": "company_sustainability_report", "id": "S08" }, { "title": "stc Kuwait Annual Report 2025: Overview", "publisher": "stc Kuwait", "date": "Undated", "url": "https://cws.stc.com.kw/DigitalStatic/AnnualReport2025/index.html", "locator": "2025 timeline, June and July", "originalUrl": "https://cws.stc.com.kw/DigitalStatic/AnnualReport2025/index.html", "access": "full_page | Review: full_page", "relationship": "company_annual_report", "id": "S09" }, { "title": "stc Kuwait Annual Report 2025: Technology Sector", "publisher": "stc Kuwait", "date": "Undated", "url": "https://cws.stc.com.kw/DigitalStatic/AnnualReport2025/pdfs/sections/en/Technology%20sector.pdf", "locator": "Printed pp.70-71 / PDF page index0, Building Smarter IT Foundations", "originalUrl": "https://cws.stc.com.kw/DigitalStatic/AnnualReport2025/pdfs/sections/en/Technology%20sector.pdf", "access": "full_page", "relationship": "company_annual_report", "id": "S10" }, { "title": "stc Bahrain wins awards for AI-powered CX assurance and convergent billing", "publisher": "stc Bahrain", "date": "2025-10-26", "url": "https://www.stc.com.bh/media-center/template?id=4346", "locator": "AI and Machine Learning Customer Experience award paragraph", "originalUrl": "https://www.stc.com.bh/media-center/template?id=4346", "access": "full_page | Review: full_page", "relationship": "company_press_release", "id": "S11" }, { "title": "stc Bahrain signs partnership agreement with Huawei", "publisher": "stc Bahrain", "date": "2025-06-16", "url": "https://www.stc.com.bh/media-center/template?id=4188", "locator": "Main agreement and roadmap paragraphs", "originalUrl": "https://www.stc.com.bh/media-center/template?id=4188", "access": "full_page | Review: full_page", "relationship": "company_press_release", "id": "S12" }, { "title": "stc Agentic AI product page", "publisher": "stc Bahrain", "date": "Undated", "url": "https://business.stc.com.bh/ai-solutions/stc-agentic-ai", "locator": "Product introduction and Autonomous Intelligence", "originalUrl": "https://business.stc.com.bh/ai-solutions/stc-agentic-ai", "access": "indexed_excerpt", "relationship": "company_product_page", "id": "S13" }, { "title": "Zain Annual Report 2025: Technology", "publisher": "Zain", "date": "Undated", "url": "https://zain.com/AR2025/technology/", "locator": "Personal Data Privacy Policies and Framework", "originalUrl": "https://zain.com/AR2025/technology/", "access": "full_page | Review: full_page", "relationship": "company_annual_report", "id": "S14" }, { "title": "Zain Annual Report 2025: Strategy", "publisher": "Zain", "date": "Undated", "url": "https://zain.com/AR2025/strategy/", "locator": "Force 2: Digital Zain", "originalUrl": "https://zain.com/AR2025/strategy/", "access": "full_page", "relationship": "company_annual_report", "id": "S15" }, { "title": "du launches Microsoft 365 Copilot for enhanced workplace efficiency", "publisher": "du", "date": "2025-03-13", "url": "https://www.du.ae/about/media-centre/newsdetail/du-launches-microsoft-365-copilot", "locator": "Main release, first-phase paragraph", "originalUrl": "https://www.du.ae/about/media-centre/newsdetail/du-launches-microsoft-365-copilot", "access": "full_page | Review: full_page", "relationship": "company_press_release", "id": "S16" }, { "title": "du announces collaboration with Microsoft to transform AI call center operations", "publisher": "du", "date": "2025-03-05", "url": "https://www.du.ae/about/media-centre/newsdetail/du-announces-collaboration-with-microsoft", "locator": "Project technologies and subscriber-copilot paragraphs", "originalUrl": "https://www.du.ae/about/media-centre/newsdetail/du-announces-collaboration-with-microsoft", "access": "full_page", "relationship": "company_press_release", "id": "S17" }, { "title": "du launches AI Park and National Hybrid AI at GITEX 2025", "publisher": "du", "date": "2025-10-13", "url": "https://www.du.ae/about/media-centre/newsdetail/du-launches-ai-park-and-national-hybrid-ai-at-gitex-2025", "locator": "AI Park and National Hybrid AI sections", "originalUrl": "https://www.du.ae/about/media-centre/newsdetail/du-launches-ai-park-and-national-hybrid-ai-at-gitex-2025", "access": "full_page", "relationship": "company_press_release", "id": "S18" }, { "title": "Mobily Annual Report 2025: Technology and Innovation", "publisher": "Mobily", "date": "Undated", "url": "https://apisit.mobily.com.sa/wps/wcm/connect/e30b0c6d-81ce-485a-bdbb-891eaab24db8/Mobily%2BAR25%2B%2BEnglish.pdf?CACHEID=ROOTWORKSPACE-e30b0c6d-81ce-485a-bdbb-891eaab24db8-pSH9C9c&MOD=AJPERES", "locator": "Printed pp.104-105, AI, Automation and Intelligent Network Operations", "originalUrl": "https://apisit.mobily.com.sa/wps/wcm/connect/e30b0c6d-81ce-485a-bdbb-891eaab24db8/Mobily%2BAR25%2B%2BEnglish.pdf?CACHEID=ROOTWORKSPACE-e30b0c6d-81ce-485a-bdbb-891eaab24db8-pSH9C9c&MOD=AJPERES", "access": "indexed_excerpt_full_pdf_fetch_failed | Review: fetch_failed", "relationship": "company_annual_report", "id": "S19" }, { "title": "Ericsson and Mobily sign MoU at LEAP 2025 to explore intent-driven autonomous networks", "publisher": "Ericsson", "date": "2025-02-10", "url": "https://www.ericsson.com/en/press-releases/5/2025/ericsson-and-mobily-sign-mou-at-leap-2025-to-explore-intent-driven-autonomous-networks", "locator": "Main announcement paragraphs 1-2", "originalUrl": "https://www.ericsson.com/en/press-releases/5/2025/ericsson-and-mobily-sign-mou-at-leap-2025-to-explore-intent-driven-autonomous-networks", "access": "full_page | Review: full_page", "relationship": "partner_press_release", "id": "S20" }, { "title": "Vodafone Qatar Annual Report 2025", "publisher": "Vodafone Qatar", "date": "Undated", "url": "https://www.vodafone.qa/en/investor-relations/financials/Annual-Report-2025", "locator": "Telco to TechCo section", "originalUrl": "https://www.vodafone.qa/en/investor-relations/financials/Annual-Report-2025", "access": "full_page", "relationship": "company_annual_report", "id": "S21" }, { "title": "Vodafone Qatar and KAIC sign strategic MoU to advance AI innovation", "publisher": "Vodafone Qatar", "date": "2025-09-13", "url": "https://www.vodafone.qa/en/about-us/vodafone-qatar-and-kaic-sign-stategic-mou-to-advance-ai", "locator": "Main release paragraphs 1-3", "originalUrl": "https://www.vodafone.qa/en/about-us/vodafone-qatar-and-kaic-sign-stategic-mou-to-advance-ai", "access": "indexed_excerpt_full_page_fetch_failed", "relationship": "company_press_release", "id": "S22" }, { "title": "stc ownership table", "publisher": "Primary source as linked", "date": "See original", "url": "https://www.stc.com/content/dam/stc/stc-annual-report-2024/subsidiaries-investment.html", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S23" } ], "body": "# AI in Gulf telecoms: the next test is how the business operates\n\nMayden AI Native Index | Telecom research brief | 10 October 2026\n\n**Exploratory public-evidence edition.** This brief examines public disclosures from ten deliberately selected telecom entities headquartered or operating in Saudi Arabia, the UAE, Qatar, Kuwait and Bahrain. It is a structured evidence review, not a representative survey, audited maturity assessment or claim that one national market leads another. The accompanying evidence database preserves claim-level citations, deployment stages and source limitations. Company statements remain company statements.\n\n## The executive question\n\nTelecom operators can sell AI infrastructure, apply AI to network engineering and redesign their own enterprise at the same time. These are three different achievements. An operator may host sophisticated customer workloads while its finance, people and service processes remain largely conventional. Another may quietly improve internal work without publishing detailed case studies. A useful AI-native benchmark must distinguish those possibilities.\n\nThe public evidence already offers more substance than a collection of ambitious announcements. There are named employee platforms, network-assurance mechanisms, customer assistants and governance structures. The unresolved question is their penetration: which share of material work has changed, what decisions the systems can make, who owns failures and whether benefits survive the full cost of running them?\n\nThis is the opportunity for Mayden. Public research can identify the operating questions worth asking. Proprietary questionnaires and corroborating evidence can establish answers that public communications rarely supply.\n\n## What the disclosures reveal\n\n**Governance is becoming an operating system for AI.** stc reports a Responsible AI framework, reusable-model certification and a model registry. Its sustainability report also describes employee training and an internal platform soft-launched in 2023. The time distinction matters: a historical launch is evidence of activity, not proof of today's employee adoption. [stc governance report](https://www.stc.com/content/dam/groupsites/stc-sustainability-report-2025/corporate-governance/)\n\ne& describes a central governance platform, use-case review, named steering authority, risk processes and mandatory training. Its own forward agenda still includes full platform rollout across business units. A careful assessment therefore credits the disclosed mechanism while preserving incomplete coverage. Governance documents should lead to requests for actual approval, rejection, monitoring and incident records. [e& 2025 report, printed pp.128-129](https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf)\n\n**Network intelligence offers concrete workflows to inspect.** An indexed passage from Mobily's report identifies potential leads on network automation. Direct PDF access failed in this review. These leads are excluded from scoring and are not treated as verified factual conclusions here. The next source-recovery step should verify deployment scope, override arrangements and the denominator behind any claimed savings. [Mobily 2025 report, printed pp.104-105](https://apisit.mobily.com.sa/wps/wcm/connect/e30b0c6d-81ce-485a-bdbb-891eaab24db8/Mobily%2BAR25%2B%2BEnglish.pdf?CACHEID=ROOTWORKSPACE-e30b0c6d-81ce-485a-bdbb-891eaab24db8-pSH9C9c&MOD=AJPERES)\n\nThe boundary between deployed practice and future design is equally visible. Ericsson and Mobily's intent-driven-network announcement is an MoU to explore specified capabilities. e& UAE's June 2026 autonomous-network blueprint describes a roadmap towards higher autonomy. Neither announcement alone demonstrates completed transformation. [Ericsson announcement](https://www.ericsson.com/en/press-releases/5/2025/ericsson-and-mobily-sign-mou-at-leap-2025-to-explore-intent-driven-autonomous-networks), [e& roadmap](https://www.eand.com/en/news/22-06-26-eand-tm-strategic-blueprint.html)\n\n**The most revealing word in a press release may be “pilot.”** du announced an internal Copilot launch, but the body specifies testing with selected employees. Its Microsoft contact-centre announcement sets out proposed call insights, analytics and a VoiceBot. These are useful, bounded transformation signals. They do not establish sustained deployment at scale. The assessment follows the implementation described in the body, rather than the ambition of the headline. [du employee pilot](https://www.du.ae/about/media-centre/newsdetail/du-launches-microsoft-365-copilot), [du contact-centre programme](https://www.du.ae/about/media-centre/newsdetail/du-announces-collaboration-with-microsoft)\n\n**AI delivery to customers needs a separate lens.** Ooredoo reports deploying AI-enabled care tools across markets and a live sovereign AI cloud launch through Syntys. The first concerns how the operator works; the second concerns what it sells. du's AI Park announcement combines a phased infrastructure plan with a hybrid AI offering. Both may be commercially significant, but infrastructure capacity is not an organizational capability score. [Ooredoo annual report](https://www.ooredoo.com/wp-content/uploads/2026/03/2025-Ooredoo-Annual-Report_English_FINAL.pdf), [Ooredoo results](https://www.ooredoo.com/en/media/news_view/ooredoo-group-fy-2025/), [du AI Park](https://www.du.ae/about/media-centre/newsdetail/du-launches-ai-park-and-national-hybrid-ai-at-gitex-2025)\n\nVodafone Qatar's reviewed materials emphasize Microsoft collaboration and a proposed FlowTeam solution with KAIC for client needs. They support a customer-service innovation narrative. They do not, in this bounded source set, support scoring internal AI-native work practices. “Unknown from these sources” is the appropriate result. It would be misleading to translate limited disclosure into weak capability. [Vodafone Qatar report](https://www.vodafone.qa/en/investor-relations/financials/Annual-Report-2025), [KAIC announcement](https://www.vodafone.qa/en/about-us/vodafone-qatar-and-kaic-sign-stategic-mou-to-advance-ai)\n\n**Practical adoption often begins with a specific job.** Zain describes a Kuwait FME Mate field-engineering initiative for access to fault data, topology and recommended corrections. The report mixes trial and deployment language, supporting a limited documented implementation rather than an established scaled rollout. It also describes trial portfolios and broader closed-loop ambitions. Its reported efficiency improvement belongs to the specified engineering use case; expected avoided downtime is a forecast. These distinctions prevent promising engineering evidence becoming a claim about every employee or subsidiary. [Zain technology report](https://zain.com/AR2025/technology/)\n\nstc Kuwait reports an AI Unit, AI Office and employee AI Week. Those disclosures identify accountable structures and capability-building activity. They leave the assessment to determine whether ownership changes investment decisions, whether labs become maintained production services and whether employees continue using the tools after training. Its detailed technology review also illustrates why ordinary cloud, billing and provisioning improvements should not automatically earn AI credit. [stc Kuwait overview](https://cws.stc.com.kw/DigitalStatic/AnnualReport2025/index.html), [technology review](https://cws.stc.com.kw/DigitalStatic/AnnualReport2025/pdfs/sections/en/Technology%20sector.pdf)\n\n**Bahrain provides two distinct operating examples.** Batelco's Basma assistant has been extended to billing and other service tasks with backend integration. Beyon Cyber's security platform is described as supporting alert investigation with human oversight in managed security operations. The annual report spells it Oryx AI; the sustainability report uses Orryx.AI. This is bounded service-operation evidence, not proof of whole-Beyon autonomy. These practices belong to different entities within Beyon. stc Bahrain separately reports a deployed AI customer-experience assurance framework using topology and fault correlation, while its Huawei agreement concerns future joint development. [Beyon sustainability report, pp.58-59](https://beyon.com/wp-content/uploads/2026/03/Beyon_ESG2025_English.pdf), [Beyon annual report, pp.26-31](https://beyon.com/wp-content/uploads/2026/03/Beyon-AR2025-ENGLISH.pdf), [stc Bahrain assurance](https://www.stc.com.bh/media-center/template?id=4346), [Huawei agreement](https://www.stc.com.bh/media-center/template?id=4188)\n\n## The evidence that would change an executive decision\n\nThe proposed company study asks each participating operator to nominate three material workflows and provide a small, reproducible evidence pack. Each workflow should show its eligible volume, actual AI-assisted share, exceptions, escalation route and business owner. The organization should identify which decisions remain advisory, which require confirmation and which execute automatically. A sample of successful cases is insufficient: the pack needs failures, rework and reversals.\n\nOutcome measurement should retain handling time, elapsed resolution time, quality, customer impact, model cost, review effort and control failures separately. The same apparent productivity gain can represent faster work, transferred work or reduced service quality. Baseline period, case mix and comparison method should accompany any percentage improvement.\n\nFor employee adoption, licensed seats and training attendance are useful inputs. Sustained use, changed tasks, demonstrated skills and supervision burden are more informative. For governance, existence of a committee is a starting signal. The consequential evidence is a decision it made and a risk it controlled.\n\n## A credible comparison needs comparable entities\n\nThe launch cohort mixes international groups with local operators. stc group, stc Kuwait and stc Bahrain overlap: stc's 2024 subsidiary table records full ownership of Bahrain and 51.84% of Kuwait. e& also reports Mobily among its associates. Country labels in this review identify headquarters or the local operating market; they do not isolate where every disclosed capability operates. [stc ownership table](https://www.stc.com/content/dam/stc/stc-annual-report-2024/subsidiaries-investment.html), [e& report, associates note](https://www.eand.com/content/dam/eand/assets/docs/annual-report/2025/eand-integrated-annualreport-en.pdf)\n\nThe first public comparison should therefore display group and local-operator strata, source coverage and explicit unknowns. A future local-market study should scope evidence to Saudi stc operations, e& UAE, Ooredoo Qatar, Zain Kuwait and Batelco Bahrain, rather than silently relabel group disclosures. Parent and subsidiary records should remain linked.\n\n## A stronger headline for the benchmark\n\n“Who is most AI-native?” is attractive but premature as a factual conclusion from this corpus. The defensible launch story is more useful: **Gulf telecoms are showing how AI enters real operating work. The next test is coverage, accountability and verified value.**\n\nPublish a workflow map, deployment-stage timeline and evidence profile beside the company table. Let an executive inspect the claim, scope and source behind every mark. Invite companies to correct facts and contribute evidence under a consistent protocol. Over successive waves, the research can establish which practices spread, which gains persist and which organizational changes explain the difference. That longitudinal record is the beginning of a proprietary benchmark.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/telecom-public-evidence.pdf", "cover": "/research/index/2026-10/telecom-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 90524, "pdfSha256": "bf657aa8655efdf6c8e2eaef8d243ff33734a99bbe68999f556b35d795309c43", "pdfPages": 16 }, { "slug": "energy-public-evidence", "number": 4, "title": "AI in Gulf energy and utilities: examine the operating record", "shortTitle": "Intelligence in industrial operations", "dek": "Production, maintenance and control: evidence across ten selected operating entities.", "sector": "Energy & Utilities", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Distinguish prediction from control. Aramco describes automatic process adjustment at named sites; ADNOC describes equipment monitoring and maintenance prediction. Ask where authority passes from an operator to a system.", "Evaluate the AI as well as its output. DEWA reports agents generating and executing software tests. Request evidence that the agents themselves meet quality and failure-handling criteria.", "Confirm routine use after a field test. Bapco’s supplier describes an initial upstream model supporting engineers’ scenario analysis. Confirm sustained operation and intervention rules before treating that pilot as an established operating capability." ], "executiveSources": [ { "claimIds": [ "ARAMCO01", "ARAMCO02", "ADNOC01" ], "sources": [ { "claimId": "ARAMCO01", "url": "https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data", "sourceTitle": "AI and Big Data" }, { "claimId": "ARAMCO02", "url": "https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data", "sourceTitle": "AI and Big Data" }, { "claimId": "ADNOC01", "url": "https://www.adnoc.ae/en/news-and-media/press-releases/2024/adnoc-deploys-pioneering-ai-enabled-process-optimization-technology", "sourceTitle": "ADNOC Deploys Pioneering AI-Enabled Process Optimization Technology" } ] }, { "claimIds": [ "DEWA01" ], "sources": [ { "claimId": "DEWA01", "url": "https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/7/dewa-deploys-agentic-ai-across-its-digital-platforms", "sourceTitle": "DEWA deploys agentic AI across its digital platforms" } ] }, { "claimIds": [ "BAPCOU01", "BAPCOU02" ], "sources": [ { "claimId": "BAPCOU01", "url": "https://www.automation.com/article/geminus-ai-bapco-energies-slb-physics-native-ai-bahrain-energy-infrastructure", "sourceTitle": "Geminus AI, Bapco Energies and SLB to Deploy Physics-Native AI to Optimize Bahrain’s Energy Infrastructure" }, { "claimId": "BAPCOU02", "url": "https://www.automation.com/article/geminus-ai-bapco-energies-slb-physics-native-ai-bahrain-energy-infrastructure", "sourceTitle": "Geminus AI, Bapco Energies and SLB to Deploy Physics-Native AI to Optimize Bahrain’s Energy Infrastructure" } ] } ], "evidenceAgeContext": "Dated sources run from 2023-03-08 to 2026-07-06. 8 of 30 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "Who acts on the industrial signal?", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 36, "label": "recorded claims" }, { "value": 30, "label": "eligible claims" }, { "value": 2, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "Who acts on the industrial signal?", "subtitle": "Different decision arrangements in selected source accounts; no ordering implies that more autonomy is better.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "Predict a maintenance need", "unit": "Equipment at named facilities", "finding": "ADNOC describes Neuron5 using sensor readings at NEB and Taweelah to predict maintenance needs.", "boundary": "Prediction is not autonomous maintenance execution.", "implication": "Who accepts a prediction and schedules the intervention?", "claimIds": [ "ADNOC01" ], "sources": [ { "claimId": "ADNOC01", "url": "https://www.adnoc.ae/en/news-and-media/press-releases/2024/adnoc-deploys-pioneering-ai-enabled-process-optimization-technology", "sourceTitle": "ADNOC Deploys Pioneering AI-Enabled Process Optimization Technology" } ] }, { "label": "Adjust a process", "unit": "Oil stabilisation at Abqaiq", "finding": "Aramco describes AI/ML automatically adjusting the oil-stabilisation process.", "boundary": "Named site; group-wide control authority is not established.", "implication": "What operating envelope, intervention and fallback have been tested?", "claimIds": [ "ARAMCO02" ], "sources": [ { "claimId": "ARAMCO02", "url": "https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data", "sourceTitle": "AI and Big Data" } ] }, { "label": "Compare operating scenarios", "unit": "Bapco upstream field test", "finding": "The supplier account describes a model through which engineers test scenarios and choose setpoints.", "boundary": "Pilot coding retained; continuing routine use unresolved.", "implication": "Which decisions remained with engineers after the field test?", "claimIds": [ "BAPCOU02" ], "sources": [ { "claimId": "BAPCOU02", "url": "https://www.automation.com/article/geminus-ai-bapco-energies-slb-physics-native-ai-bahrain-energy-infrastructure", "sourceTitle": "Geminus AI, Bapco Energies and SLB to Deploy Physics-Native AI to Optimize Bahrain’s Energy Infrastructure" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "aramco", "label": "Aramco", "values": [ 3, 3, 3, 3, null, 2, 3, 2, null ], "score": 70, "rank": 1 }, { "id": "dewa", "label": "Dubai Electricity and Water Authority", "values": [ 3, 3, 2, 3, null, null, null, 3, null ], "score": 52, "rank": 2 }, { "id": "adnoc", "label": "ADNOC", "values": [ 3, 3, 3, null, null, 2, null, 2, null ], "score": 48, "rank": 3 }, { "id": "koc", "label": "Kuwait Oil Company", "values": [ 2, 3, 2, null, null, 1, null, null, null ], "score": 30, "rank": 4 }, { "id": "qelng", "label": "QatarEnergy LNG", "values": [ 2, 2, null, null, null, null, null, null, null ], "score": 15, "rank": 5 }, { "id": "bapco-upstream", "label": "Bapco Upstream", "values": [ 1, null, 1, 1, null, null, null, null, null ], "score": 11, "rank": 6 }, { "id": "sec", "label": "Saudi Energy Company (formerly Saudi Electricity Company)", "values": [ null, 1, null, null, null, null, null, null, 2 ], "score": 11, "rank": 6 }, { "id": "kahramaa", "label": "KAHRAMAA", "values": [ 1, null, null, null, null, null, null, null, null ], "score": 4, "rank": 8 }, { "id": "bapco-refining", "label": "Bapco Refining", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "knpc", "label": "Kuwait National Petroleum Company", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "Aramco", "value": 70, "secondary": 7, "rank": 1 }, { "label": "Dubai Electricity and Water Authority", "value": 52, "secondary": 5, "rank": 2 }, { "label": "ADNOC", "value": 48, "secondary": 5, "rank": 3 }, { "label": "Kuwait Oil Company", "value": 30, "secondary": 4, "rank": 4 }, { "label": "QatarEnergy LNG", "value": 15, "secondary": 2, "rank": 5 }, { "label": "Bapco Upstream", "value": 11, "secondary": 3, "rank": 6 }, { "label": "Saudi Energy Company (formerly Saudi Electricity Company)", "value": 11, "secondary": 2, "rank": 6 }, { "label": "KAHRAMAA", "value": 4, "secondary": 1, "rank": 8 }, { "label": "Bapco Refining", "value": null, "secondary": 0, "rank": null }, { "label": "Kuwait National Petroleum Company", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 30; one practice may support multiple distinct themes", "maximum": 13, "rows": [ { "label": "Specific commitment / pilot", "value": 7 }, { "label": "Implemented practice", "value": 10 }, { "label": "Implemented + operating detail", "value": 13 } ], "note": "Source: claim ledger. 6 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "Aramco", "value": 70.4, "after": 18.52 }, { "label": "Dubai Electricity and Water Authority", "value": 51.9, "after": 22.22 }, { "label": "ADNOC", "value": 48.1, "after": 22.22 }, { "label": "Kuwait Oil Company", "value": 29.6, "after": 14.81 }, { "label": "QatarEnergy LNG", "value": 14.8, "after": 0.0 }, { "label": "Bapco Upstream", "value": 11.1, "after": 0.0 }, { "label": "Saudi Energy Company (formerly Saudi Electricity Company)", "value": 11.1, "after": 0.0 }, { "label": "KAHRAMAA", "value": 3.7, "after": 0.0 }, { "label": "Bapco Refining", "value": null, "after": null }, { "label": "Kuwait National Petroleum Company", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "aramco", "title": "Aramco", "scope": "Saudi Aramco group; each claim restricted to named site or function.", "claims": [ { "id": "ARAMCO01", "claim": "Fadhili permanently adopted reinforcement-learning control for acid-gas removal after its October 2024 to April 2025 pilot.", "url": "https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data", "locator": "Fadhili Gas Plant", "level": 3, "eligible": true, "limitation": "Claim scope: Saudi Aramco group; each claim restricted to named site or function.. Site-level; rollout elsewhere remained assessment. Undated live page. Application: internal-operating-practice.", "rationale": "Fadhili passage states permanent adoption after a dated pilot and identifies reinforcement-learning acid-gas control. Evaluation remains ongoing; no group-wide deployment inferred." }, { "id": "ARAMCO02", "claim": "At Abqaiq, AI/ML automatically adjusts the oil-stabilization process.", "url": "https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data", "locator": "Abqaiq Plants", "level": 3, "eligible": true, "limitation": "Claim scope: Saudi Aramco group; each claim restricted to named site or function.. No override, fail-safe or decision-log details. Application: internal-operating-practice.", "rationale": "Abqaiq passage expressly attributes automatic process adjustment to AI/ML. Scope is that plant; safety authority and override controls are not demonstrated." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Name the asset and the decision." }, { "companyId": "dewa", "title": "Dubai Electricity and Water Authority", "scope": "DEWA utility operations and internal digital functions; external product sales excluded.", "claims": [ { "id": "DEWA01", "claim": "Testing Centre AI agents generate test artifacts from specifications, execute tests and report results in the engineering pipeline.", "url": "https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/7/dewa-deploys-agentic-ai-across-its-digital-platforms", "locator": "Boosting digital testing efficiency", "level": 3, "eligible": true, "limitation": "Claim scope: DEWA utility operations and internal digital functions; external product sales excluded.. This tests digital services; it does not establish evaluation of AI models. Application: internal-operating-practice.", "rationale": "The Testing Centre paragraph specifies embedded test generation, execution and reporting in an engineering pipeline. It demonstrates software-testing automation, not controlled evaluation of AI impact." }, { "id": "DEWA02", "claim": "AI agents help design and engineering staff generate components and detect deviations from approved standards.", "url": "https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/7/dewa-deploys-agentic-ai-across-its-digital-platforms", "locator": "Enhancing customer experience", "level": 3, "eligible": true, "limitation": "Claim scope: DEWA utility operations and internal digital functions; external product sales excluded.. No staff adoption denominator or controlled productivity study. Application: internal-operating-practice.", "rationale": "Named design and engineering roles use AI-generated components and standards-deviation detection. This is workforce augmentation; training completion and competence are separate." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Name the asset and the decision." }, { "companyId": "bapco-refining", "title": "Bapco Refining", "scope": "Refining subsidiary of Bapco Energies; not upstream sibling or whole group.", "claims": [ { "id": "BAPCOR01", "claim": "Indexed annual-report passage describes intended AI optimization of the Bapco Modernization Project.", "url": "https://bapco-phase2.objects.frb.io/Media-Centre/Reports/Annual-Report-2024.pdf", "locator": "Printed28–29 indexed lead", "level": null, "eligible": false, "limitation": "Claim scope: Refining subsidiary of Bapco Energies; not upstream sibling or whole group.. Public report download truncated; only indexed passage read. Future language is not deployment. Application: internal-operating-practice. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score.", "rationale": "Excluded pending full-document access. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score." }, { "id": "BAPCOR02", "claim": "Indexed tender requests engineering services covering APC, RTO and AI applications.", "url": "https://www.tenderboard.gov.bh/tenderdetails/?id=745%2F2025%2FBTB+%28T88046+%2861%29%29", "locator": "Indexed tender description", "level": null, "eligible": false, "limitation": "Claim scope: Refining subsidiary of Bapco Energies; not upstream sibling or whole group.. Tender is not award or deployment; fetched page inaccessible. Application: internal-operating-practice. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score.", "rationale": "Excluded pending full-source read. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Name the asset and the decision." } ], "agenda": [ { "question": "Name the asset and the decision.", "evidence": "Plant/site boundary, control-system connection and accountable operator." }, { "question": "Show how the model fails safely.", "evidence": "Alarm handling, interventions, downtime and process-safety controls." }, { "question": "Trace the move from pilot to routine use.", "evidence": "Commissioning evidence, recurring usage and maintained operating responsibility." } ], "sources": [ { "title": "AI and Big Data", "publisher": "Aramco", "date": "Undated", "url": "https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data", "locator": "Fadhili Gas Plant", "originalUrl": "https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-page", "id": "S01" }, { "title": "How AI helps Aramco turn old data into new opportunity", "publisher": "Aramco", "date": "2025-03-12", "url": "https://europe.aramco.com/en/news-media/news/2025/aramco-deploys-ai", "locator": "Final substantive paragraph", "originalUrl": "https://europe.aramco.com/en/news-media/news/2025/aramco-deploys-ai", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-article", "id": "S02" }, { "title": "DEWA deploys agentic AI across its digital platforms", "publisher": "DEWA", "date": "2026-07-06", "url": "https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/7/dewa-deploys-agentic-ai-across-its-digital-platforms", "locator": "Boosting digital testing efficiency", "originalUrl": "https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/7/dewa-deploys-agentic-ai-across-its-digital-platforms", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-release", "id": "S03" }, { "title": "DEWA’s virtual employee Rammas handled over 1.6 million inquiries in 2025", "publisher": "DEWA", "date": "2026-01-24", "url": "https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/1/dewas-virtual-employee-rammas-handled", "locator": "Openingparagraph", "originalUrl": "https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/1/dewas-virtual-employee-rammas-handled", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-release", "id": "S04" }, { "title": "ADNOC Deploys Pioneering AI-Enabled Process Optimization Technology", "publisher": "ADNOC", "date": "2024-08-27", "url": "https://www.adnoc.ae/en/news-and-media/press-releases/2024/adnoc-deploys-pioneering-ai-enabled-process-optimization-technology", "locator": "Release paragraphs1–5", "originalUrl": "https://www.adnoc.ae/en/news-and-media/press-releases/2024/adnoc-deploys-pioneering-ai-enabled-process-optimization-technology", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-release", "id": "S05" }, { "title": "2025: A Year in Review", "publisher": "ADNOC", "date": "Undated", "url": "https://reports.adnoc.ae/e/2025-a-year-in-review", "locator": "Printedslide9", "originalUrl": "https://reports.adnoc.ae/e/2025-a-year-in-review", "access": "full-source-read | Review: full-source-read", "relationship": "annual-review", "id": "S06" }, { "title": "First batch of Certified Technicians Graduate from AI Academy in the UK", "publisher": "Kuwait Oil Company", "date": "2025-11-03", "url": "https://www.kockw.com/sites/EN/EMagazine/Pages/Events/1082.aspx", "locator": "Bodyparagraphs1,4,6", "originalUrl": "https://www.kockw.com/sites/EN/EMagazine/Pages/Events/1082.aspx", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-release", "id": "S07" }, { "title": "AI Innovation Center inauguration", "publisher": "Kuwait Oil Company", "date": "2025-08-07", "url": "https://www.kockw.com/sites/EN/EMagazine/Pages/Events/1000.aspx", "locator": "Bodyparagraphs1,4", "originalUrl": "https://www.kockw.com/sites/EN/EMagazine/Pages/Events/1000.aspx", "access": "full-source-read", "relationship": "corporate-release", "id": "S08" }, { "title": "SLB Announces Fourth-Quarter and Full-Year 2023 Results", "publisher": "SLB", "date": "2024-01-19", "url": "https://investorcenter.slb.com/news-releases/news-release-details/slb-announces-fourth-quarter-and-full-year-2023-results/", "locator": "Quarterly Highlights KOC paragraph", "originalUrl": "https://investorcenter.slb.com/news-releases/news-release-details/slb-announces-fourth-quarter-and-full-year-2023-results/", "access": "full-source-read", "relationship": "partner-results-release", "id": "S09" }, { "title": "SLB Announces Third-Quarter 2023 Results", "publisher": "SLB", "date": "2023-10-20", "url": "https://investorcenter.slb.com/news-releases/news-release-details/slb-announces-third-quarter-2023-results/", "locator": "Digital Operations KOC paragraph", "originalUrl": "https://investorcenter.slb.com/news-releases/news-release-details/slb-announces-third-quarter-2023-results/", "access": "full-source-read", "relationship": "partner-results-release", "id": "S10" }, { "title": "Sustainability Report 2024", "publisher": "QatarEnergy LNG", "date": "2025-11-12", "url": "https://www.qatarenergylng.qa/Portals/0/DNNGalleryPro/uploads/2025/11/12/QatarEnergyLNGSustainabilityReport2024-EN.pdf", "locator": "PDF38; printed74–75; Adoption of AI", "originalUrl": "https://www.qatarenergylng.qa/Portals/0/DNNGalleryPro/uploads/2025/11/12/QatarEnergyLNGSustainabilityReport2024-EN.pdf", "access": "full-source-read", "relationship": "sustainability-report", "id": "S11" }, { "title": "Geminus AI, Bapco Energies and SLB to Deploy Physics-Native AI to Optimize Bahrain’s Energy Infrastructure", "publisher": "Geminus AI release carried by Automation.com", "date": "2026-04-14", "url": "https://www.automation.com/article/geminus-ai-bapco-energies-slb-physics-native-ai-bahrain-energy-infrastructure", "locator": "Initialusecase and field-test paragraphs", "originalUrl": "https://www.automation.com/article/geminus-ai-bapco-energies-slb-physics-native-ai-bahrain-energy-infrastructure", "access": "full-mirror-read | Review: full-mirror-read", "relationship": "company-issued-syndication", "id": "S12" }, { "title": "Annual Report 2024", "publisher": "Saudi Electricity Company", "date": "2025", "url": "https://www.se.com.sa/-/media/sec/Investors/Saudi-Electricity-Company-Annual-Report-2024_Eng_V21.ashx", "locator": "PDF62; printed122–123", "originalUrl": "https://www.se.com.sa/-/media/sec/Investors/Saudi-Electricity-Company-Annual-Report-2024_Eng_V21.ashx", "access": "full-source-read", "relationship": "annual-report", "id": "S13" }, { "title": "KAHRAMAA Launches AI Platform, in Collaboration with Microsoft and KPMG", "publisher": "KAHRAMAA", "date": "2023-03-08", "url": "https://km.qa/MediaCenter/Pages/NewsDetails.aspx?ItemID=431", "locator": "Body paragraphs1 and final technical paragraph", "originalUrl": "https://km.qa/MediaCenter/Pages/NewsDetails.aspx?ItemID=431", "access": "full-source-read", "relationship": "corporate-release", "id": "S14" }, { "title": "KAHRAMAA Google Cloud initiatives", "publisher": "KAHRAMAA", "date": "2025-12-11", "url": "https://www.km.qa/MediaCenter/Pages/NewsDetails.aspx?ItemID=503", "locator": "Indexed lead only", "originalUrl": "https://www.km.qa/MediaCenter/Pages/NewsDetails.aspx?ItemID=503", "access": "snippet-only", "relationship": "corporate-release", "id": "S15" }, { "title": "Bapco Energies Annual Report 2024", "publisher": "Bapco Energies", "date": "2025", "url": "https://bapco-phase2.objects.frb.io/Media-Centre/Reports/Annual-Report-2024.pdf", "locator": "Printed28–29 indexed lead", "originalUrl": "https://bapco-phase2.objects.frb.io/Media-Centre/Reports/Annual-Report-2024.pdf", "access": "inaccessible", "relationship": "annual-report", "id": "S16" }, { "title": "Tender745/2025/BTB T88046(61): APC RTO and AI engineering", "publisher": "Bahrain Tender Board", "date": "2025-12-08", "url": "https://www.tenderboard.gov.bh/tenderdetails/?id=745%2F2025%2FBTB+%28T88046+%2861%29%29", "locator": "Indexed tender description", "originalUrl": "https://www.tenderboard.gov.bh/tenderdetails/?id=745%2F2025%2FBTB+%28T88046+%2861%29%29", "access": "snippet-only", "relationship": "official-tender", "id": "S17" }, { "title": "KNPC Technology Magazine, issue1", "publisher": "KNPC", "date": "Undated", "url": "https://www.knpc.com/knpc/media/pdf/tech-magazine-1_1.pdf", "locator": "PDF8; lines324–329", "originalUrl": "https://www.knpc.com/knpc/media/pdf/tech-magazine-1_1.pdf", "access": "full-source-read", "relationship": "company-magazine", "id": "S18" }, { "title": "annual-report archive", "publisher": "Primary source as linked", "date": "See original", "url": "https://www.knpc.com/en/media-center/publications/reports/annual-reports", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S19" } ], "body": "# AI in energy and utilities: where intelligence enters the operating system\n\nMayden AI Native Index | Energy & Utilities expansion | Research cutoff: 10 October 2026\n\n**Research status:** exploratory source review with a separate AI-agent challenge completed. Human and academic validation remain pending. This report does not rank actual company maturity. It covers ten selected entities across Saudi Arabia, the UAE, Qatar, Kuwait and Bahrain; Oman is outside this edition. Eight have proposed qualifying claims in this corpus. Two remain unknown records because current qualifying evidence was not established. Their inclusion makes the disclosure limits visible.\n\n## The executive finding\n\nThe most useful distinction in industrial AI is the decision being changed. A system may detect a condition, recommend an intervention, or directly adjust a process. Those are different operating arrangements. Announcements often blur them; organizational measurement must keep them separate.\n\nFor Mayden, an AI-native energy enterprise should therefore be investigated through observable work: the signal entering a model, the operating decision it changes, the accountable role, the permitted action, and the feedback used to improve it. A portfolio of impressive demonstrations cannot answer those questions by itself. Neither can a large digital-transformation budget.\n\nThis review finds evidence ranging from automated process control to maintenance prediction, engineering assistance, employee learning and customer service. It also finds procurement intentions and digital platforms whose AI features are described as future additions. These categories remain distinct in the database. Company and supplier statements are attributed disclosures, not independent assurance.\n\n## Start with the boundary between advice and action\n\nAramco describes permanent adoption at Fadhili and automatic adjustment at Abqaiq. These are useful starting points for examining bounded industrial autonomy. The database records each named site separately from the wider group. It does not infer enterprise-wide autonomy from a successful local installation. [Aramco operating examples](https://www.aramco.com/en/what-we-do/energy-innovation/digitalization/ai-and-big-data)\n\nThe next investigative step is practical: identify the approved operating envelope, the conditions requiring operator intervention and the fallback when model confidence or data quality deteriorates. The public passage is a starting point for that inquiry, not an answer to every control question.\n\nADNOC's Neuron5 release supports a different description: equipment monitoring and maintenance prediction at named facilities. It should not be recast as autonomous maintenance execution. Its pilot benefit figures describe potential improvements, so the outcome ledger preserves that qualification. [ADNOC deployment release](https://www.adnoc.ae/en/news-and-media/press-releases/2024/adnoc-deploys-pioneering-ai-enabled-process-optimization-technology)\n\nBapco's upstream case illustrates a third boundary. A vendor-issued release describes an initial field-tested model that supports engineers' scenario analysis. Routine post-test operation and wider expansion have not been established. This is evidence of decision support; the source does not establish independent machine authority over production. The release names an upstream executive, which helps prevent its claims being transferred to Bapco Refining. [Geminus-issued release carried by Automation.com](https://www.automation.com/article/geminus-ai-bapco-energies-slb-physics-native-ai-bahrain-energy-infrastructure)\n\nThe analytical implication is that autonomy should be assessed through delegated authority, not the sophistication of a model name. For a questionnaire, ask an operations leader to select one live workflow and show exactly where responsibility passes between people and software.\n\n## Distinguish AI work from digital work\n\nQatarEnergy LNG's 2024 sustainability report explicitly identifies machine learning in predictive maintenance and Bypass Override Risk Assessment. The same section describes broader automation initiatives. Mayden records the AI-specific statement without turning the surrounding digital portfolio into an AI count. The report also separates general digital literacy from formal AI certification. [QatarEnergy LNG report, printed pages 74-75](https://www.qatarenergylng.qa/Portals/0/DNNGalleryPro/uploads/2025/11/12/QatarEnergyLNGSustainabilityReport2024-EN.pdf)\n\nThis distinction is important for executive assessments. Ask what technique is actually used, where its output enters the workflow and whether the workflow continues to run in production. A rules-based workflow can be valuable without being machine learning. Calling every automation project AI makes comparisons less meaningful.\n\nThe Saudi Electricity Company 2024 report provides a useful edge case. It calls contracting RPA artificial intelligence but gives no corresponding inference or learning mechanism. That item is retained as an ambiguous, ineligible lead. Separately, the report explicitly puts AI integration within the Technical Committee's remit; this supports a narrower governance claim, without establishing model-level controls. [Annual report, printed pages 156-157 and 252-253](https://www.se.com.sa/-/media/sec/Investors/Saudi-Electricity-Company-Annual-Report-2024_Eng_V21.ashx)\n\nKNPC's accessible technology magazine is another caution. It describes existing digital maintenance processes while presenting AI predictive maintenance as a future upgrade. Its publication timing is unclear and nearby charts refer to 2018. The record is consequently historical context, not evidence of current AI deployment. The newer annual-report archive was found, but its report contents were not successfully retrieved during this review. [KNPC magazine](https://www.knpc.com/knpc/media/pdf/tech-magazine-1_1.pdf), [annual-report archive](https://www.knpc.com/en/media-center/publications/reports/annual-reports)\n\n## Follow how staff actually use the tools\n\nEmployee capability is easier to discuss than to measure. A participant count may show programme coverage; it does not prove proficiency or sustained use. Aramco's training disclosure and ADNOC's annual review provide candidate coverage indicators, but neither justifies a claim that every trained employee can independently operate or evaluate an AI system. [Aramco workforce disclosure](https://europe.aramco.com/en/news-media/news/2025/aramco-deploys-ai), [ADNOC 2025 review](https://reports.adnoc.ae/e/2025-a-year-in-review)\n\nKuwait Oil Company's academy disclosure is more specific about programme design: completed training, a defined duration and practical oil-and-gas applications. Its evidence is still bounded to the reported cohort. Separate SLB disclosures describe a mud-loss workflow and an earlier proof-of-value project; the latter remains a pilot in this dataset. A supplier release is an interested primary account, not independent corroboration. [KOC academy](https://www.kockw.com/sites/EN/EMagazine/Pages/Events/1082.aspx), [SLB 2023 year-end results](https://investorcenter.slb.com/news-releases/news-release-details/slb-announces-fourth-quarter-and-full-year-2023-results/), [SLB third-quarter 2023 results](https://investorcenter.slb.com/news-releases/news-release-details/slb-announces-third-quarter-2023-results/)\n\nDEWA's reported engineering and service workflows extend the inquiry beyond industrial plants. Their relevance is the integration of AI into routine work. However, AI generating software test artifacts is not evidence that the AI itself is systematically evaluated. Website-grounded answers likewise do not, by themselves, establish accuracy or knowledge freshness. [DEWA engineering deployment](https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/7/dewa-deploys-agentic-ai-across-its-digital-platforms), [Rammas disclosure](https://dewa.gov.ae/en/about-us/media-publications/latest-news/2026/1/dewas-virtual-employee-rammas-handled)\n\nA useful company report should show adoption and proficiency separately: eligible staff, active users, assessed competence, workflow dependence and exception rates. These remain proposed measures for future proprietary research, not statistics collected in this edition.\n\n## Unknowns are research priorities\n\nKAHRAMAA's accessible platform launch identifies a specific initiative, but its operating applications are largely future-tense. A subsequent Google-related lead could not be read in full and is excluded. Bapco Refining's indexed report and procurement leads likewise do not establish qualifying deployed practice. The group annual-report download was incomplete, and the tender page was inaccessible. [KAHRAMAA launch](https://km.qa/MediaCenter/Pages/NewsDetails.aspx?ItemID=431)\n\nThese cases do not show that the organizations lack AI capability. They show what this retrieval exercise could establish. A responsible explorer should display the unknown state and its reason directly, rather than hide it beneath a zero labelled maturity.\n\nThe ten records also do not form ten independent corporate groups. KOC and KNPC share KPC ownership; Bapco Upstream and Bapco Refining are siblings. KAHRAMAA is a public utility corporation. Country labels describe the selected entity's base and do not constrain the geography of group evidence. A cross-country league table would overstate comparability.\n\n## Questions that would turn disclosures into organizational evidence\n\nA company interview should move through one named live workflow before asking about the broader portfolio:\n\n1. What decision or task changed, and on what date did the system enter routine use?\n2. Which assets, teams or transactions are eligible, and what share actually uses it?\n3. Can the system act directly, or does a named role approve its recommendation?\n4. What input, model version, output, approval and override can be reconstructed afterward?\n5. What tests and operating thresholds decide whether it is deployed, changed or withdrawn?\n6. How are failures, false alerts and near misses captured, including cases where staff bypass it?\n7. Who maintains the knowledge or training data, and how are permissions and revisions controlled?\n8. What outcome changed against a defined baseline, with what denominator and alternative explanations?\n9. Which skills are assessed, rather than inferred from course attendance?\n10. What remains reusable when a vendor, model or individual employee changes?\n\nThese questions require different respondents and artifacts. Operations can explain decision authority; engineering can demonstrate evaluation; workforce teams can substantiate learning; risk and assurance functions can show approvals and incidents. A single confident executive cannot substitute for all four perspectives.\n\n## What this expansion can support now\n\nThe immediate product is a transparent evidence collection and an informed executive discussion. It is not a representative estimate of Gulf adoption, a causal evaluation of business impact, or a validated maturity instrument. A separate agent challenged every proposed level-3 claim and sampled lower levels. Its decisions are recorded beside the source proposals; this is not human or academic validation.\n\nThe underlying JSON preserves source proposals, excluded leads, access states, source locators, implementation states, limitations and per-company search logs. The separate review downgraded the three Bapco field-pilot claims and excluded a general DEWA governance-framework statement. Outcomes are stored separately and do not increase evidence points. Missing themes mean no qualifying evidence was located in this bounded corpus.\n\nThe next research wave should prioritize current report retrieval for weak-access entities, a systematic Arabic search, direct evidence of decision lineage and AI-specific control operation, and a consistent refresh window. Mayden can then invite organizations to correct factual scope and supply additional artifacts through a documented process, without selling a higher score or implying that payment establishes capability.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/energy-public-evidence.pdf", "cover": "/research/index/2026-10/energy-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 88794, "pdfSha256": "5023e2ac605a07fe7c0685610a0fbc6c7063ec7a8d8b84f95e751a4689b24a90", "pdfPages": 16 }, { "slug": "logistics-public-evidence", "number": 5, "title": "AI in Gulf aviation and logistics: trace the operating change", "shortTitle": "AI across the movement of goods and people", "dek": "A review of selected airlines, logistics groups and airport operators.", "sector": "Aviation & Logistics", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Follow implementation through its rollout. Wiremind reports Qatar Airways Cargo’s staged deployment of demand, inventory and overbooking tools. Ask which routes and decisions are covered, and what remains outside the system.", "Ask what model validation changed. Wiremind describes testing and iterating ML approaches with the cargo team. Request acceptance criteria and comparative results before inferring better commercial performance.", "Keep group claims within their operating scope. Emirates reports an IT AI gateway and structured software-delivery records. Examine active use and control operation within IT before extending conclusions across the airline or group." ], "executiveSources": [ { "claimIds": [ "LOG-017" ], "sources": [ { "claimId": "LOG-017", "url": "https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser", "sourceTitle": "Qatar Airways Cargo goes live with CARGOSTACK Optimiser" } ] }, { "claimIds": [ "LOG-018" ], "sources": [ { "claimId": "LOG-018", "url": "https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser", "sourceTitle": "Qatar Airways Cargo goes live with CARGOSTACK Optimiser" } ] }, { "claimIds": [ "LOG-004", "LOG-005", "LOG-006" ], "sources": [ { "claimId": "LOG-004", "url": "https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/", "sourceTitle": "Catalysing the next wave of innovation: ForsaTEK 2026 concludes successfully" }, { "claimId": "LOG-005", "url": "https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/", "sourceTitle": "Catalysing the next wave of innovation: ForsaTEK 2026 concludes successfully" }, { "claimId": "LOG-006", "url": "https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/", "sourceTitle": "Catalysing the next wave of innovation: ForsaTEK 2026 concludes successfully" } ] } ], "evidenceAgeContext": "Dated sources run from 2021-10-31 to 2026-09-28. 9 of 32 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "Match the logistics model to its unit of work", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 37, "label": "recorded claims" }, { "value": 32, "label": "eligible claims" }, { "value": 1, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "Match the logistics model to its unit of work", "subtitle": "Three separate allocation and information workflows; coverage must follow the named operation.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "Cargo allocation", "unit": "Demand, inventory and overbooking decision", "finding": "Wiremind describes progressive rollout of optimisation tools within Qatar Airways Cargo.", "boundary": "Cargo scope; complete network and module coverage unestablished.", "implication": "Which decisions use the recommendation, and which exceptions revert to people?", "claimIds": [ "LOG-017" ], "sources": [ { "claimId": "LOG-017", "url": "https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser", "sourceTitle": "Qatar Airways Cargo goes live with CARGOSTACK Optimiser" } ] }, { "label": "Training allocation", "unit": "Instructor and facility schedule", "finding": "Microsoft describes Saudia Academy implementing AI-enabled scheduling in training management.", "boundary": "Academy workflow; not the airline’s wider operating system.", "implication": "How many eligible scheduling decisions use it, and what does a planner override?", "claimIds": [ "LOG-008" ], "sources": [ { "claimId": "LOG-008", "url": "https://www.microsoft.com/en/customers/story/1782079017420911107-saudia-academy-dynamics-365-customer-service-travel-and-transportation-en-saudi-arabia", "sourceTitle": "Saudia Academy transforms aviation training with Dynamics 365" } ] }, { "label": "Executive information", "unit": "Live operational question", "finding": "Gulf Air’s FalconEye account combines flight, passenger and crew information for executive questions.", "boundary": "Airline information interface; not autonomous operational command.", "implication": "Can the answer be traced to current data before an executive acts?", "claimIds": [ "LOG-033", "LOG-034" ], "sources": [ { "claimId": "LOG-033", "url": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/public-preview-voice-native-agents-in-microsoft-foundry/4502756", "sourceTitle": "Public preview: Voice-native agents in Microsoft Foundry" }, { "claimId": "LOG-034", "url": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/public-preview-voice-native-agents-in-microsoft-foundry/4502756", "sourceTitle": "Public preview: Voice-native agents in Microsoft Foundry" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "emirates-group", "label": "Emirates Group", "values": [ 2, 3, null, null, 3, 2, null, null, 3 ], "score": 48, "rank": 1 }, { "id": "qatar-airways-group", "label": "Qatar Airways Group", "values": [ 3, null, 2, null, null, 3, null, null, 1 ], "score": 33, "rank": 2 }, { "id": "dp-world", "label": "DP World", "values": [ 3, 1, null, 3, null, null, null, null, null ], "score": 26, "rank": 3 }, { "id": "bahri", "label": "Bahri", "values": [ 3, null, 1, null, null, null, null, 2, null ], "score": 22, "rank": 4 }, { "id": "gulf-air-group", "label": "Gulf Air Group", "values": [ null, 1, null, 3, null, null, null, 2, null ], "score": 22, "rank": 4 }, { "id": "saudia-group", "label": "Saudia Group", "values": [ 3, null, 2, null, null, null, null, null, null ], "score": 19, "rank": 6 }, { "id": "makhazen", "label": "Makhazen (Agility Public Warehousing)", "values": [ null, null, 1, 2, null, null, null, null, null ], "score": 11, "rank": 7 }, { "id": "jazeera-airways", "label": "Jazeera Airways", "values": [ 2, null, null, null, null, null, null, null, null ], "score": 7, "rank": 8 }, { "id": "milaha", "label": "Milaha", "values": [ 1, null, 1, null, null, null, null, null, null ], "score": 7, "rank": 8 }, { "id": "bahrain-airport-company", "label": "Bahrain Airport Company", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "Emirates Group", "value": 48, "secondary": 5, "rank": 1 }, { "label": "Qatar Airways Group", "value": 33, "secondary": 4, "rank": 2 }, { "label": "DP World", "value": 26, "secondary": 3, "rank": 3 }, { "label": "Bahri", "value": 22, "secondary": 3, "rank": 4 }, { "label": "Gulf Air Group", "value": 22, "secondary": 3, "rank": 4 }, { "label": "Saudia Group", "value": 19, "secondary": 2, "rank": 6 }, { "label": "Makhazen (Agility Public Warehousing)", "value": 11, "secondary": 2, "rank": 7 }, { "label": "Jazeera Airways", "value": 7, "secondary": 1, "rank": 8 }, { "label": "Milaha", "value": 7, "secondary": 2, "rank": 8 }, { "label": "Bahrain Airport Company", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 32; one practice may support multiple distinct themes", "maximum": 13, "rows": [ { "label": "Specific commitment / pilot", "value": 13 }, { "label": "Implemented practice", "value": 9 }, { "label": "Implemented + operating detail", "value": 10 } ], "note": "Source: claim ledger. 5 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "Emirates Group", "value": 48.1, "after": 14.81 }, { "label": "Qatar Airways Group", "value": 33.3, "after": 3.7 }, { "label": "DP World", "value": 25.9, "after": 11.11 }, { "label": "Bahri", "value": 22.2, "after": 3.7 }, { "label": "Gulf Air Group", "value": 22.2, "after": 14.81 }, { "label": "Saudia Group", "value": 18.5, "after": 0.0 }, { "label": "Makhazen (Agility Public Warehousing)", "value": 11.1, "after": 3.7 }, { "label": "Jazeera Airways", "value": 7.4, "after": 3.7 }, { "label": "Milaha", "value": 7.4, "after": 3.7 }, { "label": "Bahrain Airport Company", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "emirates-group", "title": "Emirates Group", "scope": "Emirates Group, including Emirates airline and dnata; claim-specific IT and Airport Services scopes", "claims": [ { "id": "LOG-004", "claim": "More than 2,000 Group IT staff had joined its AI gateway by August 2026.", "url": "https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/", "locator": "ARRIVE and AI gateway paragraphs near article end", "level": 3, "eligible": true, "limitation": "Claim scope: Emirates Group, including Emirates airline and dnata; claim-specific IT and Airport Services scopes. IT subset only; total eligible staff, usage frequency and productivity effects undisclosed. Application: internal-operating-practice.", "rationale": "Primary Emirates page paragraphs115–116 explicitly reports IT onboarding count. Retain level3 for bounded workforce coverage; do not infer active usage or all employees." }, { "id": "LOG-005", "claim": "ARRIVE records intent, scope, risk, evidence and decisions in AI-assisted software delivery.", "url": "https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/", "locator": "ARRIVE software-engineering framework paragraph", "level": 3, "eligible": true, "limitation": "Claim scope: Emirates Group, including Emirates airline and dnata; claim-specific IT and Airport Services scopes. IT software-delivery scope; record completeness and audit results undisclosed. Application: internal-operating-practice.", "rationale": "Primary Emirates paragraph115 describes records for intent, scope, risk, evidence and decisions in the software lifecycle. Retain3 for this IT lineage mechanism, not enterprise-wide traceability." } ], "boundary": "IT gateway and software-governance evidence does not establish workforce-wide or dnata-wide deployment.", "question": "Follow one shipment or journey end to end." }, { "companyId": "qatar-airways-group", "title": "Qatar Airways Group", "scope": "Qatar Airways Group; cargo revenue-management implementation and group transformation plans", "claims": [ { "id": "LOG-017", "claim": "Qatar Airways Cargo went live with CARGOSTACK Optimiser, gradually rolling out demand, inventory and overbooking optimization.", "url": "https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser", "locator": "Go-live paragraph and phased implementation details", "level": 3, "eligible": true, "limitation": "Claim scope: Qatar Airways Group; cargo revenue-management implementation and group transformation plans. Cargo scope; staged rollout means no inference of complete network or module coverage. Application: internal-operating-practice.", "rationale": "Reopened Wiremind primary implementation account94–95: go-live, progressive rollout, demand/inventory/overbooking and bid-price ML. Retain3 for Cargo revenue-management implementation, not all Qatar Airways operations." }, { "id": "LOG-018", "claim": "The cargo team and Wiremind tested and iterated multiple machine-learning approaches during model validation.", "url": "https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser", "locator": "Collaboration and model-development/validation paragraphs", "level": 3, "eligible": true, "limitation": "Claim scope: Qatar Airways Group; cargo revenue-management implementation and group transformation plans. Validation criteria, datasets and comparative performance results are not public. Application: internal-operating-practice.", "rationale": "Reopened Wiremind paragraph106: company-team implementation input and explicit multiple ML approaches tested and iterated. Retain3 under evaluation-process anchor; no criteria, effect size or rigorous independent validation established." } ], "boundary": "Cargo implementation must not be generalized to passenger operations or all group functions.", "question": "Follow one shipment or journey end to end." }, { "companyId": "bahrain-airport-company", "title": "Bahrain Airport Company", "scope": "Bahrain Airport Company W.L.L., operator of Bahrain International Airport and subsidiary of Gulf Air Group", "claims": [ { "id": "LOG-037", "claim": "Indexed town-hall coverage mentions ambitions involving AI alongside other digital technologies.", "url": "https://www.bahrainairport.bh/corporate/news/bahrain-airport-company-bac-hosts-journey-to-excellence-town-hall-and-launches-the-tamayouz-programme", "locator": "Search-indexed article passage; full page unavailable", "level": null, "eligible": false, "limitation": "Claim scope: Bahrain Airport Company W.L.L., operator of Bahrain International Airport and subsidiary of Gulf Air Group. No exact full-source passage read. Parent and airline implementations are not attributed to BAC. Application: excluded-inaccessible-and-generic. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score.", "rationale": "Excluded: inaccessible source and generic ambition do not establish a named internal AI practice. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Complete readable source support not established; indexed-only evidence cannot score." } ], "boundary": "Overlaps the Gulf Air Group parent. Parent/airline AI evidence is not inherited by this subsidiary. Retrieved airport material does not establish a qualifying operating case.", "question": "Follow one shipment or journey end to end." } ], "agenda": [ { "question": "Follow one shipment or journey end to end.", "evidence": "The task, handoffs, route/site coverage and accountable operating team." }, { "question": "Separate prediction from execution.", "evidence": "Which decisions remain advisory and which actions the system can trigger." }, { "question": "Test resilience under disruption.", "evidence": "Fallback procedures, exceptions and time-critical performance under non-routine demand." } ], "sources": [ { "title": "ForsaTEK 2025 offers platform for high-impact scalable innovation", "publisher": "Emirates Group", "date": "2025-06-23", "url": "https://www.emirates.com/media-centre/forsatek-2025-offers-platform-for-high-impact-scalable-innovation/", "locator": "AI-powered solutions paragraph; fully launched Airport Services initiative", "originalUrl": "https://www.emirates.com/media-centre/forsatek-2025-offers-platform-for-high-impact-scalable-innovation/", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S01" }, { "title": "Emirates Group collaborates with OpenAI to accelerate AI adoption and innovation", "publisher": "Emirates Group", "date": "2025-11-21", "url": "https://www.emirates.com/media-centre/emirates-group-collaborates-with-openai-to-accelerate-ai-adoption-and-innovation/", "locator": "Enterprise ChatGPT, training and internal champion network paragraphs", "originalUrl": "https://www.emirates.com/media-centre/emirates-group-collaborates-with-openai-to-accelerate-ai-adoption-and-innovation/", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S02" }, { "title": "Catalysing the next wave of innovation: ForsaTEK 2026 concludes successfully", "publisher": "Emirates Group", "date": "2026-09-28", "url": "https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/", "locator": "ARRIVE and AI gateway paragraphs near article end", "originalUrl": "https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-disclosure", "id": "S03" }, { "title": "Qatar Airways Cargo goes live with CARGOSTACK Optimiser", "publisher": "Wiremind", "date": "2024-11-06", "url": "https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser", "locator": "Go-live paragraph and phased implementation details", "originalUrl": "https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser", "access": "full-source-read | Review: full-source-read", "relationship": "official-partner-publication", "id": "S04" }, { "title": "Qatar Airways and Accenture embark on AI-powered partnership", "publisher": "Accenture", "date": "2025-08-13", "url": "https://newsroom.accenture.com/news/2025/qatar-airways-and-accenture-embark-on-ai-powered-partnership-to-set-new-benchmark-in-aviation-excellence", "locator": "Responsible AI and joint operating model paragraphs", "originalUrl": "https://newsroom.accenture.com/news/2025/qatar-airways-and-accenture-embark-on-ai-powered-partnership-to-set-new-benchmark-in-aviation-excellence", "access": "full-source-read", "relationship": "official-partner-publication", "id": "S05" }, { "title": "Is artificial intelligence the key to resilient supply chains?", "publisher": "DP World", "date": "Undated", "url": "https://www.dpworld.com/en/insights/is-artificial-intelligence-the-key-to-resilient-supply-chains", "locator": "What does AI look like in action / Jafza and TOS+ paragraphs", "originalUrl": "https://www.dpworld.com/en/insights/is-artificial-intelligence-the-key-to-resilient-supply-chains", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-disclosure", "id": "S06" }, { "title": "How AI is transforming global trade possibilities", "publisher": "DP World", "date": "Undated", "url": "https://www.dpworld.com/en/insights/how-ai-is-transforming-global-trade-possibilities", "locator": "Digital twin and CARGOES TOS+ paragraphs", "originalUrl": "https://www.dpworld.com/en/insights/how-ai-is-transforming-global-trade-possibilities", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-disclosure", "id": "S07" }, { "title": "Bahri Annual Report 2025", "publisher": "Bahri", "date": "Undated", "url": "https://www.bahri.sa/media/go4a3eke/bahri-ar25_english-final-to-tadawul.pdf", "locator": "PDF page 54, printed pages 106-107, Digital Transformation", "originalUrl": "https://www.bahri.sa/media/go4a3eke/bahri-ar25_english-final-to-tadawul.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-disclosure", "id": "S08" }, { "title": "Bahri partners with Aventra Group to accelerate digital transformation journey", "publisher": "Bahri", "date": "2022-08-08", "url": "https://www.bahri.sa/en/news-insights/news-insight/2022/bahri-partners-with-aventra-group-to-accelerate-digital-transformation-journey/", "locator": "Partnership and platform-function paragraphs", "originalUrl": "https://www.bahri.sa/en/news-insights/news-insight/2022/bahri-partners-with-aventra-group-to-accelerate-digital-transformation-journey/", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S09" }, { "title": "Gulf Air Group strengthens security and collaboration with Microsoft 365 E5", "publisher": "Microsoft", "date": "2025-02-17", "url": "https://www.microsoft.com/en/customers/story/22484-gulf-air-group-microsoft-365-e5", "locator": "Power BI Copilot future-development passage", "originalUrl": "https://www.microsoft.com/en/customers/story/22484-gulf-air-group-microsoft-365-e5", "access": "full-source-read", "relationship": "official-partner-publication", "id": "S10" }, { "title": "Public preview: Voice-native agents in Microsoft Foundry", "publisher": "Microsoft", "date": "2026-03-16", "url": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/public-preview-voice-native-agents-in-microsoft-foundry/4502756", "locator": "Customer example: Gulf Air FalconEye and Ahmed Naeemi statement", "originalUrl": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/public-preview-voice-native-agents-in-microsoft-foundry/4502756", "access": "full-source-read | Review: full-source-read", "relationship": "official-partner-publication", "id": "S11" }, { "title": "What is new in Microsoft Foundry Agent Service at Build 2026", "publisher": "Microsoft", "date": "2026-06-02", "url": "https://devblogs.microsoft.com/foundry/agent-service-build2026/", "locator": "Real-time voice / Gulf Air example", "originalUrl": "https://devblogs.microsoft.com/foundry/agent-service-build2026/", "access": "full-source-read", "relationship": "official-partner-publication", "id": "S12" }, { "title": "Advancing speech innovation with Azure Speech in Microsoft Foundry", "publisher": "Microsoft", "date": "2025-11-20", "url": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/advancing-speech-innovation-with-azure-speech-in-microsoft-foundry/4471461", "locator": "Gulf Air customer example and company statement concerning testing", "originalUrl": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/advancing-speech-innovation-with-azure-speech-in-microsoft-foundry/4471461", "access": "full-source-read", "relationship": "official-partner-publication", "id": "S13" }, { "title": "Saudia Academy transforms aviation training with Dynamics 365", "publisher": "Microsoft", "date": "2024-06-20", "url": "https://www.microsoft.com/en/customers/story/1782079017420911107-saudia-academy-dynamics-365-customer-service-travel-and-transportation-en-saudi-arabia", "locator": "Technology implementation and instructor scheduling passages", "originalUrl": "https://www.microsoft.com/en/customers/story/1782079017420911107-saudia-academy-dynamics-365-customer-service-travel-and-transportation-en-saudi-arabia", "access": "full-source-read | Review: full-source-read", "relationship": "official-partner-publication", "id": "S14" }, { "title": "Agility Annual Report 2024", "publisher": "Makhazen", "date": "Undated", "url": "https://makhazen.com/wp-content/uploads/2025/05/Agility-Annual-Report-2024-En.pdf", "locator": "PDF page 27, IT security controls list", "originalUrl": "https://makhazen.com/wp-content/uploads/2025/05/Agility-Annual-Report-2024-En.pdf", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S15" }, { "title": "Agility signs MOU with Huawei to transform logistics industrial parks to smart campuses", "publisher": "Makhazen", "date": "2021-10-31", "url": "https://makhazen.com/en/agility-signs-mou-with-huawei-to-transform-its-logistics-industrial-parks-to-smart-campuses/", "locator": "Technology evaluation and Mina Abdullah campus paragraphs", "originalUrl": "https://makhazen.com/en/agility-signs-mou-with-huawei-to-transform-its-logistics-industrial-parks-to-smart-campuses/", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S16" }, { "title": "Makhazen Annual Report 2025", "publisher": "Makhazen", "date": "Undated", "url": "https://makhazen.com/wp-content/uploads/2026/05/Makhazen-Annual-Report-2025-English.pdf", "locator": "Corporate governance and internal-control sections", "originalUrl": "https://makhazen.com/wp-content/uploads/2026/05/Makhazen-Annual-Report-2025-English.pdf", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S17" }, { "title": "Jazeera Airways chooses SkyBreathe 360 eco-flying platform", "publisher": "OpenAirlines", "date": "2022-06-14", "url": "https://www.prnewswire.com/news-releases/jazeera-airways-chooses-skybreathe-360-eco-flying-platform-to-cut-carbon-emissions-301567557.html", "locator": "Platform selection and analytics-description paragraphs", "originalUrl": "https://www.prnewswire.com/news-releases/jazeera-airways-chooses-skybreathe-360-eco-flying-platform-to-cut-carbon-emissions-301567557.html", "access": "full-source-read", "relationship": "official-partner-publication", "id": "S18" }, { "title": "Jazeera Airways Annual Report 2024", "publisher": "Jazeera Airways", "date": "Undated", "url": "https://investorrelations.jazeeraairways.com/media/1367/j9-fy24-annual-report-english-mar-4-2025.pdf", "locator": "PDF pages 5-6 and 11, SkyBreathe fuel-efficiency references; corroboration L-S14", "originalUrl": "https://investorrelations.jazeeraairways.com/media/1367/j9-fy24-annual-report-english-mar-4-2025.pdf", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S19" }, { "title": "Jazeera Airways partners with TCS to power AI-led transformation", "publisher": "TCS", "date": "2025-05-03", "url": "https://www.tcs.com/who-we-are/newsroom/press-release/jazeera-airways-partners-with-tcs-to-power-ai-led-transformation-reimagine-digital-customer-experience", "locator": "Digital customer experience and AI-led architecture paragraphs", "originalUrl": "https://www.tcs.com/who-we-are/newsroom/press-release/jazeera-airways-partners-with-tcs-to-power-ai-led-transformation-reimagine-digital-customer-experience", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S20" }, { "title": "Milaha Annual Report 2025", "publisher": "Milaha", "date": "Undated", "url": "https://www.milaha.com/sites/default/files/2026-02/Milaha-Annual-Report-2025-EN.pdf", "locator": "PDF page 18, printed pages 30-31, Digital Transformation", "originalUrl": "https://www.milaha.com/sites/default/files/2026-02/Milaha-Annual-Report-2025-EN.pdf", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S21" }, { "title": "Milaha enhances data and AI solutions with Google Cloud", "publisher": "Milaha", "date": "2024-11-24", "url": "https://milaha.com/en/news-and-media/milaha-enhances-data-ai-solutions-google-cloud", "locator": "Data and AI collaboration paragraphs", "originalUrl": "https://milaha.com/en/news-and-media/milaha-enhances-data-ai-solutions-google-cloud", "access": "full-source-read", "relationship": "corporate-disclosure", "id": "S22" }, { "title": "BAC hosts Journey to Excellence town hall and launches Tamayouz programme", "publisher": "Bahrain Airport Company", "date": "2024-03-05", "url": "https://www.bahrainairport.bh/corporate/news/bahrain-airport-company-bac-hosts-journey-to-excellence-town-hall-and-launches-the-tamayouz-programme", "locator": "Search-indexed article passage; full page unavailable", "originalUrl": "https://www.bahrainairport.bh/corporate/news/bahrain-airport-company-bac-hosts-journey-to-excellence-town-hall-and-launches-the-tamayouz-programme", "access": "snippet-only", "relationship": "corporate-disclosure", "id": "S23" } ], "body": "# AI at work in GCC aviation and logistics\n\nExploratory public-evidence research | Evidence cutoff: 10 October 2026 | Source claims challenged for the public evidence edition\n\nThe most informative question for aviation and logistics executives is where AI changes an operating decision. A passenger-facing assistant can attract attention, but a cargo forecast, instructor-allocation system or terminal workflow provides a more direct view into how a company works. This review examines ten entities headquartered in Saudi Arabia, the UAE, Qatar, Kuwait and Bahrain. It finds several specific operating cases and substantial unanswered questions about their reach, controls and results.\n\nThis is a purposeful research cohort, not a representative survey of the region. The observations mix airlines, a training business, shipping, ports and logistics groups. Countries identify headquarters. They do not identify the territory in which every application operates. The accompanying evidence file retains disclosure-specificity coding and separate claim-level challenges for all proposed top anchors. It does not certify organizational maturity or produce a national league table.\n\n## The operating decision is the useful unit of analysis\n\nThree kinds of implementation help make the concept of an AI-native enterprise concrete: improving allocation, connecting information to work, and governing how systems are developed. These are useful research categories rather than a new scoring system. The Mayden framework retains its nine established dimensions, and repeated announcements do not accumulate points.\n\nSaudia Academy illustrates allocation in a training operation. Microsoft describes an implemented AI scheduling system delivered with Netways. The record identifies resources and workflow, rather than merely calling the organization digital. Its reported instructor and facility counts provide context. They do not establish adoption across Saudia Group, and the wider transformation outcomes cannot be attributed to AI alone. An interview should ask who can override a schedule and how conflicts, fairness and utilization are evaluated. [Microsoft customer account](https://www.microsoft.com/en/customers/story/1782079017420911107-saudia-academy-dynamics-365-customer-service-travel-and-transportation-en-saudi-arabia)\n\nQatar Airways Cargo offers a commercial allocation case. Wiremind reports that CARGOSTACK Optimiser went live through a gradual rollout and describes model testing with airline staff. That combination makes the implementation more informative than a partnership announcement: a business process, deployment stage and validation activity are visible. The account still leaves important questions unanswered, including model-comparison results, acceptance criteria and the extent of live coverage. These would be appropriate requests for the proprietary questionnaire. [Wiremind implementation account](https://www.wiremind.io/blog/qatar-airways-cargo-cargostack-optimiser)\n\nBahri reports vessel-data applications supporting maintenance and voyage-related decisions, alongside a shared enterprise information platform. The operational detail is concentrated in particular shipping businesses. It should not be turned into an assertion that every group function has the same capabilities. For executives, the useful follow-up is whether model recommendations change maintenance decisions and whether their contribution can be separated from sensor installation, process redesign and other investments. [Bahri 2025 annual report, printed pages 106–107](https://www.bahri.sa/media/go4a3eke/bahri-ar25_english-final-to-tadawul.pdf)\n\n## An interface matters when it connects to work\n\nDP World describes its Jafza digital twin as part of an AI-driven operating suite, with real-time information connected to basic terminal processes. That supports a bounded internal operating case. It does not establish that all terminal decisions are autonomous or that the same implementation is present throughout DP World's international network. The next layer of evidence would specify actions the system can execute, exceptions requiring approval, fallback arrangements and safety evaluation. [DP World operating description](https://www.dpworld.com/en/insights/how-ai-is-transforming-global-trade-possibilities)\n\nGulf Air's FalconEye is a different interface case: executives ask questions of live operational information and can use voice to open dashboards. Microsoft describes the airline example in its voice-agent publications. This is evidence about a particular managerial interface. It does not establish automated executive authority, universal answer accuracy or adoption by Bahrain Airport Company. A credible company report would investigate which answers require source references and how users handle conflicting information. [Microsoft voice-agent account](https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/public-preview-voice-native-agents-in-microsoft-foundry/4502756), [related Microsoft account](https://devblogs.microsoft.com/foundry/agent-service-build2026/)\n\nInterfaces should therefore be assessed through the work they support. Does an answer lead to a recommendation, a prepared transaction, or an executed action? Who remains responsible? Can the action be reversed? A public statement that a company has adopted a conversational tool rarely answers all four questions. Mayden's interview should.\n\n## The organizational system is harder to observe\n\nEmirates Group provides a useful IT example. Its September 2026 disclosure describes ARRIVE, an internal software-engineering framework, and staff onboarding through an AI gateway. The documentation connects AI-assisted development with recorded decisions and human accountability. This is valuable evidence about an operating mechanism. It remains bounded to the described IT context; it is not an enterprise-wide assurance finding. [Emirates ForsaTEK 2026 disclosure](https://www.emirates.com/media-centre/catalysing-the-next-wave-of-innovation-at-the-emirates-group-forsatek-2026-concludes-successfully/)\n\nThe distinction matters because an AI-native enterprise needs repeatable organizational practices, not simply a catalogue of applications. A strong diagnostic should ask how experimentation changes production standards, how staff know which tools they may use, and whether decision records survive across models and suppliers. The public cases suggest where to investigate these practices. They cannot settle the questions without direct evidence.\n\n## Plans need to remain visible as plans\n\nMilaha's annual report describes AI integration beginning at its logistics centre and outlines warehouse-related partnerships. Its Google Cloud announcement also identifies intended data and AI work. These are relevant signals of a named programme. The retrieved wording does not establish the production state, model configuration or operational coverage of every component. The dataset therefore preserves in-progress and planned statuses instead of converting a transformation narrative into completed deployment. [Milaha 2025 annual report](https://www.milaha.com/sites/default/files/2026-02/Milaha-Annual-Report-2025-EN.pdf), [Milaha cloud and AI announcement](https://milaha.com/en/news-and-media/milaha-enhances-data-ai-solutions-google-cloud)\n\nJazeera Airways requires a different reading across time. Its supplier announced selection of an AI-enabled fuel-analysis platform in 2022. The airline's 2024 report subsequently refers to continued participation in that programme. Together these sources support a bounded implementation record. The supplier's fuel-efficiency comparison is not a Mayden ranking, and it is not evidence of the airline's overall AI maturity. Jazeera's separate customer-experience transformation agreement is recorded but does not, by itself, establish an internal operating capability. [Jazeera 2024 annual report](https://investorrelations.jazeeraairways.com/media/1367/j9-fy24-annual-report-english-mar-4-2025.pdf), [OpenAirlines announcement](https://www.prnewswire.com/news-releases/jazeera-airways-chooses-skybreathe-360-eco-flying-platform-to-cut-carbon-emissions-301567557.html)\n\n## Scope and missing evidence change the interpretation\n\nMakhazen demonstrates why corporate identity belongs in the research database. Historical Agility reporting and current group structures cannot be treated as interchangeable. The 2024 report lists an AI-enabled security control, but its consolidated perimeter predates the subsequent restructuring. Employee-AI disclosures from the separately listed UAE Agility Global entity are not automatically evidence for the Kuwait-headquartered company. Readers should be able to see the historical boundary beside the claim. [Historical Agility 2024 report on the current issuer domain](https://makhazen.com/wp-content/uploads/2025/05/Agility-Annual-Report-2024-En.pdf), [Makhazen 2025 report](https://makhazen.com/wp-content/uploads/2026/05/Makhazen-Annual-Report-2025-English.pdf)\n\nBahrain Airport Company exposes another measurement problem: accessible public evidence may be sparse. The identified town-hall coverage was only recoverable through indexed material and contained broad technology ambitions. It is excluded from scoring. The airport subsidiary also does not inherit the airline's FalconEye evidence. All unknown dimensions mean that this bounded search did not establish a qualifying case. They do not mean that the company lacks AI capability.\n\nThese limitations are part of the product's value. An executive should be able to correct an entity boundary, supply a missing source, or challenge an implementation status. That creates a better benchmark than presenting a precise-looking number with hidden assumptions.\n\n## Questions for the operating team\n\nThe proprietary research should follow one live workflow from input to consequence. Six questions provide a practical starting point:\n\n1. Which operating decision has changed, and who uses the system regularly?\n2. What proportion of the relevant work does it cover, with what exceptions?\n3. What can it recommend or execute, and who can override it?\n4. Which evaluation or comparison justified deployment and continued use?\n5. What evidence links the outcome to AI rather than the wider transformation?\n6. How are errors, incidents and model changes recorded and resolved?\n\nA company should be able to answer with examples and supporting artifacts, while protecting confidential operational information. The resulting report can distinguish a promising pilot from a repeatable organizational capability and identify the next useful improvement. That is a stronger engagement proposition than promising a maturity verdict from public announcements.\n\nFor publication, use the specific operating cases as the story. Preserve the separation between disclosure strength, implementation state and demonstrated benefit. The research supports a serious discussion about how work is changing; a representative claim about the whole GCC would require a different sampling design and direct company participation.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/logistics-public-evidence.pdf", "cover": "/research/index/2026-10/logistics-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 89063, "pdfSha256": "a50c2d2c5f07ced65a19b36df1f27da663d432e67743f2ce1b04653868869328", "pdfPages": 16 }, { "slug": "manufacturing-public-evidence", "number": 6, "title": "AI in Gulf manufacturing: from process signals to operating decisions", "shortTitle": "From process signals to operating decisions", "dek": "Production, maintenance and quality across ten selected industrial entities.", "sector": "Manufacturing & Materials", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Start at the production decision. EGA reports AI vision for carbon-anode inspection at Al Taweelah. Ask how detected defects affect operator action, rework and accepted quality.", "Separate a completed model from sustained use. Alba describes a density-prediction model; EGA describes a planned autonomous-potline pilot. Confirm production use and decision authority before interpreting either as autonomous manufacturing.", "Measure the planning workflow end to end. Ducab reports AI within manufacturing order planning. Examine accepted schedules, planner overrides and actual operating results; improvement targets are not achieved outcomes." ], "executiveSources": [ { "claimIds": [ "MAN-EGA01" ], "sources": [ { "claimId": "MAN-EGA01", "url": "https://media.ega.ae/ega-designated-industry-40-global-lighthouse-by-world-economic-forum-in-uae-and-world-aluminium-industry-first/", "sourceTitle": "EGA designated Industry 4.0 Global Lighthouse" } ] }, { "claimIds": [ "MAN-ALBA01", "MAN-EGA02" ], "sources": [ { "claimId": "MAN-ALBA01", "url": "https://www.albasmelter.com/en/article/alba-joins-forces-with-nasser-artificial-intelligence-research-and-development-centre-to-advance-in-ai", "sourceTitle": "Alba and NAIRDC artificial intelligence collaboration" }, { "claimId": "MAN-EGA02", "url": "https://media.ega.ae/egas-ex-technology-pilot-to-be-used-to-prove-most-advanced-industry-40-capabilities-in-aluminium-industry/", "sourceTitle": "EX technology pilot and Industry 4.0 capabilities" } ] }, { "claimIds": [ "MAN-DUCAB01" ], "sources": [ { "claimId": "MAN-DUCAB01", "url": "https://ducab.com/media-center/press-releases/ducab-group-expedites-operations-with-the-launch-of-blade-the-smart-factory-project-of-the-future-based-on-industry-4-0-technologies/", "sourceTitle": "Ducab launches Blade smart factory project" } ] } ], "evidenceAgeContext": "Dated sources run from 2023-06-13 to 2026-09-23. 2 of 17 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "Three factory decisions. Three evidence tests.", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 17, "label": "recorded claims" }, { "value": 17, "label": "eligible claims" }, { "value": 5, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "Three factory decisions. Three evidence tests.", "subtitle": "Inspection, safety prioritisation and planning change different operating objects.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "Inspect quality", "unit": "Carbon-anode quality decision", "finding": "EGA reports real-time AI vision checking carbon-anode quality at Al Taweelah.", "boundary": "Site-specific inspection; group-wide autonomy not established.", "implication": "What is rejected, reviewed or reworked after the signal?", "claimIds": [ "MAN-EGA01" ], "sources": [ { "claimId": "MAN-EGA01", "url": "https://media.ega.ae/ega-designated-industry-40-global-lighthouse-by-world-economic-forum-in-uae-and-world-aluminium-industry-first/", "sourceTitle": "EGA designated Industry 4.0 Global Lighthouse" } ] }, { "label": "Prioritise safety attention", "unit": "High-risk operating area", "finding": "Alba describes AI using incident patterns and risk drivers to identify higher-risk areas.", "boundary": "Identification is reported; causal safety improvement is not verified.", "implication": "What action changes, and how are false alerts and missed risks reviewed?", "claimIds": [ "MAN-ALBA04" ], "sources": [ { "claimId": "MAN-ALBA04", "url": "https://www.albasmelter.com/uploads/albayan2026_q1.pdf", "sourceTitle": "Al Bayan Q1 2026" } ] }, { "label": "Plan production", "unit": "Manufacturing order schedule", "finding": "Ducab states that AI software is used in order planning within Blade.", "boundary": "Facility coverage undisclosed; targets are not achieved outcomes.", "implication": "Which schedule changes are accepted, overridden and measured?", "claimIds": [ "MAN-DUCAB01" ], "sources": [ { "claimId": "MAN-DUCAB01", "url": "https://ducab.com/media-center/press-releases/ducab-group-expedites-operations-with-the-launch-of-blade-the-smart-factory-project-of-the-future-based-on-industry-4-0-technologies/", "sourceTitle": "Ducab launches Blade smart factory project" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "alba", "label": "Aluminium Bahrain (Alba)", "values": [ 3, 3, 1, 3, null, 1, null, 1, 2 ], "score": 52, "rank": 1 }, { "id": "sabic", "label": "SABIC", "values": [ 3, 2, null, null, null, 1, null, 2, 2 ], "score": 37, "rank": 2 }, { "id": "ega", "label": "Emirates Global Aluminium", "values": [ 3, null, null, null, null, 1, null, null, null ], "score": 15, "rank": 3 }, { "id": "maaden", "label": "Maaden", "values": [ 2, null, 2, null, null, null, null, null, null ], "score": 15, "rank": 3 }, { "id": "ducab", "label": "Ducab", "values": [ 2, null, null, null, null, null, null, null, null ], "score": 7, "rank": 5 }, { "id": "equate", "label": "EQUATE", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "garmco", "label": "GARMCO", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "industries-qatar", "label": "Industries Qatar", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "pic", "label": "Petrochemical Industries Company", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "qapco", "label": "QAPCO", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "Aluminium Bahrain (Alba)", "value": 52, "secondary": 7, "rank": 1 }, { "label": "SABIC", "value": 37, "secondary": 5, "rank": 2 }, { "label": "Emirates Global Aluminium", "value": 15, "secondary": 2, "rank": 3 }, { "label": "Maaden", "value": 15, "secondary": 2, "rank": 3 }, { "label": "Ducab", "value": 7, "secondary": 1, "rank": 5 }, { "label": "EQUATE", "value": null, "secondary": 0, "rank": null }, { "label": "GARMCO", "value": null, "secondary": 0, "rank": null }, { "label": "Industries Qatar", "value": null, "secondary": 0, "rank": null }, { "label": "Petrochemical Industries Company", "value": null, "secondary": 0, "rank": null }, { "label": "QAPCO", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 17; one practice may support multiple distinct themes", "maximum": 7, "rows": [ { "label": "Specific commitment / pilot", "value": 5 }, { "label": "Implemented practice", "value": 7 }, { "label": "Implemented + operating detail", "value": 5 } ], "note": "Source: claim ledger. 0 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "Aluminium Bahrain (Alba)", "value": 51.9, "after": 25.93 }, { "label": "SABIC", "value": 37.0, "after": 0.0 }, { "label": "Emirates Global Aluminium", "value": 14.8, "after": 3.7 }, { "label": "Maaden", "value": 14.8, "after": 7.41 }, { "label": "Ducab", "value": 7.4, "after": 0.0 }, { "label": "EQUATE", "value": null, "after": null }, { "label": "GARMCO", "value": null, "after": null }, { "label": "Industries Qatar", "value": null, "after": null }, { "label": "Petrochemical Industries Company", "value": null, "after": null }, { "label": "QAPCO", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "alba", "title": "Aluminium Bahrain (Alba)", "scope": "Aluminium Bahrain B.S.C.; retained evidence concerns Bahrain smelter and named central functions, not all newly acquired operations. Maaden shareholding creates overlap.", "claims": [ { "id": "MAN-ALBA02", "claim": "Alba reports a three-day AI course underway for 35 employees, covering generative AI and prompting.", "url": "https://www.albasmelter.com/en/article/alba-partners-with-nvtcs-randd-arm-to-upskill-employees-with-artificial-intelligence-knowledge", "locator": "Body paragraphs on participants, duration and course topics", "level": 3, "eligible": true, "limitation": "Claim scope: Aluminium Bahrain B.S.C.; retained evidence concerns Bahrain smelter and named central functions, not all newly acquired operations. Maaden shareholding creates overlap.. 35 participants are not 35 certified completions; no productivity effect established. Application: internal-operating-practice.", "rationale": "A CEO visit to students confirms the course was underway; duration, participation and topics are specified. Completion and proficiency are not established." }, { "id": "MAN-ALBA04", "claim": "Alba describes deployed AI using incident patterns and risk drivers to identify high-risk operating areas.", "url": "https://www.albasmelter.com/uploads/albayan2026_q1.pdf", "locator": "Physical24, printed46–47, AI safety article", "level": 3, "eligible": true, "limitation": "Claim scope: Aluminium Bahrain B.S.C.; retained evidence concerns Bahrain smelter and named central functions, not all newly acquired operations. Maaden shareholding creates overlap.. No verified predictive accuracy or causal safety improvement. Application: internal-operating-practice.", "rationale": "Issuer newsletter explicitly describes deployment using incident patterns to prioritise smelter risk areas; accuracy and safety effects remain unverified." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Locate the production constraint." }, { "companyId": "sabic", "title": "SABIC", "scope": "SABIC group global manufacturing; claim scope restricted to named functions. Aramco ownership creates cross-sector overlap.", "claims": [ { "id": "MAN-SABIC01", "claim": "AI process and predictive analytics tools are deployed at a majority of global manufacturing plants.", "url": "https://www.sabic.com/en/Images/SABIC-Integrated-Annual-Report-2025-EN_tcm1010-49452.pdf", "locator": "Digital Transformation, printed60 / physical60, deployment paragraph", "level": 3, "eligible": true, "limitation": "Claim scope: SABIC group global manufacturing; claim scope restricted to named functions. Aramco ownership creates cross-sector overlap.. Mixed predictive analytics/AI tool portfolio; not every tool is AI or every plant covered. Application: internal-operating-practice.", "rationale": "Current deployment and majority-of-plants coverage are explicit for a mixed AI/predictive-analytics portfolio; no all-plants or AI-only coverage inferred." }, { "id": "MAN-SABIC03", "claim": "GenAI platforms and AI infrastructure connected previously isolated laboratory systems.", "url": "https://www.sabic.com/en/Images/SABIC-Integrated-Annual-Report-2025-EN_tcm1010-49452.pdf", "locator": "Digital Transformation, printed60 / physical60, R&D paragraph", "level": 2, "eligible": true, "limitation": "Claim scope: SABIC group global manufacturing; claim scope restricted to named functions. Aramco ownership creates cross-sector overlap.. No documented knowledge maintenance, retrieval evaluation or lineage controls. Application: internal-operating-practice.", "rationale": "GenAI platforms and infrastructure are said to have connected laboratory systems; knowledge maintenance or traceability is not established." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Locate the production constraint." }, { "companyId": "equate", "title": "EQUATE", "scope": "EQUATE group; not unrelated academic projects named EQUATE. PIC equity ownership creates overlap.", "claims": [], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Locate the production constraint." } ], "agenda": [ { "question": "Locate the production constraint.", "evidence": "Named line/site, quality or maintenance decision and accepted operating baseline." }, { "question": "Inspect integration with plant control.", "evidence": "Action authority, operator override, validation and safety boundaries." }, { "question": "Distinguish a useful model from a scaled practice.", "evidence": "Utilization, covered assets, retraining ownership and reproducible results." } ], "sources": [ { "title": "Alba and NAIRDC artificial intelligence collaboration", "publisher": "Alba", "date": "2023-07-12", "url": "https://www.albasmelter.com/en/article/alba-joins-forces-with-nasser-artificial-intelligence-research-and-development-centre-to-advance-in-ai", "locator": "Body paragraphs on completed Green Anode Density Prediction project", "originalUrl": "https://www.albasmelter.com/en/article/alba-joins-forces-with-nasser-artificial-intelligence-research-and-development-centre-to-advance-in-ai", "access": "full-source-read", "relationship": "corporate-operating-announcement", "id": "S01" }, { "title": "Alba partners with NVTC R&D arm to upskill employees", "publisher": "Alba", "date": "2024-08-21", "url": "https://www.albasmelter.com/en/article/alba-partners-with-nvtcs-randd-arm-to-upskill-employees-with-artificial-intelligence-knowledge", "locator": "Body paragraphs on participants, duration and course topics", "originalUrl": "https://www.albasmelter.com/en/article/alba-partners-with-nvtcs-randd-arm-to-upskill-employees-with-artificial-intelligence-knowledge", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-announcement", "id": "S02" }, { "title": "Alba adopts AI-powered Seeq platform", "publisher": "Alba", "date": "2025-04-27", "url": "https://www.albasmelter.com/en/article/alba-sets-regional-benchmark-as-first-middle-east-smelter-to-adopt-ai-powered-seeq-platform", "locator": "Body paragraphs on agreement and future integration", "originalUrl": "https://www.albasmelter.com/en/article/alba-sets-regional-benchmark-as-first-middle-east-smelter-to-adopt-ai-powered-seeq-platform", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-announcement", "id": "S03" }, { "title": "Al Bayan Q1 2026", "publisher": "Alba", "date": "2026", "url": "https://www.albasmelter.com/uploads/albayan2026_q1.pdf", "locator": "Physical24, printed46–47, AI safety article", "originalUrl": "https://www.albasmelter.com/uploads/albayan2026_q1.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "company-magazine", "id": "S04" }, { "title": "Updated Code of Conduct disclosure", "publisher": "Alba", "date": "2026-09-23", "url": "https://www.albasmelter.com/uploads/Disclosure_of_MI_Alba_Reinforces_Ethical_Leadership_Through_Updated_Code_of_Conduct_23.09.2026_1.pdf", "locator": "Physical2, launch and new guidance paragraphs", "originalUrl": "https://www.albasmelter.com/uploads/Disclosure_of_MI_Alba_Reinforces_Ethical_Leadership_Through_Updated_Code_of_Conduct_23.09.2026_1.pdf", "access": "full-source-read", "relationship": "regulatory-disclosure", "id": "S05" }, { "title": "SABIC Integrated Annual Report 2025", "publisher": "SABIC", "date": "2026", "url": "https://www.sabic.com/en/Images/SABIC-Integrated-Annual-Report-2025-EN_tcm1010-49452.pdf", "locator": "Digital Transformation, printed60 / physical60, deployment paragraph", "originalUrl": "https://www.sabic.com/en/Images/SABIC-Integrated-Annual-Report-2025-EN_tcm1010-49452.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S06" }, { "title": "EGA designated Industry 4.0 Global Lighthouse", "publisher": "EGA", "date": "2025-01-15", "url": "https://media.ega.ae/ega-designated-industry-40-global-lighthouse-by-world-economic-forum-in-uae-and-world-aluminium-industry-first/", "locator": "Body paragraph describing real-time carbon-anode quality assessment", "originalUrl": "https://media.ega.ae/ega-designated-industry-40-global-lighthouse-by-world-economic-forum-in-uae-and-world-aluminium-industry-first/", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-announcement", "id": "S07" }, { "title": "EX technology pilot and Industry 4.0 capabilities", "publisher": "EGA", "date": "2024-12-09", "url": "https://media.ega.ae/egas-ex-technology-pilot-to-be-used-to-prove-most-advanced-industry-40-capabilities-in-aluminium-industry/", "locator": "Body paragraphs on ten pilot reduction cells and autonomous-control goal", "originalUrl": "https://media.ega.ae/egas-ex-technology-pilot-to-be-used-to-prove-most-advanced-industry-40-capabilities-in-aluminium-industry/", "access": "full-source-read", "relationship": "corporate-operating-announcement", "id": "S08" }, { "title": "Tech innovations", "publisher": "Maaden", "date": "Undated", "url": "https://www.maaden.com/our-innovation/tech-innovations", "locator": "Exploration technology paragraph", "originalUrl": "https://www.maaden.com/our-innovation/tech-innovations", "access": "full-source-read", "relationship": "corporate-operating-page", "id": "S09" }, { "title": "Partnerships", "publisher": "Maaden", "date": "Undated", "url": "https://www.maaden.com/discover-us/partnerships", "locator": "Ivanhoe Electric partnership paragraph", "originalUrl": "https://www.maaden.com/discover-us/partnerships", "access": "full-source-read", "relationship": "corporate-operating-page", "id": "S10" }, { "title": "Ducab launches Blade smart factory project", "publisher": "Ducab", "date": "2023-06-13", "url": "https://ducab.com/media-center/press-releases/ducab-group-expedites-operations-with-the-launch-of-blade-the-smart-factory-project-of-the-future-based-on-industry-4-0-technologies/", "locator": "Body paragraph on AI software and planning orders", "originalUrl": "https://ducab.com/media-center/press-releases/ducab-group-expedites-operations-with-the-launch-of-blade-the-smart-factory-project-of-the-future-based-on-industry-4-0-technologies/", "access": "full-source-read", "relationship": "corporate-operating-announcement", "id": "S11" } ], "body": "# The factory decision is the test\n## Manufacturing & Materials — Mayden AI Native Index research edition\n**Evidence cutoff: 10 October 2026. Ten purposefully selected entities across five markets. Public evidence, not a measurement of actual enterprise maturity.**\n\nA factory can announce an AI partnership without changing a production decision. It can also run a useful predictive model for years without describing it publicly. An index that treats either possibility carelessly will reward communications activity and punish silence. The manufacturing research therefore asks a narrower, more useful question: what operating practice can an external reader actually trace to a primary source, and how far does that source support the claim?\n\nThe strongest cases in this corpus describe a decision close to production: identifying an anode-quality issue, prioritising a safety risk, or using AI in an order-planning workflow. These are more informative than an undifferentiated count of digital initiatives. They reveal where information enters work. They do not establish that the whole enterprise is AI-native, that a model is accurate, or that a reported improvement was caused by AI.\n\nThis report turns those cases into questions executives can use. Its central proposition is an editorial interpretation: an AI-native manufacturer would need repeatable arrangements for deploying, supervising, maintaining and learning from models across operating conditions. A successful isolated application is evidence of an application. The additional organisational arrangements remain an empirical question.\n\n## What the cohort covers\n\nThe ten entities are SABIC and Maaden in Saudi Arabia; Emirates Global Aluminium and Ducab in the UAE; Industries Qatar and QAPCO in Qatar; EQUATE and Petrochemical Industries Company in Kuwait; and Aluminium Bahrain and GARMCO in Bahrain. Selection creates a small, purposeful view of chemicals, mining, metals and industrial manufacturing. It is not a census, a sample of the largest ten firms, or a representative estimate of GCC adoption. Oman is outside this five-market expansion.\n\nCorporate boundaries are consequential. Industries Qatar has exposure to QAPCO; PIC has an ownership relationship with EQUATE. These are not independent observations of entirely separate economic systems. Maaden's investment in Alba also creates overlap, while SABIC has an ownership relationship with an energy-sector company covered elsewhere in the Index. No operating claim is transferred from an investee to its shareholder simply because that relationship exists. Alba's retained cases concern its Bahrain operation or a specified central function, rather than every operation in its changing corporate perimeter.\n\nThe cohort contains both integrated operators and investment-group structures. Comparing their disclosure evidence can be useful if that distinction is visible. Comparing the resulting numbers as if they represented identical factories would be misleading. Readers should open the company scope and supporting passage before interpreting a difference.\n\n## Production evidence is strongest when the unit is visible\n\nSABIC's 2025 report states that its AI and predictive tools are deployed across a majority of global manufacturing plants. It separately describes pilots in demand forecasting, pricing and planning; connected laboratory systems; workforce adoption programmes; and risk assessments that include AI. The separation matters: a broadly deployed production-tool portfolio does not turn each commercial pilot into a production system. Nor does it establish autonomous control. [SABIC Integrated Annual Report 2025, digital and cybersecurity sections, printed pages 60–62](https://www.sabic.com/en/Images/SABIC-Integrated-Annual-Report-2025-EN_tcm1010-49452.pdf)\n\nThe executive lesson is to insist on a denominator with a stable definition. “Most plants” is more informative than “many use cases,” but the next discussion still needs to identify eligible sites, processes covered, usage frequency and retirement of unsuccessful models. A portfolio count can rise because a company creates more prototypes, because it divides applications differently, or because it has genuinely expanded productive use. Only the last interpretation necessarily changes operating capability.\n\nEGA provides a more localised example. Its January 2025 lighthouse announcement describes implemented AI vision for real-time carbon-anode quality assessment. That is a named production task with a recognisable observation mechanism. The surrounding announcement also discusses a broader Industry 4.0 portfolio. Those portfolio totals and general digital-skilling numbers should not be relabelled as AI totals. [EGA announcement, 15 January 2025](https://media.ega.ae/ega-designated-industry-40-global-lighthouse-by-world-economic-forum-in-uae-and-world-aluminium-industry-first/)\n\nA quality model creates several different questions. Does it flag an item for inspection, change a process setting, or reject output automatically? Who resolves disagreement between the model and an experienced operator? What happens when raw-material characteristics change? The public case supports the existence of a particular application; these questions define the deeper evidence needed for an organisational benchmark.\n\nDucab's Blade announcement identifies AI software used in manufacturing-order planning. The same release describes expected improvements in production and downtime. This review retains the stated planning use and keeps those expected benefits separate. It does not present targets as achieved results or treat every sensor in the wider project as AI. [Ducab Blade announcement, 13 June 2023](https://ducab.com/media-center/press-releases/ducab-group-expedites-operations-with-the-launch-of-blade-the-smart-factory-project-of-the-future-based-on-industry-4-0-technologies/)\n\nFor a planning application, usefulness depends on more than forecast accuracy. An apparently better plan may violate an equipment constraint, create downstream inventory, or increase schedule instability. A serious company interview should connect the recommendation to the supervisor who accepts it and the production consequences that person monitors.\n\n## A pilot deserves recognition without being called a rollout\n\nEGA's December 2024 EX announcement is explicit about an experimental facility and an intended future autonomous production system. It describes the construction of pilot reduction cells, prospective AI capabilities and a longer industrialisation horizon. Existing developed use cases are discussed alongside these ambitions. The future autonomous-control objective remains a different claim from an implemented quality-inspection application. [EGA EX pilot announcement, 9 December 2024](https://media.ega.ae/egas-ex-technology-pilot-to-be-used-to-prove-most-advanced-industry-40-capabilities-in-aluminium-industry/)\n\nThis distinction is useful even when a later announcement confirms that a physical pilot has started producing metal. Commissioning the industrial asset does not, by itself, establish that a particular AI control is operating routinely. The evidence needed to cross that boundary is a dated account of the AI system entering service, its operating envelope and the accountable owner.\n\nAlba's 2023 announcement describes completion of a green-anode density prediction model and a dashboard developed with NAIRDC. The review treats this as a specific completed development or pilot rather than assuming sustained production deployment. Separately, the 2025 Seeq announcement contains future integration language despite its adoption-oriented headline. It is retained as a named commitment. [Alba–NAIRDC project, 12 July 2023](https://www.albasmelter.com/en/article/alba-joins-forces-with-nasser-artificial-intelligence-research-and-development-centre-to-advance-in-ai); [Alba–Seeq announcement, 27 April 2025](https://www.albasmelter.com/en/article/alba-sets-regional-benchmark-as-first-middle-east-smelter-to-adopt-ai-powered-seeq-platform)\n\nExecutives should not read a lower evidence stage as criticism of experimentation. A well-run pilot can prevent an expensive operational mistake. The relevant management question is whether there is a disciplined transition: a baseline, a success criterion, a safety review, an owner, a decision date and a route to withdrawal. A benchmark should be able to recognise that discipline without pretending that intent has already produced enterprise-wide change.\n\n## The organisation appears in the less theatrical applications\n\nAlba's first-quarter 2026 company magazine describes AI used to identify high-risk operating areas from incident patterns and risk drivers. It also describes a Scope of Work assistant that helps prepare structured contract documents for an approval process. A conduct-policy chatbot is described with less conclusive deployment language. These are different levels of support, even though they appear in the same publication. The safety narrative does not establish a causal reduction in accidents; the drafting assistant does not establish autonomous contracting. [Alba Al Bayan Q1 2026, physical pages 24–25](https://www.albasmelter.com/uploads/albayan2026_q1.pdf)\n\nThese examples direct attention beyond the furnace or production line. An enterprise changes when a tool becomes part of how work is specified, reviewed and escalated. Contract drafting can affect procurement quality. Safety prioritisation can influence the allocation of supervisory attention. Both deserve scrutiny because their consequences depend on human interpretation and follow-through.\n\nThe next interview should ask for a recent decision reconstructed end to end. Which information entered the tool? Which recommendation emerged? Who inspected it? What changed as a result? Could the organisation reproduce the reasoning after a dispute or incident? A useful answer would identify records and responsibilities. A generic assurance that humans remain in control would leave the most important details unresolved.\n\nAlba's August 2024 course announcement provides another bounded organisational signal: a three-day AI programme underway for 35 employees, with specified generative-AI and prompting topics. Those are participants in an announced course, not independently verified completions or evidence that the workforce has mastered AI. [Alba training announcement, 21 August 2024](https://www.albasmelter.com/en/article/alba-partners-with-nvtcs-randd-arm-to-upskill-employees-with-artificial-intelligence-knowledge)\n\nA training measure should distinguish attendance, assessed competence and changed work. Different roles need different capabilities. A process engineer may need to recognise model drift; a procurement specialist may need to verify generated specifications; a manager may need to judge when automation is inappropriate. Counting all three people as identically trained loses information that matters to performance.\n\nAlba also disclosed an updated Code of Conduct with responsible-AI guidance in September 2026. This supports the presence of an issued policy. The disclosure does not supply enough detail to assess whether particular model controls have been applied effectively. [Alba regulatory disclosure, 23 September 2026](https://www.albasmelter.com/uploads/Disclosure_of_MI_Alba_Reinforces_Ethical_Leadership_Through_Updated_Code_of_Conduct_23.09.2026_1.pdf)\n\n## Exploration is a distinct operating setting\n\nMaaden's technology and partnership pages describe AI and data modelling in mineral exploration, including work with Ivanhoe Electric. They are useful primary descriptions of an operating activity and a named collaboration. They are live undated pages, however, and do not provide a deployment date, model-performance evaluation or a clear coverage denominator. [Maaden technology page](https://www.maaden.com/our-innovation/tech-innovations); [Maaden partnerships](https://www.maaden.com/discover-us/partnerships)\n\nExploration is not the same decision environment as continuous process control. Its evidence can involve sparse observations, uncertain geological interpretation and long feedback cycles. A common organisational framework needs sector-specific questions. The principle is to compare how an organisation makes and governs AI-supported decisions, while resisting the temptation to compare unlike model metrics.\n\nAn exploration team could explain how recommendations change survey priorities, how new field observations are incorporated, and how unsuccessful predictions are documented. A smelter could explain how a model reacts to process changes and how operators override recommendations. These are different technical accounts of a common organisational obligation: keeping a decision system reliable in use.\n\n## What remains unknown is part of the result\n\nIndustries Qatar's examined annual-report passages mention AI in the economic and market backdrop. Such language does not demonstrate an implemented group practice. QAPCO's report archive and older PDF links presented access problems; indexed references to a predictive-maintenance plan were not promoted into verified source evidence. EQUATE's accessible executive summary did not yield a qualifying AI practice in the passages searched. For PIC and GARMCO, the bounded company-site search and identity material did not provide qualifying operating evidence.\n\nThese findings do not establish that those companies lack AI. They establish the limits of this research corpus. Report access failures, different disclosure customs, corporate structure and search vocabulary can all affect visibility. The correct label is unknown within the bounded search, accompanied by the search log.\n\nA reader must also distinguish a failed retrieval from a source that was read and found inapplicable. The former leaves an unresolved lead. The latter may rule out a particular interpretation, such as treating a general technology investment as AI. The dataset records these differences so that a subsequent researcher can make a targeted correction rather than repeat the entire search.\n\n## Questions that can turn disclosure research into a benchmark\n\nA proprietary company assessment should start with two material decisions selected by the firm and one selected by the interviewer. This reduces the chance that the interview covers only a flagship demonstration. For each decision, the respondent should identify the task, operating owner, affected sites, user population, deployment date and current stage. The assessment should then request a small evidence sample rather than unrestricted access to operational systems.\n\nThe questions should cover nine dimensions with sector-specific anchors. Where is AI materially involved in core work? Which employees have demonstrated role-relevant competence? What responsibilities sit with technology partners? How do outputs enter systems or work queues? Can a material decision be reconstructed from retained records? What evidence determines whether an experiment proceeds? Which actions can the system take, and who can stop it? How is operational knowledge maintained? Which AI-specific controls have actually been applied?\n\nThese are proposed interview questions, not findings about the cohort. They should be tested with operators for clarity before being used to create a validated organisational measure. A polished questionnaire alone cannot establish validity. Mayden would still need consistent coding, independent adjudication, evidence of reliable responses and investigation of whether the resulting dimensions relate to meaningful external outcomes.\n\n## Reading the manufacturing evidence exhibits\n\nThe heatmap compares the ten selected entities across nine dimensions. U means no qualifying evidence was located in the bounded corpus; it does not mean the capability is absent. Level 1 denotes a specific commitment or pilot, level 2 an implemented practice, and level 3 an implemented practice with concrete operating detail. Each cell takes the highest supported, adjudicated level for that dimension. Several descriptions of one application do not accumulate extra points.\n\nThe points comparison uses a fixed denominator of 27: nine dimensions with a maximum of three points each. The displayed value is the supported total divided by 27 and multiplied by 100, rounded to a whole number. Unknown cells add no documented points, but remain unknown in interpretation. A higher total therefore describes a stronger public evidence record under these rules; it is not a percentage of enterprise maturity or proof of better factory performance. Equal underlying totals share a position, and records without qualifying evidence remain unranked.\n\nThe anchor-count exhibit separates commitments or pilots, implemented practices, and implemented practices with operating detail. These are claim counts, not counts of independent projects, factories or companies. One operating application may support more than one dimension. This helps explain the distinction between an implemented inspection practice and a future autonomous-control objective without treating every mention of AI as equivalent.\n\nThe source-dependence exhibit removes one source URL at a time and recalculates the points, showing the result when the most influential URL is removed. A large decrease identifies concentration in the disclosed evidence, often in an annual report. It does not simulate a factory losing its AI capability. Different URLs can also contain related company assertions, so a smaller decrease is not proof of independent corroboration. Read the chart alongside the entity scopes: shareholder and investee records can overlap even when their source URLs differ.\n\nThe practical outcome of this edition is an inspectable research base and a sharper conversation with manufacturers. A future benchmark becomes credible by preserving these distinctions over time: company versus site, intention versus operation, activity versus outcome, and undisclosed versus absent. The index should make it easier to ask the next difficult question, not make the available evidence appear more complete than it is.\n\n**Evidence and reproducibility.** The companion manufacturing-evidence.json contains company scopes, source URLs, passage locators, proposed evidence levels and per-company search limits. Separate AI-assisted source challenge is recorded separately; the released calculations use the retained adjudicated decisions. Company assertions remain company assertions. This edition includes no confidential company access, respondent survey, statistical meta-analysis or validation of actual AI-native organisational maturity.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/manufacturing-public-evidence.pdf", "cover": "/research/index/2026-10/manufacturing-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 89172, "pdfSha256": "edc2647d6dd69a34e5ffc8347d1d479bffcf46aee7ae08f5fc3bb84cfd01c2cd", "pdfPages": 17 }, { "slug": "retail-public-evidence", "number": 7, "title": "AI in Gulf consumer businesses: follow the operating workflow", "shortTitle": "AI behind the consumer experience", "dek": "Demand, inventory and service across ten selected retail and consumer entities.", "sector": "Retail & Consumer", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Examine who approves replenishment. LuLu reports UAE forecasting and replenishment with planner review and defined thresholds. Test the exception process and establish which products and stores are covered.", "Require more than a finished document. Mezzan reports using generative AI to prepare a supplier questionnaire. Assess responses and subsequent decisions; creation alone establishes neither supplier compliance nor completed deployment to every intended supplier.", "Separate a new interaction from a completed service. LuLu describes an assistant connected to customer and order systems; MAF’s City+ case concerns a named historical store. Measure successful transactions and verify current operating scope." ], "executiveSources": [ { "claimIds": [ "RET-LULU01", "RET-LULU02", "RET-LULU03" ], "sources": [ { "claimId": "RET-LULU01", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "sourceTitle": "LuLu Retail Integrated Annual Report 2025" }, { "claimId": "RET-LULU02", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "sourceTitle": "LuLu Retail Integrated Annual Report 2025" }, { "claimId": "RET-LULU03", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "sourceTitle": "LuLu Retail Integrated Annual Report 2025" } ] }, { "claimIds": [ "RET-MEZZAN01" ], "sources": [ { "claimId": "RET-MEZZAN01", "url": "https://www.mezzan.com/wp-content/uploads/2026/08/Mezzan-Sustainability-Report-2025-English-.pdf", "sourceTitle": "Mezzan Sustainability Report 2025" } ] }, { "claimIds": [ "RET-LULU05", "RET-MAF05" ], "sources": [ { "claimId": "RET-LULU05", "url": "https://www.luluretail.com/media/news/lulu-hypermarket-launches-new-ai-powered-customer-experiences-on-whatsapp-in-collaboration-with-yellowai", "sourceTitle": "AI-powered customer experiences on WhatsApp with Yellow.ai" }, { "claimId": "RET-MAF05", "url": "https://www.majidalfuttaim.com/en/media-centre/press-releases/detail/2021/09/carrefour-city-plus", "sourceTitle": "Carrefour City+ launch" } ] } ], "evidenceAgeContext": "Dated sources run from 2021-09-06 to 2026-10-01. 0 of 17 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "Where does the retail workflow actually finish?", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 18, "label": "recorded claims" }, { "value": 17, "label": "eligible claims" }, { "value": 4, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "Where does the retail workflow actually finish?", "subtitle": "Three separate completion tests; a generated output is not automatically an accepted operating result.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "Forecast to replenishment", "unit": "Approved order recommendation", "finding": "LuLu says UAE planning teams review AI outputs and exceptions under defined rules and thresholds.", "boundary": "Reported approval workflow; no inference to every market.", "implication": "Does the accepted order improve availability without shifting waste elsewhere?", "claimIds": [ "RET-LULU01", "RET-LULU02" ], "sources": [ { "claimId": "RET-LULU01", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "sourceTitle": "LuLu Retail Integrated Annual Report 2025" }, { "claimId": "RET-LULU02", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "sourceTitle": "LuLu Retail Integrated Annual Report 2025" } ] }, { "label": "Conversation to service", "unit": "Customer or order request", "finding": "LuLu describes Salem on WhatsApp integrated with customer and order systems.", "boundary": "Integration alone does not establish successful resolution.", "implication": "Is the transaction completed, escalated or repeated by the customer?", "claimIds": [ "RET-LULU05" ], "sources": [ { "claimId": "RET-LULU05", "url": "https://www.luluretail.com/media/news/lulu-hypermarket-launches-new-ai-powered-customer-experiences-on-whatsapp-in-collaboration-with-yellowai", "sourceTitle": "AI-powered customer experiences on WhatsApp with Yellow.ai" } ] }, { "label": "Draft to supplier evidence", "unit": "Completed and reviewed questionnaire", "finding": "Mezzan reports using GenAI to develop an ESG questionnaire for targeted high-spend suppliers.", "boundary": "Target population is not confirmed responses or compliance.", "implication": "Which responses change a supplier decision after human review?", "claimIds": [ "RET-MEZZAN01" ], "sources": [ { "claimId": "RET-MEZZAN01", "url": "https://www.mezzan.com/wp-content/uploads/2026/08/Mezzan-Sustainability-Report-2025-English-.pdf", "sourceTitle": "Mezzan Sustainability Report 2025" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "lulu-retail", "label": "LuLu Retail Holdings PLC", "values": [ 3, null, 3, 3, null, null, 3, null, 3 ], "score": 56, "rank": 1 }, { "id": "majid-al-futtaim", "label": "Majid Al Futtaim", "values": [ 3, 3, 3, 3, null, null, null, 2, null ], "score": 52, "rank": 2 }, { "id": "al-meera", "label": "Al Meera Consumer Goods Company", "values": [ null, null, 1, 3, null, null, null, null, null ], "score": 15, "rank": 3 }, { "id": "mezzan", "label": "Mezzan Holding", "values": [ null, null, null, 3, null, null, null, null, null ], "score": 11, "rank": 4 }, { "id": "baladna", "label": "Baladna", "values": [ 1, null, null, 1, null, null, null, null, null ], "score": 7, "rank": 5 }, { "id": "extra", "label": "United Electronics Company (eXtra)", "values": [ null, null, null, null, null, null, null, null, 2 ], "score": 7, "rank": 5 }, { "id": "alshaya", "label": "Alshaya Group", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "bmmi", "label": "BMMI", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "jarir", "label": "Jarir Marketing Company", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "trafco", "label": "Trafco Group", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "LuLu Retail Holdings PLC", "value": 56, "secondary": 5, "rank": 1 }, { "label": "Majid Al Futtaim", "value": 52, "secondary": 5, "rank": 2 }, { "label": "Al Meera Consumer Goods Company", "value": 15, "secondary": 2, "rank": 3 }, { "label": "Mezzan Holding", "value": 11, "secondary": 1, "rank": 4 }, { "label": "Baladna", "value": 7, "secondary": 2, "rank": 5 }, { "label": "United Electronics Company (eXtra)", "value": 7, "secondary": 1, "rank": 5 }, { "label": "Alshaya Group", "value": null, "secondary": 0, "rank": null }, { "label": "BMMI", "value": null, "secondary": 0, "rank": null }, { "label": "Jarir Marketing Company", "value": null, "secondary": 0, "rank": null }, { "label": "Trafco Group", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 17; one practice may support multiple distinct themes", "maximum": 12, "rows": [ { "label": "Specific commitment / pilot", "value": 3 }, { "label": "Implemented practice", "value": 2 }, { "label": "Implemented + operating detail", "value": 12 } ], "note": "Source: claim ledger. 1 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "LuLu Retail Holdings PLC", "value": 55.6, "after": 22.22 }, { "label": "Majid Al Futtaim", "value": 51.9, "after": 22.22 }, { "label": "Al Meera Consumer Goods Company", "value": 14.8, "after": 3.7 }, { "label": "Mezzan Holding", "value": 11.1, "after": 0.0 }, { "label": "Baladna", "value": 7.4, "after": 0.0 }, { "label": "United Electronics Company (eXtra)", "value": 7.4, "after": 0.0 }, { "label": "Alshaya Group", "value": null, "after": null }, { "label": "BMMI", "value": null, "after": null }, { "label": "Jarir Marketing Company", "value": null, "after": null }, { "label": "Trafco Group", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "lulu-retail", "title": "LuLu Retail Holdings PLC", "scope": "Listed GCC retail group; historical LuLu brand claims limited to GCC operations, excluding non-GCC group businesses.", "claims": [ { "id": "RET-LULU01", "claim": "LuLu reports AI forecasting and replenishment operational in the UAE using sales, seasonality, promotion and store-pattern inputs.", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "locator": "Physical105, printed104, AI adoption section", "level": 3, "eligible": true, "limitation": "Claim scope: Listed GCC retail group; historical LuLu brand claims limited to GCC operations, excluding non-GCC group businesses.. Other-country rollout underway is not completed; no verified benefit size. Application: internal-operating-practice.", "rationale": "Annual2025 printed104/physical105 describes operational UAE forecasting/replenishment with sales, seasonality, promotion and store inputs. Other-market rollout remains underway." }, { "id": "RET-LULU02", "claim": "Planning teams review AI outputs and exceptions, approving replenishment recommendations within defined rules and thresholds.", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "locator": "Physical105, printed104, oversight and automation paragraphs", "level": 3, "eligible": true, "limitation": "Claim scope: Listed GCC retail group; historical LuLu brand claims limited to GCC operations, excluding non-GCC group businesses.. Bounded automation, not fully autonomous buying; thresholds not numerically disclosed. Application: internal-operating-practice.", "rationale": "Same annual passage explicitly describes planning-team review, exception handling, approvals and bounded automated operation. This establishes specified decision controls, not unrestricted autonomy." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Trace the demand signal into an action." }, { "companyId": "majid-al-futtaim", "title": "Majid Al Futtaim", "scope": "Majid Al Futtaim group; claims restricted to Retail, named store, HR or training scope. Not Carrefour SA worldwide.", "claims": [ { "id": "RET-MAF01", "claim": "MAF launched an AI Academy with separate hands-on, technical and senior-leader learning paths.", "url": "https://www.majidalfuttaim.com/docs/default-source/reports/maf-sus-pages-full_final.pdf?sfvrsn=7c1de58c_2", "locator": "Physical/printed80–81, AI Academy and GO/DEVELOP/ELEVATE pathways", "level": 3, "eligible": true, "limitation": "Claim scope: Majid Al Futtaim group; claims restricted to Retail, named store, HR or training scope. Not Carrefour SA worldwide.. Other academy satisfaction and coverage figures cannot be attributed to AI Academy. Application: internal-operating-practice.", "rationale": "Original2024 report pages80-81 describes launched AI Academy with three role-specific pathways. Do not transfer other academy participation or satisfaction figures into AI coverage or proficiency." }, { "id": "RET-MAF02", "claim": "MAF describes an AI pricing engine considering scenarios, recommendations and promotions in grocery retail.", "url": "https://www.majidalfuttaim.com/docs/default-source/reports/maf-sus-pages-full_final.pdf?sfvrsn=7c1de58c_2", "locator": "Physical/printed70, AI customer-service spotlight", "level": 3, "eligible": true, "limitation": "Claim scope: Majid Al Futtaim group; claims restricted to Retail, named store, HR or training scope. Not Carrefour SA worldwide.. Portfolio SKU and country totals are not AI coverage denominators; no autonomous price-setting authority established. Application: internal-operating-practice.", "rationale": "Original report page70 describes an implemented pricing engine considering scenarios, recommendations and promotions. No autonomous pricing authority or measured causal impact inferred." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Trace the demand signal into an action." }, { "companyId": "alshaya", "title": "Alshaya Group", "scope": "M.H. Alshaya operating group; retail-media service evidence not inherited by every franchise brand.", "claims": [ { "id": "RET-ALSHAYA01", "claim": "Alshaya Media Connect advertises an AI-enabled retail-media service combining loyalty purchase data and intent signals.", "url": "https://www.alshaya.com/en/alshaya-media-connect/", "locator": "Offsite Display & Video service description", "level": null, "eligible": false, "limitation": "Claim scope: M.H. Alshaya operating group; retail-media service evidence not inherited by every franchise brand.. Service catalogue alone does not establish actual operating deployment; no transfer to franchise stores. Application: operator-service-offering. Exclusion: New claim has no explicit accepted review decision. No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass.", "rationale": "Full undated service catalogue advertises AI-enabled retail-media offering. It does not establish actual operating deployment or a specific internal implementation commitment. Product availability alone is insufficient under the public-method boundary. Exclusion: New claim has no explicit accepted review decision. No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Trace the demand signal into an action." } ], "agenda": [ { "question": "Trace the demand signal into an action.", "evidence": "Forecast inputs, replenishment decision, store/channel scope and override ownership." }, { "question": "Measure the whole service outcome.", "evidence": "Availability, waste, margins and service quality with relevant denominators." }, { "question": "Inspect customer and employee controls.", "evidence": "Data permissions, contested recommendations and escalation when the system is wrong." } ], "sources": [ { "title": "LuLu Retail Integrated Annual Report 2025", "publisher": "LuLu Retail", "date": "2026", "url": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "locator": "Physical105, printed104, AI adoption section", "originalUrl": "https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S01" }, { "title": "AI-powered robotic shopping experience in UAE stores", "publisher": "LuLu Retail", "date": "2026-10-01", "url": "https://www.luluretail.com/media/news/lulu-brings-ai-powered-robotic-shopping-experience-to-uae-stores/", "locator": "Body deployment and Odigo robot capabilities paragraphs", "originalUrl": "https://www.luluretail.com/media/news/lulu-brings-ai-powered-robotic-shopping-experience-to-uae-stores/", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-announcement", "id": "S02" }, { "title": "AI-powered customer experiences on WhatsApp with Yellow.ai", "publisher": "LuLu Retail", "date": "2023-02-28", "url": "https://www.luluretail.com/media/news/lulu-hypermarket-launches-new-ai-powered-customer-experiences-on-whatsapp-in-collaboration-with-yellowai", "locator": "Body launch, CRM/order integration and live-agent handoff paragraphs", "originalUrl": "https://www.luluretail.com/media/news/lulu-hypermarket-launches-new-ai-powered-customer-experiences-on-whatsapp-in-collaboration-with-yellowai", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-announcement", "id": "S03" }, { "title": "Majid Al Futtaim Sustainability Report 2024", "publisher": "Majid Al Futtaim", "date": "2025", "url": "https://www.majidalfuttaim.com/docs/default-source/reports/maf-sus-pages-full_final.pdf?sfvrsn=7c1de58c_2", "locator": "Physical/printed80–81, AI Academy and GO/DEVELOP/ELEVATE pathways", "originalUrl": "https://www.majidalfuttaim.com/docs/default-source/reports/maf-sus-pages-full_final.pdf?sfvrsn=7c1de58c_2", "access": "full-source-read | Review: full-source-read", "relationship": "sustainability-report", "id": "S04" }, { "title": "Carrefour City+ launch", "publisher": "Majid Al Futtaim", "date": "2021-09-06", "url": "https://www.majidalfuttaim.com/en/media-centre/press-releases/detail/2021/09/carrefour-city-plus", "locator": "Launch body describing Mall of the Emirates store and checkout flow", "originalUrl": "https://www.majidalfuttaim.com/en/media-centre/press-releases/detail/2021/09/carrefour-city-plus", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-announcement", "id": "S05" }, { "title": "Precision Media and Advertima AI rollout", "publisher": "Majid Al Futtaim", "date": "2025-05-13", "url": "https://www.majidalfuttaim.com/en/media-centre/press-releases/detail/2025/05/majid-al-futtaim-s-precision-media-and-advertima-set-new-retail-media-benchmark-with-ai-rollout-in-carrefour-stores", "locator": "Body paragraphs3 and8, hypermarket deployment and 3D sensors", "originalUrl": "https://www.majidalfuttaim.com/en/media-centre/press-releases/detail/2025/05/majid-al-futtaim-s-precision-media-and-advertima-set-new-retail-media-benchmark-with-ai-rollout-in-carrefour-stores", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-operating-announcement", "id": "S06" }, { "title": "Al Meera Annual Report 2024", "publisher": "Al Meera", "date": "2025", "url": "https://sacorporatewebprodqc001.blob.core.windows.net/public/documents/Al%20Meera%20AR%202024%2013%20APR%202025%20ENG%20FINAL%20(1).pdf", "locator": "Physical9–10, printed14–17, Chairman and Board reports", "originalUrl": "https://sacorporatewebprodqc001.blob.core.windows.net/public/documents/Al%20Meera%20AR%202024%2013%20APR%202025%20ENG%20FINAL%20(1).pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S07" }, { "title": "Al Meera partners with Zippin", "publisher": "Zippin", "date": "2022-08-22", "url": "https://www.getzippin.com/blog/al-meera-partners-with-zippin-to-open-first-checkout-free-stores-in-qatar", "locator": "Announcement body on planned stores and checkout-free technology", "originalUrl": "https://www.getzippin.com/blog/al-meera-partners-with-zippin-to-open-first-checkout-free-stores-in-qatar", "access": "full-source-read | Review: full-source-read", "relationship": "partner-announcement", "id": "S08" }, { "title": "Mezzan Sustainability Report 2025", "publisher": "Mezzan Holding", "date": "2026", "url": "https://www.mezzan.com/wp-content/uploads/2026/08/Mezzan-Sustainability-Report-2025-English-.pdf", "locator": "Physical53, Case Study: ESG Risk Screening for High-Spend Suppliers", "originalUrl": "https://www.mezzan.com/wp-content/uploads/2026/08/Mezzan-Sustainability-Report-2025-English-.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "sustainability-report", "id": "S09" }, { "title": "Baladna Annual Report 2025", "publisher": "Baladna", "date": "2026", "url": "https://baladna.com/storage/uploads/iKUuHjMbSSYZfRre3VkWieoWq.pdf", "locator": "Physical41, printed80–81, Looking ahead to2026", "originalUrl": "https://baladna.com/storage/uploads/iKUuHjMbSSYZfRre3VkWieoWq.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S10" }, { "title": "eXtra addendum to EGM invitation", "publisher": "United Electronics Company via Saudi Exchange", "date": "2025-04-30", "url": "https://www.saudiexchange.sa/wps/portal/saudiexchange/newsandreports/issuer-news/issuer-announcements/issuer-announcements-details/?anCat=1&anId=86829&cs=4003&locale=en", "locator": "Additional information, board-member remuneration item", "originalUrl": "https://www.saudiexchange.sa/wps/portal/saudiexchange/newsandreports/issuer-news/issuer-announcements/issuer-announcements-details/?anCat=1&anId=86829&cs=4003&locale=en", "access": "full-source-read", "relationship": "issuer-disclosure", "id": "S11" }, { "title": "Alshaya Media Connect service page", "publisher": "Alshaya", "date": "Undated", "url": "https://www.alshaya.com/en/alshaya-media-connect/", "locator": "Offsite Display & Video service description", "originalUrl": "https://www.alshaya.com/en/alshaya-media-connect/", "access": "full-source-read | Review: full-source-read", "relationship": "corporate-service-page", "id": "S12" }, { "title": "Jarir Annual Report 2025", "publisher": "Primary source as linked", "date": "See original", "url": "https://wps-media.jarir.com/wp-content/uploads/2026/04/JB_AnnualReport_2025_EN.pdf", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S13" } ], "body": "# The decision behind the shopping experience\n## Retail & Consumer — Mayden AI Native Index research edition\n**Evidence cutoff: 10 October 2026. Ten purposefully selected entities across five markets. Public evidence, not a measure of actual enterprise maturity.**\n\nThe visible face of retail AI is easy to photograph: a checkout-free store, a shopping robot or a conversational assistant. The more consequential organisational question is often less visible. How does a recommendation become a replenishment order? Who checks an AI-generated supplier questionnaire? Which person can reject an automated suggestion, and what information supports that decision?\n\nThis research examines those questions through primary disclosures from ten retail and consumer businesses. The cases show several ways AI enters work: forecasting, recruitment, customer feedback, checkout, media delivery, supplier engagement and prospective production systems. They do not establish which company is truly the most AI-native. Public communications reveal selected practices under different disclosure incentives. The purpose of this edition is to make those practices and their limitations inspectable.\n\nMayden's editorial interpretation is that an enduring organisational benchmark should follow decisions through the business. A collection of applications is a starting point. The additional evidence concerns responsibility, maintenance, employee capability, exception handling and evaluation. The cases below show why those details are worth asking about and why a named implementation must remain separate from a plan.\n\n## A deliberately broad consumer-business cohort\n\nThe cohort includes Jarir and United Electronics Company, trading as eXtra, in Saudi Arabia; Majid Al Futtaim and LuLu Retail Holdings in the UAE; Al Meera and Baladna in Qatar; Alshaya Group and Mezzan Holding in Kuwait; and BMMI and Trafco Group in Bahrain. There are two entities per market, chosen for purposeful coverage. This is not a representative GCC sample, a market-share ranking or a complete inventory of large consumer businesses.\n\n“Retail & Consumer” is intentionally broader than grocery retail. Baladna manufactures food; Mezzan and Trafco include distribution and manufacturing activities; other entities operate stores or franchise portfolios. Their tasks and economics differ. A consumer-finance subsidiary also should not automatically stand in for an electronics retailer's operating model. Company scope is therefore part of every evidence record.\n\nGeographical labels identify the company's home market in this cohort, rather than the location of every implementation. Majid Al Futtaim's named Carrefour evidence concerns its operating business, not Carrefour SA worldwide. LuLu's historical brand announcements can include countries outside the listed GCC group's perimeter; the retained interpretation is restricted to the relevant GCC operations. One company's international reach does not create observations for every country in which it sells.\n\n## Replenishment makes the management question concrete\n\nLuLu's 2025 annual report describes AI forecasting and replenishment operating in the UAE. It identifies inputs, integration with core systems, planning-team review, authorisation rules and monitoring. Rollout elsewhere is described as ongoing, not completed. This is unusually useful evidence because it explains several parts of the same operating workflow rather than presenting an application name alone. It is still a company account, not an audit of control effectiveness. [LuLu Retail Integrated Annual Report 2025, printed page 104](https://www.luluretail.com/media/pqobufrm/lulu-retail_iar2025_eng.pdf)\n\nThe analytical value lies in the relationship between a prediction and an action. A forecast has no operational effect until someone or something changes an order, allocation or stock position. A retailer therefore needs to know both whether predictions are useful and whether the process acts on them appropriately. A system can be accurate on average and still fail on the categories, stores or exceptional events that matter most.\n\nFor an executive interview, The proposed interview asks for one routine replenishment decision and one exception. The respondent should identify which inputs were available, what the recommendation was, which rule applied, whether a person intervened and what happened next. A redacted decision record would provide stronger evidence of actual responsibility than a general claim that the system is supervised.\n\nThis is also where autonomy should be defined carefully. Human approval and bounded automation can be a deliberate operating design. They should not be treated as a failed attempt at full autonomy. The question is whether authority is appropriate to the task and explicit enough to inspect. A system that makes a suggestion, one that prepares an order, and one that commits an order have different consequences.\n\n## A digital shopfront does not establish a digital organisation\n\nLuLu's October 2026 announcement describes AI shopping robots deployed in UAE stores, with navigation and promotional functions. It supports a specific customer-service implementation. It does not establish how widely shoppers use it, whether it improves conversion, or whether the broader enterprise delegates material decisions to AI. [LuLu robot announcement, 1 October 2026](https://www.luluretail.com/media/news/lulu-brings-ai-powered-robotic-shopping-experience-to-uae-stores/)\n\nAn earlier LuLu announcement describes a WhatsApp assistant implemented with Yellow.ai, connected to customer and order systems and supported by handoff to a live agent. The integration is an organisationally relevant detail: service answers can be linked to a real transaction rather than a static FAQ. The review does not reproduce the announcement's performance claims as independently verified results. [LuLu WhatsApp announcement, 28 February 2023](https://www.luluretail.com/media/news/lulu-hypermarket-launches-new-ai-powered-customer-experiences-on-whatsapp-in-collaboration-with-yellowai)\n\nFor either interface, the executive question is what happens when the system cannot complete the task. An effective escalation can matter more than an impressive demonstration. Mayden's proposed interview should examine failed requests, incorrect product guidance, accessibility, language coverage and the transfer of context to an employee. It should also ask whether the company monitors unresolved demand rather than merely counting interactions.\n\nMajid Al Futtaim's 2021 Carrefour City+ release describes AI-assisted item recognition and automatic charging at a named store in Dubai. This is evidence of a historical service launch with a defined customer journey. It is not evidence that the same configuration remains operational at the cutoff or that a wider network adopted it. [Majid Al Futtaim Carrefour City+ announcement, 6 September 2021](https://www.majidalfuttaim.com/en/media-centre/press-releases/detail/2021/09/carrefour-city-plus)\n\nHistorical evidence deserves a visible date. An index should neither discard every older implementation nor silently describe it as current. It should make continuity a separate question and seek a dated operating update in the next research cycle.\n\n## The back office changes the picture\n\nMajid Al Futtaim's 2024 sustainability report describes an AI pricing engine, AI analysis of customer feedback, AI-supported recruitment and a new AI Academy. The academy has separate learning paths for everyday users, technical specialists and senior leaders. Recruitment uses screening questions and job-fit scores. These passages broaden the observable organisational picture, but they do not establish unbiased hiring, autonomous pricing or company-wide proficiency. The source's other training metrics are not automatically AI Academy results. [Majid Al Futtaim Sustainability Report 2024, pages 70, 76 and 80–81](https://www.majidalfuttaim.com/docs/default-source/reports/maf-sus-pages-full_final.pdf?sfvrsn=7c1de58c_2)\n\nThese applications lead to different tests. A pricing recommendation should be assessed against commercial constraints and customer effects. A recruitment recommendation requires an account of the human decision and the handling of contested outcomes. Customer-feedback analysis needs a way to check whether important minority complaints disappear inside broad categories. Training needs evidence that people can apply the relevant judgement, not merely attend a course.\n\nThe common thread is the design of work. When an AI tool shortens an analysis task, the company still has to decide what happens to the time released and who owns the action that follows. Faster classification of complaints is valuable only if the organisation identifies and fixes the right problems. A benchmark should avoid converting a processing-speed claim into a claim about customer outcomes without the intermediate evidence.\n\nThere is also a knowledge question. A company may successfully classify incoming feedback without maintaining a reliable body of operational knowledge. To distinguish the two, The next company interview should examine how categories change, how corrections are retained and whether later users can inspect the source material behind a conclusion. The current public record is not enough to answer all of those questions.\n\n## Retail media needs the same operating test as every other service\n\nIn May 2025, Majid Al Futtaim described Advertima Audience AI deployed across UAE Carrefour hypermarkets, while supermarket expansion remained prospective. The release explains a sensor-to-message mechanism for retail media. The review recognises the operator's implemented service workflow. It does not treat the announcement as independent proof of advertising effectiveness or privacy safeguards. [Majid Al Futtaim–Advertima announcement, 13 May 2025](https://www.majidalfuttaim.com/en/media-centre/press-releases/detail/2025/05/majid-al-futtaim-s-precision-media-and-advertima-set-new-retail-media-benchmark-with-ai-rollout-in-carrefour-stores)\n\nAlshaya's Media Connect service page advertises AI-assisted media delivery using purchase data and intent signals. That is a specific offering, but the inspected page does not document a live campaign implementation. A separate AI-assisted source challenge excludes this catalogue claim from scoring: it does not establish deployment or a specific internal implementation commitment. The distinction is about the evidence available, not a claim that the service is inactive. [Alshaya Media Connect](https://www.alshaya.com/en/alshaya-media-connect/)\n\nConsistency is essential here. Selling an AI-labelled product is not the same as using AI to operate a service. A retailer's own media-delivery process can qualify if the source actually describes it in operation. A service catalogue alone leaves an implementation question. The same rule should apply irrespective of the prominence of the company or the attractiveness of the announcement.\n\nFor a deeper assessment, the relevant questions concern responsibility for targeting decisions, the inputs used, the interpretation of performance measures and the handling of unsuitable outputs. Those questions are proposed due diligence, not findings that either company has failed a requirement.\n\n## Recovering a source can change the evidence, without changing the standard\n\nAl Meera's older annual-report link did not initially provide a readable document. The research recovered the full report through the company's new annual-report page and its linked document host. The report describes smart carts launched in 2024, with login, scanning and purchase functions, and places them in an AI discussion. It also records the earlier opening of Al Meera Smart. The named vendor's 2022 announcement remains a separate, earlier project commitment. [Al Meera Annual Report 2024, printed pages 14–17](https://sacorporatewebprodqc001.blob.core.windows.net/public/documents/Al%20Meera%20AR%202024%2013%20APR%202025%20ENG%20FINAL%20(1).pdf); [Zippin project announcement](https://www.getzippin.com/blog/al-meera-partners-with-zippin-to-open-first-checkout-free-stores-in-qatar)\n\nThis case illustrates why access status belongs in a serious index. An indexed extract may identify a promising lead, but it is not equivalent to reading the source passage. Conversely, a broken old link should not permanently erase evidence that can be recovered from an official archive.\n\nThe company case still has boundaries. A launched shopping interface does not supply a rollout denominator, a controlled comparison or evidence about every part of the business. The report also cannot establish that every component described as smart uses AI. The retained claim is limited to the specific passage that links the cart workflow to AI.\n\n## Consumer production adds a different time horizon\n\nBaladna's 2025 report describes future AI-supported utility metering and AI-capable feed-management software, including intended integration with existing systems. Those are specific operational plans. They remain plans in this review because the passages do not establish completed AI deployment. The company's existing production, water or energy achievements must not be attributed to a future system. [Baladna Annual Report 2025, physical pages 39 and 41, printed pages 76–81](https://baladna.com/storage/uploads/iKUuHjMbSSYZfRre3VkWieoWq.pdf)\n\nA consumer-products company can have a sophisticated physical operation before an AI project enters service. The Index should not confuse general technological capability with evidence of AI-specific practice. It should also preserve the ability to record a later transition once a dated source establishes installation, routine use and a responsible operating team.\n\nMezzan provides a more bounded implemented example. Its 2025 sustainability report describes GenAI used to develop an ESG questionnaire with deployment targeting its top twenty high-spend suppliers. This supports GenAI-assisted drafting and a specified intended supplier cohort; it does not confirm twenty completed deployments or responses. It does not establish that AI autonomously scored suppliers or verified compliance. [Mezzan Sustainability Report 2025, physical page 53](https://www.mezzan.com/wp-content/uploads/2026/08/Mezzan-Sustainability-Report-2025-English-.pdf)\n\nThe distinction is valuable. Generating a questionnaire, interpreting responses and making a supplier decision are three different activities. A respondent should identify exactly where the AI contribution stops. The follow-up should establish what human review checked relevance, completeness and unintended ambiguity before the questionnaire was used.\n\n## Silence and irrelevant mentions need different labels\n\nJarir's examined annual-report language about AI-enabled products concerns merchandise, not an internal operating practice. eXtra's issuer notice provides a different kind of evidence: supervision of an Executive Management Committee for Data and AI during fiscal 2024. That supports a specific governance activity, without proving an implemented store or fulfilment application. [Jarir Annual Report 2025](https://wps-media.jarir.com/wp-content/uploads/2026/04/JB_AnnualReport_2025_EN.pdf); [eXtra issuer notice, 30 April 2025](https://www.saudiexchange.sa/wps/portal/saudiexchange/newsandreports/issuer-news/issuer-announcements/issuer-announcements-details/?anCat=1&anId=86829&cs=4003&locale=en)\n\nBMMI's latest report downloads presented access difficulties, while its accessible 2024 annual report did not yield a qualifying practice in the searched text. Trafco's examined annual-report material described technology without an explicit qualifying AI practice. Neither result justifies calling the company non-native or digitally immature.\n\nThe research also encountered Arabic news pages labelled as AI-generated summaries. Those labels describe the publisher's process, not the subject company's adoption. Similarly, a supplier portal marked as a testing environment cannot establish production deployment. These exclusions are not minor editorial choices; they prevent a keyword count from becoming a misleading organisational measure.\n\n## From an evidence page to a useful executive conversation\n\nThe proposed company interview should begin with a material workflow rather than an abstract self-rating. A retailer might choose replenishment or customer service; a consumer-products company might choose production planning. The interviewer should then ask for an additional workflow outside the flagship area to test whether practices are repeatable.\n\nFor each workflow, the discussion should cover nine dimensions: the role of AI in core work; role-specific employee capability; partner responsibilities; integration with systems and people; records that allow a decision to be reconstructed; rules for experiments; permitted actions and overrides; maintenance of knowledge; and applied AI governance. Respondents should be able to answer through a guided click-through or voice interview, but the underlying evidence request should be the same.\n\nA useful company report would separate three things: what was disclosed, what was substantiated in the assessment, and what remains unresolved. It would explain the next evidence needed for each conclusion and provide a small number of prioritised operating questions. It should not penalise a company for declining to disclose commercially sensitive details publicly; confidential participation and public evidence are different measurement channels.\n\nThis is a proposed assessment design, not a validated scale. Before claiming to benchmark organisations scientifically, Mayden needs to test question interpretation, scoring reliability, missing-data patterns and the relationship between measured dimensions and external outcomes. The public corpus provides useful material for that work, but it cannot substitute for it.\n\n## Reading the retail and consumer evidence exhibits\n\nThe heatmap shows nine dimensions. U means unknown; level 1 denotes a specific commitment or pilot, level 2 an implemented practice, and level 3 adds concrete operating detail. Each cell takes the highest supported, adjudicated level.\n\nPoints equal the supported total divided by 27 and multiplied by 100, rounded. The fixed denominator allows three points per dimension. Equal underlying totals share a position; records without qualifying evidence remain unranked. Unknowns add no documented points without proving absent capability. Positions describe this selected public corpus, not maturity, commercial success or national adoption.\n\nAnchor counts distinguish specific commitments or pilots, implemented practices, and implemented practices with operating detail. They do not count independent applications or companies: one replenishment workflow can support several dimensions. These are research-record counts, not estimates of operational AI coverage.\n\nSource dependence shows the largest points reduction when one source URL is removed. A sharp decrease identifies reliance on one disclosure, not loss of capability. A smaller decrease does not prove independent verification: URLs can repeat announcements. Historical cases, subsidiary boundaries and supplier accounts retain their claim-specific qualifications.\n\nThe enduring value of the Index will depend on disciplined updates. New evidence may strengthen a claim, reveal that a project stopped, or clarify that a widely promoted deployment had a narrower scope. Preserving the previous edition and recording the reason for each change will make those developments legible. That is how this research can become a useful continuing record of enterprise change while keeping its scientific claims proportional to the evidence.\n\n**Evidence and reproducibility.** The companion retail-evidence.json contains entity scopes, source relationships, passage locators, deployment stages and proposed evidence levels. Separate AI-assisted source challenge is retained separately and governs the released calculations. This edition uses public company and partner disclosures; it includes no private system access, respondent survey, statistical meta-analysis or validation of actual organisational maturity.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/retail-public-evidence.pdf", "cover": "/research/index/2026-10/retail-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 95115, "pdfSha256": "21f15e18003effb1aea0122eb0781bb09ddfffb5728a9e83e775bc9671edd170", "pdfPages": 18 }, { "slug": "healthcare-public-evidence", "number": 8, "title": "AI in Gulf healthcare: evidence at the point of care", "shortTitle": "Evidence at the point of care", "dek": "Clinical and administrative workflows, responsibility and the boundaries of deployment.", "sector": "Healthcare", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Follow documentation into clinical review. Dallah reports ambient AI supporting clinician documentation. Ask who verifies each note, how errors are corrected and what share of eligible encounters uses the workflow.", "Separate throughput, validation and patient benefit. M42 reports daily screening activity and a large radiograph-validation study. Evaluate each evidence layer separately; neither alone establishes improved patient outcomes.", "Test the clinician’s final authority. HMG describes imaging support and a responsible-AI framework retaining clinical authority. Request an example of a disputed output, the intervention made and the subsequent review." ], "executiveSources": [ { "claimIds": [ "HLT-006" ], "sources": [ { "claimId": "HLT-006", "url": "https://www.dallahhealth.com/Library/Assets/Downloads/Dallah_Health_English_2025.pdf", "sourceTitle": "Annual Report 2025" } ] }, { "claimIds": [ "HLT-012", "HLT-013" ], "sources": [ { "claimId": "HLT-012", "url": "https://m42.ae/media-resources/news/m42-and-ge-healthcare-join-forces-to-advance-ai-enabled-patient-care/", "sourceTitle": "M42 and GE HealthCare join forces to advance AI-Enabled Patient Care" }, { "claimId": "HLT-013", "url": "https://m42.ae/media-resources/news/m42-announces-breakthrough-results-for-its-ai-powered-tuberculosis-screening/", "sourceTitle": "M42 announces breakthrough results for its AI-powered tuberculosis screening" } ] }, { "claimIds": [ "HLT-001", "HLT-002" ], "sources": [ { "claimId": "HLT-001", "url": "https://hmg.com/ir/en/Lists/Reports_Publications/Attachments/111/HMG_Annual_Report_2025__English.pdf", "sourceTitle": "Annual Report 2025" }, { "claimId": "HLT-002", "url": "https://hmg.com/ir/en/Lists/Reports_Publications/Attachments/111/HMG_Annual_Report_2025__English.pdf", "sourceTitle": "Annual Report 2025" } ] } ], "evidenceAgeContext": "Dated sources run from 2024-12-30 to 2025-08-06. 22 of 27 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "Three evidence layers in healthcare AI", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 32, "label": "recorded claims" }, { "value": 27, "label": "eligible claims" }, { "value": 3, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "Three evidence layers in healthcare AI", "subtitle": "Activity, model evaluation and clinical authority answer different questions. Cases are not pooled into a clinical-quality score.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "Operating activity", "unit": "Screening examinations", "finding": "M42 reports AIRIS-TB conducting 2,000 AI chest-X-ray examinations per day.", "boundary": "Issuer-reported throughput; staffing/comparator details incomplete.", "implication": "How many eligible examinations reach a reviewed clinical decision?", "claimIds": [ "HLT-012" ], "sources": [ { "claimId": "HLT-012", "url": "https://m42.ae/media-resources/news/m42-and-ge-healthcare-join-forces-to-advance-ai-enabled-patient-care/", "sourceTitle": "M42 and GE HealthCare join forces to advance AI-Enabled Patient Care" } ] }, { "label": "Retrospective evaluation", "unit": "Radiographs assessed in a study", "finding": "The original AIRIS-TB study evaluates 1.04 million chest X-rays from January 2022 to January 2024 in an observational retrospective analysis.", "boundary": "M42/CHSC-affiliated research; model evaluation is not prospective patient benefit.", "implication": "Does performance hold in the intended workflow, population and prevalence?", "claimIds": [ "HLT-013" ], "sources": [ { "claimId": "HLT-013", "url": "https://m42.ae/media-resources/news/m42-announces-breakthrough-results-for-its-ai-powered-tuberculosis-screening/", "sourceTitle": "M42 announces breakthrough results for its AI-powered tuberculosis screening" }, { "claimId": "L34", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12241593/", "sourceTitle": "Population-scale cross-sectional observational study for AI-powered TB screening on one million CXRs" } ] }, { "label": "Clinical authority", "unit": "A clinician’s accepted or rejected output", "finding": "HMG says its Responsible AI Framework retains clinician final authority and includes fairness monitoring.", "boundary": "A different provider and evidence layer; no assurance result supplied.", "implication": "Can an actual disagreement show how review, intervention and follow-up operate?", "claimIds": [ "HLT-002" ], "sources": [ { "claimId": "HLT-002", "url": "https://hmg.com/ir/en/Lists/Reports_Publications/Attachments/111/HMG_Annual_Report_2025__English.pdf", "sourceTitle": "Annual Report 2025" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "health-purehealth", "label": "PureHealth", "values": [ 3, 2, null, 3, null, 1, null, 2, 1 ], "score": 44, "rank": 1 }, { "id": "health-dallah", "label": "Dallah Healthcare", "values": [ 3, 1, 2, 3, null, null, null, null, 2 ], "score": 41, "rank": 2 }, { "id": "health-hmg", "label": "Dr. Sulaiman Al Habib Medical Services Group", "values": [ 3, null, 1, 2, null, null, null, null, 3 ], "score": 33, "rank": 3 }, { "id": "health-hmc", "label": "Hamad Medical Corporation", "values": [ 2, 2, null, null, null, 1, null, null, 3 ], "score": 30, "rank": 4 }, { "id": "health-m42", "label": "M42", "values": [ 3, null, 1, null, null, 3, null, null, null ], "score": 26, "rank": 5 }, { "id": "health-bsh", "label": "Bahrain Specialist Hospital", "values": [ null, null, null, 3, null, null, null, null, null ], "score": 11, "rank": 6 }, { "id": "health-phcc", "label": "Primary Health Care Corporation", "values": [ 2, null, null, null, null, null, null, null, null ], "score": 7, "rank": 7 }, { "id": "health-kuwait-hospital", "label": "Kuwait Hospital", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "health-rbh", "label": "Royal Bahrain Hospital", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "health-taiba", "label": "Taiba Hospital", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "PureHealth", "value": 44, "secondary": 6, "rank": 1 }, { "label": "Dallah Healthcare", "value": 41, "secondary": 5, "rank": 2 }, { "label": "Dr. Sulaiman Al Habib Medical Services Group", "value": 33, "secondary": 4, "rank": 3 }, { "label": "Hamad Medical Corporation", "value": 30, "secondary": 4, "rank": 4 }, { "label": "M42", "value": 26, "secondary": 3, "rank": 5 }, { "label": "Bahrain Specialist Hospital", "value": 11, "secondary": 1, "rank": 6 }, { "label": "Primary Health Care Corporation", "value": 7, "secondary": 1, "rank": 7 }, { "label": "Kuwait Hospital", "value": null, "secondary": 0, "rank": null }, { "label": "Royal Bahrain Hospital", "value": null, "secondary": 0, "rank": null }, { "label": "Taiba Hospital", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 27; one practice may support multiple distinct themes", "maximum": 11, "rows": [ { "label": "Specific commitment / pilot", "value": 6 }, { "label": "Implemented practice", "value": 11 }, { "label": "Implemented + operating detail", "value": 10 } ], "note": "Source: claim ledger. 5 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "PureHealth", "value": 44.4, "after": 0.0 }, { "label": "Dallah Healthcare", "value": 40.7, "after": 7.41 }, { "label": "Dr. Sulaiman Al Habib Medical Services Group", "value": 33.3, "after": 7.41 }, { "label": "Hamad Medical Corporation", "value": 29.6, "after": 0.0 }, { "label": "M42", "value": 25.9, "after": 11.11 }, { "label": "Bahrain Specialist Hospital", "value": 11.1, "after": 0.0 }, { "label": "Primary Health Care Corporation", "value": 7.4, "after": 0.0 }, { "label": "Kuwait Hospital", "value": null, "after": null }, { "label": "Royal Bahrain Hospital", "value": null, "after": null }, { "label": "Taiba Hospital", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "health-purehealth", "title": "PureHealth", "scope": "PureHealth Holding PJSC consolidated group; includes Daman insurance and overseas HHG, not a UAE hospital-only measure.", "claims": [ { "id": "HLT-016", "claim": "PureHealth’s AI Factory became operational inQ3 2025 and supports document, radiology, pharmacy and claims workflows.", "url": "https://purehealth.ae/wp-content/uploads/2026/03/PureHealth_Integrated-Annual-Report_2025_EN.pdf", "locator": "PDF40–41/printed76–79, AI Factory", "level": 3, "eligible": true, "limitation": "Claim scope: PureHealth Holding PJSC consolidated group; includes Daman insurance and overseas HHG, not a UAE hospital-only measure.. Group architecture assertion does not imply uniform adoption across all acquired entities. Application: internal-operating-practice.", "rationale": "Operational timing, initial platforms and cross-function deployment are specified. Uniform group-wide adoption is not inferred." }, { "id": "HLT-017", "claim": "Daman deployed in-house AI classification, validation and extraction for documents.", "url": "https://purehealth.ae/wp-content/uploads/2026/03/PureHealth_Integrated-Annual-Report_2025_EN.pdf", "locator": "PDF39/printed74–75, Digital Transformation", "level": 3, "eligible": true, "limitation": "Claim scope: PureHealth subsidiary Daman health-insurance operations, UAE.. Insurance subsidiary scope; annual60million figure is issuer-reported and not an external performance audit. Application: internal-operating-practice.", "rationale": "Daman document classification, validation and extraction are deployed with reported annual processing scope; insurance subsidiary only." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Identify the exact clinical or administrative task." }, { "companyId": "health-dallah", "title": "Dallah Healthcare", "scope": "Listed Dallah Healthcare Company and hospital operations in Saudi Arabia.", "claims": [ { "id": "HLT-006", "claim": "Dallah reports Sahl ambient AI documenting patient encounters for more than175 clinicians.", "url": "https://www.dallahhealth.com/Library/Assets/Downloads/Dallah_Health_English_2025.pdf", "locator": "PDF24/printed46–47, Investing in responsible AI", "level": 3, "eligible": true, "limitation": "Claim scope: Listed Dallah Healthcare Company and hospital operations in Saudi Arabia.. No denominator of eligible clinicians; provider-reported scale, not independently verified. Application: internal-operating-practice.", "rationale": "Ambient documentation is explicitly in use by a stated clinician population. Reported time savings are not independently verified." }, { "id": "HLT-007", "claim": "Dallah deployed Thakaa Med machine learning to detect oral pathologies in radiographs.", "url": "https://www.dallahhealth.com/Library/Assets/Downloads/Dallah_Health_English_2025.pdf", "locator": "PDF24/printed46–47", "level": 3, "eligible": true, "limitation": "Claim scope: Listed Dallah Healthcare Company and hospital operations in Saudi Arabia.. Sensitivity, validation dataset, override rates and deployment-site count not reported here. Application: internal-operating-practice.", "rationale": "Deployment and machine-learning analysis of dental radiographs are explicit. No clinical-outcome inference." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Identify the exact clinical or administrative task." }, { "companyId": "health-kuwait-hospital", "title": "Kuwait Hospital", "scope": "Private Kuwait Hospital, Sabah Al-Salem, kuwaithospital.com.kw; excludes UAE hospitals with the same name.", "claims": [], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Identify the exact clinical or administrative task." } ], "agenda": [ { "question": "Identify the exact clinical or administrative task.", "evidence": "Care setting, patient/workflow scope, decision owner and deployment status." }, { "question": "Separate assistance from clinical authority.", "evidence": "Clinician supervision, escalation, validation and permitted-use boundaries." }, { "question": "Ask for prospective operating evidence.", "evidence": "Patient safety, subgroup performance, drift and a comparator appropriate to the care pathway." } ], "sources": [ { "title": "Integrated Annual Report 2025", "publisher": "PureHealth", "date": "Undated", "url": "https://purehealth.ae/wp-content/uploads/2026/03/PureHealth_Integrated-Annual-Report_2025_EN.pdf", "locator": "PDF40–41/printed76–79, AI Factory", "originalUrl": "https://purehealth.ae/wp-content/uploads/2026/03/PureHealth_Integrated-Annual-Report_2025_EN.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S01" }, { "title": "Annual Report 2025", "publisher": "Dallah Healthcare", "date": "Undated", "url": "https://www.dallahhealth.com/Library/Assets/Downloads/Dallah_Health_English_2025.pdf", "locator": "PDF24/printed46–47, Investing in responsible AI", "originalUrl": "https://www.dallahhealth.com/Library/Assets/Downloads/Dallah_Health_English_2025.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S02" }, { "title": "Artificial Intelligence in Dentistry: Enhancing Patient Care and Service Quality", "publisher": "Dallah Hospital", "date": "2024-12-30", "url": "https://www.dallah-hospital.com/english/media-center/news--events/artificial-intelligence-in-dentistry-enhancing-patient-care-and-service-quality", "locator": "Successful implementation opening paragraph", "originalUrl": "https://www.dallah-hospital.com/english/media-center/news--events/artificial-intelligence-in-dentistry-enhancing-patient-care-and-service-quality", "access": "full-source-read", "relationship": "official-implementation-account", "id": "S03" }, { "title": "Annual Report 2025", "publisher": "HMG", "date": "Undated", "url": "https://hmg.com/ir/en/Lists/Reports_Publications/Attachments/111/HMG_Annual_Report_2025__English.pdf", "locator": "PDF51/printed99–100, Radiology Services", "originalUrl": "https://hmg.com/ir/en/Lists/Reports_Publications/Attachments/111/HMG_Annual_Report_2025__English.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S04" }, { "title": "Sustainability2024: Maintain sound business practices", "publisher": "HMG", "date": "Undated", "url": "https://hmg.com/ir/ESGReports/2024/maintain-sound-business-practices.html", "locator": "Innovation and technology, AI applications paragraph", "originalUrl": "https://hmg.com/ir/ESGReports/2024/maintain-sound-business-practices.html", "access": "full-source-read", "relationship": "sustainability-report", "id": "S05" }, { "title": "Precision Medicine Annual Report 2021–2022", "publisher": "Hamad Medical Corporation", "date": "Undated", "url": "https://www.hamad.qa/EN/Hospitals-and-services/CPMGC/Annual-Report/Documents/Precision%20Medicine%20Annual%20Report%202021-2022.pdf", "locator": "PDF39/printed37, CIPAC", "originalUrl": "https://www.hamad.qa/EN/Hospitals-and-services/CPMGC/Annual-Report/Documents/Precision%20Medicine%20Annual%20Report%202021-2022.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S06" }, { "title": "M42 and GE HealthCare join forces to advance AI-Enabled Patient Care", "publisher": "M42", "date": "2025-07-08", "url": "https://m42.ae/media-resources/news/m42-and-ge-healthcare-join-forces-to-advance-ai-enabled-patient-care/", "locator": "Final AIRIS-TB paragraph, line69", "originalUrl": "https://m42.ae/media-resources/news/m42-and-ge-healthcare-join-forces-to-advance-ai-enabled-patient-care/", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S07" }, { "title": "M42 announces breakthrough results for its AI-powered tuberculosis screening", "publisher": "M42", "date": "2025-08-06", "url": "https://m42.ae/media-resources/news/m42-announces-breakthrough-results-for-its-ai-powered-tuberculosis-screening/", "locator": "Study design, results and demographic consistency paragraphs", "originalUrl": "https://m42.ae/media-resources/news/m42-announces-breakthrough-results-for-its-ai-powered-tuberculosis-screening/", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S08" }, { "title": "Voiceoc patient interaction: Bahrain Specialist Hospital testimonial", "publisher": "Voiceoc", "date": "Undated", "url": "https://customerexperience.voiceoc.com/", "locator": "Nadim Navde testimonial, lines222–232", "originalUrl": "https://customerexperience.voiceoc.com/", "access": "full-source-read | Review: full-source-read", "relationship": "supplier-case-study", "id": "S09" }, { "title": "We are working on introducing mandatory annual oral screening for students", "publisher": "PHCC", "date": "2025-03-06", "url": "https://www.phcc.gov.qa/media/news/2025/03/10/we-are-working-on-introducing-mandatory-annual-oral-screening-for-students---dr.-hamad-al-mudhahka", "locator": "Breast and bowel screening paragraph", "originalUrl": "https://www.phcc.gov.qa/media/news/2025/03/10/we-are-working-on-introducing-mandatory-annual-oral-screening-for-students---dr.-hamad-al-mudhahka", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S10" }, { "title": "Annual Report 2024", "publisher": "PHCC", "date": "Undated", "url": "https://www.phcc.gov.qa/docs/default-source/default-document-library/phcc%27s-annual-report-2024-eng-final.pdf", "locator": "PDF23, conference paragraph", "originalUrl": "https://www.phcc.gov.qa/docs/default-source/default-document-library/phcc%27s-annual-report-2024-eng-final.pdf", "access": "full-source-read", "relationship": "annual-report", "id": "S11" }, { "title": "MRI Scan page: KIMSHEALTH FAQs", "publisher": "Royal Bahrain Hospital / KIMSHEALTH", "date": "Undated", "url": "https://www.royalbahrainhospital.com/procedures/magnetic-resonance-imaging-mri", "locator": "FAQ: What advanced technologies are available at KIMSHEALTH?", "originalUrl": "https://www.royalbahrainhospital.com/procedures/magnetic-resonance-imaging-mri", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S12" }, { "title": "Taiba Hospital proudly recognizes Dr Asmaa Al Kandari’s achievement", "publisher": "Taiba Hospital", "date": "2025-03-05", "url": "https://www.taibahospital.com/taiba-hospital-proudly-recognizes-dr-asmaa-al-kandaris-achievement-in-a-global-gastrointestinal-endoscopy-conference/", "locator": "Indexed conference passage; full source unavailable", "originalUrl": "https://www.taibahospital.com/taiba-hospital-proudly-recognizes-dr-asmaa-al-kandaris-achievement-in-a-global-gastrointestinal-endoscopy-conference/", "access": "snippet-only", "relationship": "official-implementation-account", "id": "S13" }, { "title": "Population-scale cross-sectional observational study for AI-powered TB screening on one million CXRs", "publisher": "npj Digital Medicine; full text in PMC", "date": "2025-07-09", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12241593/", "locator": "Study methods and results; see institutional affiliations and limitations", "relationship": "original-observational-model-evaluation", "id": "S14" }, { "title": "Kuwait Hospital", "publisher": "Primary source as linked", "date": "See original", "url": "https://kuwaithospital.com.kw/", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S15" } ], "body": "# Healthcare: the AI-native test is a change in care delivery\n\n*Mayden AI Native Index — exploratory public-evidence research. Cutoff: 10 October 2026. This is a purposeful sample of ten healthcare organisations, not a clinical-quality ranking or an estimate of national adoption.*\n\nThe meaningful question for a healthcare executive is whether AI changes how work gets done while preserving accountable clinical judgement. A hospital can acquire advanced equipment, sponsor conferences and announce technology partnerships without changing a routine workflow. Another can implement a narrow, well-supervised intervention that makes a real difference to documentation, screening or patient access. Those situations should not receive the same interpretation.\n\nOur reviewed material makes the distinction concrete. It contains operating accounts of AI-assisted imaging, clinical documentation, insurance administration and appointment support. It also contains pilots, future organisational plans, broad supplier promises and statements whose entity scope cannot be established. Mayden retains these differences in the evidence register. An attractive narrative is useful only if a reader can inspect the passage, understand whose operation it describes and see why the claim qualifies.\n\nThis edition examines Dr. Sulaiman Al Habib Medical Services Group and Dallah Healthcare in Saudi Arabia; M42 and PureHealth in the UAE; Hamad Medical Corporation and Primary Health Care Corporation in Qatar; Taiba Hospital and Kuwait Hospital in Kuwait; and Bahrain Specialist Hospital and Royal Bahrain Hospital in Bahrain. Selection was purposeful, with two organisations headquartered in each market. Qatar's two organisations are public providers and are labelled accordingly. PureHealth and M42 are integrated international groups. The sample therefore supports comparison of public evidence, not a contest between equivalent hospitals.\n\n## What the strongest disclosures explain\n\nHMG's 2025 report describes cardiac-imaging AI and an AI framework that preserves clinicians' final authority and includes fairness monitoring. These are separate questions: where AI assists care, and how its use is controlled. The report also identifies clinical documentation tools. A partnership described as established to support development and research remains a commitment where completed joint work is not demonstrated. Its evidence is narrower than its ambition. [HMG Annual Report 2025, PDF pages 39 and 51](https://hmg.com/ir/en/Lists/Reports_Publications/Attachments/111/HMG_Annual_Report_2025__English.pdf)\n\nDallah identifies Sahl ambient documentation, reporting use by more than 175 clinicians, and deployed Thakaa Med radiograph analysis. Its responsible-AI framework is described at a higher level than HMG's specific authority and monitoring account. A training paragraph sits in a forward-looking section and is treated as a planned commitment, not a completed workforce programme. This distinction survived a separate AI-assisted challenge of the initial research coding. [Dallah Annual Report 2025, PDF pages 24–26](https://www.dallahhealth.com/Library/Assets/Downloads/Dallah_Health_English_2025.pdf)\n\nM42 reports AIRIS-TB operating at 2,000 chest examinations daily. That issuer-reported throughput is separate from model evaluation. The original AIRIS-TB paper evaluates historical screening images; its estimated automation potential is not a measured reduction in routine clinical workload. Prospective workflow validation remains necessary. Executives should request both technical evaluation and evidence of what changed in care delivery. The GE HealthCare collaboration remains a forward-looking commitment. [M42 operating account](https://m42.ae/media-resources/news/m42-and-ge-healthcare-join-forces-to-advance-ai-enabled-patient-care/), [original study, accessible full text](https://pmc.ncbi.nlm.nih.gov/articles/PMC12241593/).\n\nPureHealth's report connects a shared AI infrastructure to particular operations. Daman's document processing is insurance administration; SEHA Clinics' retinal screening is clinical delivery; HHG's nurse scheduling and executive insights concern overseas operations. These cases belong in an explicitly consolidated group profile. They must not be silently converted into evidence that every UAE hospital uses the same systems. The physician-documentation pilot and proposed AI Office also retain their own statuses. [PureHealth Integrated Annual Report 2025, PDF pages 29, 31, 36 and 39–41](https://purehealth.ae/wp-content/uploads/2026/03/PureHealth_Integrated-Annual-Report_2025_EN.pdf)\n\nThese examples suggest a practical reporting standard: name the workflow, identify the operating entity, state whether use is live, describe the mechanism and explain the decision boundary. A list of technologies is less informative than a clear account of who does what differently. Executives can use that standard immediately, even before collecting additional quantitative measures.\n\n## Historical evidence and current claims need different treatment\n\nHMC provides unusually specific historical material. Its 2021–2022 precision-medicine report describes an imaging AI committee, internal educational sessions and integrated imaging trials. The committee has review and monitoring responsibilities. The Siemens and Koios applications are explicitly trials, even though integration into imaging systems is technically detailed. We retain the pilot cap and label the committee evidence historical. The document cannot confirm that the same arrangements remain active in October 2026. [HMC Precision Medicine Annual Report, PDF page 39](https://www.hamad.qa/EN/Hospitals-and-services/CPMGC/Annual-Report/Documents/Precision%20Medicine%20Annual%20Report%202021-2022.pdf)\n\nPHCC's March 2025 account says AI was integrated into its 2024 screening programme to assist image interpretation. That supports an implemented practice. It does not identify enough about the model, coverage or evaluation to support a more detailed operating anchor. Its annual report's conference discussion is not automatically an internal workforce-development programme. [PHCC screening account](https://www.phcc.gov.qa/media/news/2025/03/10/we-are-working-on-introducing-mandatory-annual-oral-screening-for-students---dr.-hamad-al-mudhahka), [PHCC Annual Report 2024](https://www.phcc.gov.qa/docs/default-source/default-document-library/phcc%27s-annual-report-2024-eng-final.pdf)\n\nAge matters because technology and organisations change. A programme may be replaced, expanded, discontinued or absorbed into another service. Historical evidence can still explain an organisation's development and inform follow-up questions. It should not acquire a false current date because a researcher accessed the webpage recently. Publication date, reporting period, observation date and research access date are distinct fields.\n\nFor a future recurring edition, The next research cycle should ask organisations to reconfirm older operating claims with dated implementation records. The correct question is not simply whether the old claim remains on the website. It is whether the operation, accountable owner, user population and control arrangement remain materially the same. Changes should create a documented update rather than quietly rewriting the historical record.\n\n## What the gaps actually mean\n\nThe Kuwait and Bahrain searches illustrate why an unknown result needs a visible explanation. Taiba's indexed news points to a doctor's AI-assisted endoscopy demonstration at an overseas conference. The full page could not be recovered through the reviewed access route. Neither an inaccessible passage nor an overseas demonstration establishes routine use at the Kuwait hospital. Kuwait Hospital's official identity and service inventory were checked, but the bounded search recovered no qualifying AI operating claim. This is a research limit, not a judgement about clinical capability. [Taiba source lead, excluded](https://www.taibahospital.com/taiba-hospital-proudly-recognizes-dr-asmaa-al-kandaris-achievement-in-a-global-gastrointestinal-endoscopy-conference/), [Kuwait Hospital](https://kuwaithospital.com.kw/)\n\nBahrain Specialist Hospital has a named IT director's testimonial on Voiceoc's site describing conversational AI in appointment and call support, including a reported share of traffic. This is direct implementation-participant evidence, but supplier-hosted and undated. General performance claims elsewhere on the supplier page are not hospital results. Royal Bahrain Hospital's MRI page includes a broader KIMSHEALTH FAQ mentioning AI diagnostics; that template does not identify the Bahrain hospital's own implementation. It remains excluded from local scoring. [BSH testimonial](https://customerexperience.voiceoc.com/), [Royal Bahrain Hospital FAQ](https://www.royalbahrainhospital.com/procedures/magnetic-resonance-imaging-mri)\n\nA public-evidence index will partly reflect disclosure capacity. Large listed groups produce extensive reports and have communications resources. Smaller private providers may operate effective systems without describing them publicly. Replacing quiet organisations with easier-to-score ones would make the cohort look more impressive while weakening its integrity. The better response is to preserve the selected entities, publish the search boundary and invite evidence through a standard verification process.\n\n## Nine operating questions for healthcare leaders\n\n**AI in core work.** Which recurring clinical or administrative workflow uses AI today? Identify the service, facility, operating owner and eligible population. Distinguish assistance from a system that completes a task. Describe the ordinary workflow after implementation, including what happens when the tool is unavailable. This makes the answer testable without requiring proprietary model code.\n\n**Workforce.** Which roles have changed, and how is competence checked? Training attendance alone does not establish effective use. Ask whether clinicians and administrative staff understand limitations, escalation and correction. Separate staff training from AI used to allocate staff: both concern work, but they are different organisational practices and need different evidence.\n\n**Ecosystem.** What does the organisation do with partners beyond signing an agreement? Useful evidence identifies implementation responsibilities, integration work, support arrangements and who controls changes. A partnership announcement can establish a specific commitment. Continued delivery requires a more substantive account, especially when a vendor's software is embedded in a consequential care workflow.\n\n**Interfaces.** Where does a patient or employee encounter AI, and how is the interaction connected to the underlying service? An appointment assistant should be assessed through completed service transactions, unresolved requests and human escalation. A documentation assistant should be assessed through review and correction in the record workflow. Conversation volume alone does not explain whether work was completed reliably.\n\n**Lineage.** Can an organisation reconstruct which information, model version and instruction produced an important output? Generic electronic records and cybersecurity controls do not answer this question. The interview should seek a concrete example of tracing an AI-assisted decision or document back through its inputs and changes, with privacy-preserving evidence suitable for review by a named human assessor.\n\n**Experimentation.** How are new tools evaluated before routine adoption and after material change? A pilot should have a stated question, target population, comparison approach, acceptance criteria and stopping rule. Detailed integration is not a substitute for evaluation. Equally, a completed evaluation should not be confused with proof of general clinical benefit across settings.\n\n**Autonomy.** What can the system do without a person's approval? This should be answered task by task, with explicit limits, override mechanisms and responsibility. Automatically drafting text differs from submitting a claim, changing a schedule or influencing a clinical action. The label “agentic” does not supply these missing details and should not determine an anchor.\n\n**Knowledge.** How does the organisation turn AI-assisted information into reusable operational learning? Examples could include evaluated retrieval, reviewed summaries or structured feedback into a clinical protocol. A dashboard described as AI-enabled establishes less than a documented learning loop. Ask who checks the content and how errors alter future practice.\n\n**Governance.** Who may approve, pause or retire a system, and what evidence informs that decision? Policies become more informative when they specify authority and operating routines. Ask for an example of a model change, an incident review or a rejected use case. This reveals whether oversight is part of ordinary management rather than a document created for external communication.\n\nThese questions are a proposed diagnostic instrument, not validated survey scales. The company study must test clarity with clinical, operational and technical respondents before treating answers as comparable measurement. Some concepts require multiple informants. A chief executive may know strategic ownership while a clinical lead knows exception handling and a technical owner knows model versioning. Disagreement can be useful evidence of an incomplete operating model.\n\n## Measurement that executives can trust\n\nThe public-evidence layer uses discrete anchors. A specific named commitment or pilot can establish level 1. An explicitly implemented internal practice can establish level 2. Implementation with a concrete mechanism, coverage, evaluation or authority can establish level 3, subject to a separate AI-assisted source challenge. Unknown means that the bounded search found no qualifying evidence. It does not mean that the organisation lacks the practice.\n\nThe aggregate must remain a summary of documentation across the nine dimensions. It is not the probability that an organisation is AI-native, a clinical-safety grade or a measure of patient benefit. Repeating the same case in several press releases must not increase its score. The maximum supported anchor within a dimension is retained, and the evidence reader should show the contributing claim rather than a mysterious total.\n\nOutcome claims need a separate ledger. A reported reduction in documentation time, annual document throughput, screening discrimination and appointment traffic are not interchangeable benefits. Each has a different denominator and study design. Before comparing results, record the baseline, time period, sample, exclusions and relevant operating conditions. Where those fields are missing, label the result as an attributed organisational claim. Do not average unlike percentages into a sector impact statistic.\n\nA stronger proprietary assessment would add submitted evidence to the public record while preserving provenance. Mayden could request a workflow map, redacted approval record, role-specific adoption figures and a description of evaluation and monitoring. The organisation should approve only the factual accuracy and disclosure permissions of its submitted material, not its eventual score. A separate reviewer should decide whether the evidence meets the published anchor.\n\nThe first pilot should test respondent burden as well as measurement quality. Start with an executive interview, route detailed questions to the appropriate owners and ask for a small number of useful artefacts. A long questionnaire filled with hypothetical best practice can reward confident language rather than real capability. A shorter interview that follows one implemented workflow end to end can uncover responsibility, exceptions and learning much more clearly.\n\n## The reporting opportunity\n\nThe media opportunity is the gap between visible applications and visible operating systems, illustrated through scoped cases and unanswered questions. Two unlike providers cannot establish a country winner. Unknown entries do not mean lagging capability, and international groups do not represent national health systems.\n\nThe heatmap shows documented evidence and unknowns. Points summarise that profile under the declared rules; they do not measure clinical quality or enterprise maturity. The evidence-anchor chart separates commitments and pilots from implementation and operating detail, counting claims rather than independent systems. Source-URL removal shows the effect of omitting the most influential retained URL: a large change signals concentrated documentation, not weak capability. None establishes national adoption or patient benefit.\n\nParticipants should receive an evidence profile, unanswered operating questions and prioritised verification gaps. Repeated editions can track commitments becoming operations, stopped pilots, improved controls and better-evidenced outcomes. Consistent definitions and traceable revisions make that comparison useful.\n\n**Review transparency.** Extraction and source challenge were AI-assisted, without human expert sign-off or independent assurance.", "pdf": "/research/index/2026-10/healthcare-public-evidence.pdf", "cover": "/research/index/2026-10/healthcare-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "pdfPages": 17, "pdfSha256": "6248ec43cb77d932b0662bac05d7d9d774261f622aa59e388b3b89bccbc5c083", "bytes": 89460 }, { "slug": "realestate-public-evidence", "number": 9, "title": "AI in Gulf real estate: beyond the smart-city promise", "shortTitle": "Beyond the smart-city promise", "dek": "Property, construction and asset operations across ten selected entities.", "sector": "Real Estate & Construction", "kind": "Industry evidence review", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "Locate the capability within the group. Aldar describes district and parking applications; ST Engineering describes Lusail city operations. Establish the operating entity and its responsibilities before generalising to a developer’s wider portfolio.", "Trace better information into a management decision. RSG describes CoralNet analysis and comparison with skilled analysts. Ask how survey findings alter conservation action; faster monitoring does not itself establish ecological improvement.", "Refresh historical cases before claiming current coverage. Winnow’s Emaar account describes a 2019 hospitality implementation. Verify continuing use and the properties covered before treating it as evidence of today’s wider operating model." ], "executiveSources": [ { "claimIds": [ "REA-008", "REA-020", "REA-021" ], "sources": [ { "claimId": "REA-008", "url": "https://app-aprp01saasa21mtp001.cms.optimizely.com/globalassets/aldar-innovation-report-2025---eng---final-09-april.pdf", "sourceTitle": "Aldar Innovation Report2025" }, { "claimId": "REA-020", "url": "https://www.stengg.com/en/innovation/artificial-intelligence/how-ai-is-powering-the-next-generation-of-smart-cities/", "sourceTitle": "How AI is powering the next generation of smart cities" }, { "claimId": "REA-021", "url": "https://www.stengg.com/en/innovation/artificial-intelligence/how-ai-is-powering-the-next-generation-of-smart-cities/", "sourceTitle": "How AI is powering the next generation of smart cities" } ] }, { "claimIds": [ "REA-003", "REA-004" ], "sources": [ { "claimId": "REA-003", "url": "https://www.redseaglobal.com/en/media-center/blog/innovation-meets-preservation-technology-s-impact-on-red-sea-biodiversity-conservation/", "sourceTitle": "Innovation meets preservation: technology’s impact on Red Sea biodiversity conservation" }, { "claimId": "REA-004", "url": "https://www.redseaglobal.com/en/media-center/blog/innovation-meets-preservation-technology-s-impact-on-red-sea-biodiversity-conservation/", "sourceTitle": "Innovation meets preservation: technology’s impact on Red Sea biodiversity conservation" } ] }, { "claimIds": [ "REA-005", "REA-006" ], "sources": [ { "claimId": "REA-005", "url": "https://www.winnowsolutions.com/resources/news/emaar-hospitality-group-first-to-bring-cutting-edge-technology-into-uae-kitchen", "sourceTitle": "Emaar Hospitality Group first to bring AI technology into UAE kitchens" }, { "claimId": "REA-006", "url": "https://www.winnowsolutions.com/resources/news/emaar-hospitality-group-first-to-bring-cutting-edge-technology-into-uae-kitchen", "sourceTitle": "Emaar Hospitality Group first to bring AI technology into UAE kitchens" } ] } ], "evidenceAgeContext": "Dated sources run from 2019-05-14 to 2026-03-18. 11 of 20 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E05", "displayNumber": 1, "title": "The company name is not the operating boundary", "kind": "generated" }, { "id": "E01", "displayNumber": 2, "title": "Inspect the operating profile, not only the total", "kind": "generated" }, { "id": "E02", "displayNumber": 3, "title": "Documented evidence points in the selected cohort", "kind": "generated" }, { "id": "E03", "displayNumber": 4, "title": "Separate a commitment from a deployed practice", "kind": "generated" }, { "id": "E04", "displayNumber": 5, "title": "How much of the profile rests on one retained source URL?", "kind": "generated" } ], "exhibitOrder": [ "E05", "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 10, "label": "selected entity records" }, { "value": 27, "label": "recorded claims" }, { "value": 20, "label": "eligible claims" }, { "value": 4, "label": "unranked records" } ], "exhibits": [ { "id": "E05", "type": "case-comparison", "title": "The company name is not the operating boundary", "subtitle": "Three units within developer groups require different attribution and completion evidence.", "comparisonAxis": "Operating unit / source account / unresolved boundary / next verification", "rows": [ { "label": "District and parking", "unit": "Aldar district / Aldar Estates operations", "finding": "Aldar describes district AI with Siemens and parking applications through Aldar Estates/Parkonic.", "boundary": "Specific operating entities; design capacity is not active deployment.", "implication": "Which assets are live, and who owns the consequential decision?", "claimIds": [ "REA-008" ], "sources": [ { "claimId": "REA-008", "url": "https://app-aprp01saasa21mtp001.cms.optimizely.com/globalassets/aldar-innovation-report-2025---eng---final-09-april.pdf", "sourceTitle": "Aldar Innovation Report2025" } ] }, { "label": "Environmental monitoring", "unit": "RSG marine-survey operation", "finding": "RSG describes CoralNet processing survey imagery after training on human-labelled images.", "boundary": "Monitoring capability is not evidence of ecological improvement.", "implication": "Which management action follows a survey result, and what is observed afterwards?", "claimIds": [ "REA-003", "REA-004" ], "sources": [ { "claimId": "REA-003", "url": "https://www.redseaglobal.com/en/media-center/blog/innovation-meets-preservation-technology-s-impact-on-red-sea-biodiversity-conservation/", "sourceTitle": "Innovation meets preservation: technology’s impact on Red Sea biodiversity conservation" }, { "claimId": "REA-004", "url": "https://www.redseaglobal.com/en/media-center/blog/innovation-meets-preservation-technology-s-impact-on-red-sea-biodiversity-conservation/", "sourceTitle": "Innovation meets preservation: technology’s impact on Red Sea biodiversity conservation" } ] }, { "label": "Hospitality kitchen", "unit": "Emaar Hospitality implementation", "finding": "Winnow’s 2019 account describes AI-enabled food-waste recognition after a proof of concept.", "boundary": "Historical supplier account; current use and coverage not confirmed.", "implication": "Which kitchens still use it, and how does the signal change purchasing or production?", "claimIds": [ "REA-005" ], "sources": [ { "claimId": "REA-005", "url": "https://www.winnowsolutions.com/resources/news/emaar-hospitality-group-first-to-bring-cutting-edge-technology-into-uae-kitchen", "sourceTitle": "Emaar Hospitality Group first to bring AI technology into UAE kitchens" } ] } ], "note": "Source: selected claims in PE-2026.10-v4. Company and supplier reports are attributed accounts. The questions are Mayden editorial interpretations, not measured sector results or independent assurance.", "displayNumber": 1 }, { "id": "E01", "type": "heatmap", "title": "Inspect the operating profile, not only the total", "subtitle": "Highest eligible anchor per theme; U = no qualifying evidence located", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "id": "real-aldar", "label": "Aldar Properties", "values": [ 3, 3, 3, 3, null, 1, 2, 3, null ], "score": 67, "rank": 1 }, { "id": "real-qatari-diar", "label": "Qatari Diar", "values": [ 3, null, 1, 3, null, null, null, null, null ], "score": 26, "rank": 2 }, { "id": "real-rsg", "label": "Red Sea Global", "values": [ 3, null, null, null, null, 3, null, null, null ], "score": 22, "rank": 3 }, { "id": "real-emaar", "label": "Emaar Properties", "values": [ 3, null, 2, null, null, null, null, null, null ], "score": 19, "rank": 4 }, { "id": "real-udc", "label": "United Development Company", "values": [ null, null, null, 3, null, null, null, null, null ], "score": 11, "rank": 5 }, { "id": "real-roshn", "label": "ROSHN Group", "values": [ null, null, 1, null, null, 1, null, null, null ], "score": 7, "rank": 6 }, { "id": "real-diyar", "label": "Diyar Al Muharraq", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "real-mabanee", "label": "Mabanee", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "real-naseej", "label": "Naseej", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null }, { "id": "real-tamdeen", "label": "Tamdeen Group", "values": [ null, null, null, null, null, null, null, null, null ], "score": null, "rank": null } ], "note": "Source: Mayden PE-2026.10-v4. Each row is a named entity scope. U does not mean absent capability. Anchors are ordinal evidence-specificity judgements; scopes and disclosure access differ.", "displayNumber": 2 }, { "id": "E02", "type": "bars", "title": "Documented evidence points in the selected cohort", "subtitle": "Fixed denominator of 27; normalized to 0–100; ties use unrounded totals", "maximum": 100, "rows": [ { "label": "Aldar Properties", "value": 67, "secondary": 7, "rank": 1 }, { "label": "Qatari Diar", "value": 26, "secondary": 3, "rank": 2 }, { "label": "Red Sea Global", "value": 22, "secondary": 2, "rank": 3 }, { "label": "Emaar Properties", "value": 19, "secondary": 2, "rank": 4 }, { "label": "United Development Company", "value": 11, "secondary": 1, "rank": 5 }, { "label": "ROSHN Group", "value": 7, "secondary": 2, "rank": 6 }, { "label": "Diyar Al Muharraq", "value": null, "secondary": 0, "rank": null }, { "label": "Mabanee", "value": null, "secondary": 0, "rank": null }, { "label": "Naseej", "value": null, "secondary": 0, "rank": null }, { "label": "Tamdeen Group", "value": null, "secondary": 0, "rank": null } ], "note": "Source: frozen release calculation. These are public-evidence points, not a maturity percentage, operational audit or population ranking. No-evidence records stay unranked.", "displayNumber": 3 }, { "id": "E03", "type": "stages", "title": "Separate a commitment from a deployed practice", "subtitle": "Eligible claim records, n = 20; one practice may support multiple distinct themes", "maximum": 13, "rows": [ { "label": "Specific commitment / pilot", "value": 4 }, { "label": "Implemented practice", "value": 3 }, { "label": "Implemented + operating detail", "value": 13 } ], "note": "Source: claim ledger. 7 excluded records are outside these bars. Counts are claims, not independent systems, companies or estimates of adoption.", "displayNumber": 4 }, { "id": "E04", "type": "dependency", "title": "How much of the profile rests on one retained source URL?", "subtitle": "Baseline points versus points after removing the most influential primary source URL", "maximum": 100, "rows": [ { "label": "Aldar Properties", "value": 66.7, "after": 22.22 }, { "label": "Qatari Diar", "value": 25.9, "after": 3.7 }, { "label": "Red Sea Global", "value": 22.2, "after": 0.0 }, { "label": "Emaar Properties", "value": 18.5, "after": 0.0 }, { "label": "United Development Company", "value": 11.1, "after": 0.0 }, { "label": "ROSHN Group", "value": 7.4, "after": 0.0 }, { "label": "Diyar Al Muharraq", "value": null, "after": null }, { "label": "Mabanee", "value": null, "after": null }, { "label": "Naseej", "value": null, "after": null }, { "label": "Tamdeen Group", "value": null, "after": null } ], "note": "Source: deterministic URL-removal check. Remaining URLs can repeat the same disclosure; required corroborating-source dependencies are not removed. This is not a confidence interval or a test of source truth.", "displayNumber": 5 } ], "cases": [ { "companyId": "real-aldar", "title": "Aldar Properties", "scope": "Aldar Properties PJSC group; cases distinguish corporate teams, Aldar Estates, construction and district operations.", "claims": [ { "id": "REA-008", "claim": "Aldar reports deployed district AI with Siemens and AI parking through Aldar Estates/Parkonic.", "url": "https://app-aprp01saasa21mtp001.cms.optimizely.com/globalassets/aldar-innovation-report-2025---eng---final-09-april.pdf", "locator": "PDF10, Partnership Highlights", "level": 3, "eligible": true, "limitation": "Claim scope: Aldar Properties PJSC group; cases distinguish corporate teams, Aldar Estates, construction and district operations.. Parking200,000+ is design capacity, not confirmed active spaces; no causal energy outcome established. Application: internal-operating-practice.", "rationale": "Original innovation report page10 explicitly describes deployed Siemens district system and Parkonic parking operation. 200,000+ spaces is design capacity; no achieved energy reduction inferred." }, { "id": "REA-009", "claim": "Aldar embedded AI and geospatial specialists in its Innovation and Strategy team for several months.", "url": "https://app-aprp01saasa21mtp001.cms.optimizely.com/globalassets/aldar-innovation-report-2025---eng---final-09-april.pdf", "locator": "PDF14, Market Intelligence Project", "level": 3, "eligible": true, "limitation": "Claim scope: Aldar Properties PJSC group; cases distinguish corporate teams, Aldar Estates, construction and district operations.. No enterprise-wide workforce competence or training claim; specialist integration only. Application: internal-operating-practice.", "rationale": "Page14 describes AI and geospatial specialists embedded for several months in a named team. Retain for operating workforce arrangement, not workforce-wide proficiency." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Name the asset, phase and operating team." }, { "companyId": "real-qatari-diar", "title": "Qatari Diar", "scope": "Qatari Diar Real Estate Investment Company; Lusail operated through Lusail Real Estate Development Company. No inference to global portfolio.", "claims": [ { "id": "REA-020", "claim": "ST Engineering describes AI city operations deployed across80 Lusail sites, connecting video and operational systems.", "url": "https://www.stengg.com/en/innovation/artificial-intelligence/how-ai-is-powering-the-next-generation-of-smart-cities/", "locator": "Lusail case, lines902–909 and934", "level": 3, "eligible": true, "limitation": "Claim scope: Qatari Diar through LREDC/Lusail city operations, not global Qatari Diar portfolio.. Undated page observed at cutoff; contract runs to2027, so do not claim all contractual work complete. Application: internal-operating-practice.", "rationale": "Supplier full account explicitly connects deployed Lusail operations to video, sensor and operational systems across80 sites. Vendor assertion, not independent operating audit." }, { "id": "REA-021", "claim": "ST Engineering describes Lusail’s CityAI Assistant supporting operator questions and workflow commands.", "url": "https://www.stengg.com/en/innovation/artificial-intelligence/how-ai-is-powering-the-next-generation-of-smart-cities/", "locator": "CityAI Assistant and deployment summary, lines924,929–934", "level": 3, "eligible": true, "limitation": "Claim scope: Qatari Diar Real Estate Investment Company; Lusail operated through Lusail Real Estate Development Company. No inference to global portfolio.. No independently observed use, response accuracy or authorisation details; underlying supplier account. Application: internal-operating-practice.", "rationale": "Supplier explicitly includes CityAI Assistant in the Lusail deployment, with operator natural-language queries and workflow commands. No inference that all described emergency actions operate without human approval." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Name the asset, phase and operating team." }, { "companyId": "real-diyar", "title": "Diyar Al Muharraq", "scope": "Bahrain integrated master developer Diyar Al Muharraq; excludes unrelated similarly named developers.", "claims": [ { "id": "REA-026", "claim": "Diyar sponsored a smart-city summit whose agenda included AI.", "url": "https://www.diyar.bh/en/press/97/diyar-al-muharraq-official-platinum-sponsor-for-the-bahrain-smart-cities-summit-2024", "locator": "Conference agenda and sponsorship paragraphs", "level": null, "eligible": false, "limitation": "Claim scope: Bahrain integrated master developer Diyar Al Muharraq; excludes unrelated similarly named developers.. Excluded: smart meters, LEDs and solar trees are not themselves AI deployments. Application: excluded-or-scope-unresolved. Exclusion: New claim has no explicit accepted review decision. Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass.", "rationale": "Full release establishes conference sponsorship and AI as an agenda topic. It does not establish a named internal AI operating practice or implementation partnership. Uphold exclusion. Exclusion: New claim has no explicit accepted review decision. Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Excluded by separate AI-assisted source-review pass." } ], "boundary": "Company/group scope follows each cited claim. Headquarters is not the geography of all operations. Public disclosures and search completeness differ; no national or population inference.", "question": "Name the asset, phase and operating team." } ], "agenda": [ { "question": "Name the asset, phase and operating team.", "evidence": "Design, build, sales or facilities workflow; asset scope and recurring usage." }, { "question": "Connect the insight to an accountable action.", "evidence": "Who may change schedules, maintenance or commercial decisions and under what authority." }, { "question": "Separate a smart-city promise from delivered operations.", "evidence": "Commissioned systems, operating contracts, utilization and realized outcomes." } ], "sources": [ { "title": "Aldar Innovation Report2025", "publisher": "Aldar Properties", "date": "Undated", "url": "https://app-aprp01saasa21mtp001.cms.optimizely.com/globalassets/aldar-innovation-report-2025---eng---final-09-april.pdf", "locator": "PDF10, Partnership Highlights", "originalUrl": "https://app-aprp01saasa21mtp001.cms.optimizely.com/globalassets/aldar-innovation-report-2025---eng---final-09-april.pdf", "access": "full-source-read | Review: full-source-read", "relationship": "innovation-report", "id": "S01" }, { "title": "Annual Report2025 — full issuer document mirrored by FinancialFilings", "publisher": "Aldar Properties; mirror FinancialFilings", "date": "2026-03-18", "url": "https://financialfilings.com/filings/aldar-properties-pjsc/annual-report/2026/48374610/", "locator": "Mirror lines881–883; Live Aldar section", "originalUrl": "https://financialfilings.com/filings/aldar-properties-pjsc/annual-report/2026/48374610/", "access": "full-mirror-read | Review: full-mirror-read", "relationship": "annual-report", "id": "S02" }, { "title": "Q4/FY2025 financial results", "publisher": "Aldar Properties", "date": "2026-02-09", "url": "https://www.aldar.com/en/news-and-media/aldar-q4-fy-2025-financial-results", "locator": "Group & Corporate Updates, DigitalSpine bullet", "originalUrl": "https://www.aldar.com/en/news-and-media/aldar-q4-fy-2025-financial-results", "access": "full-source-read", "relationship": "official-implementation-account", "id": "S03" }, { "title": "How AI is powering the next generation of smart cities", "publisher": "ST Engineering", "date": "Undated", "url": "https://www.stengg.com/en/innovation/artificial-intelligence/how-ai-is-powering-the-next-generation-of-smart-cities/", "locator": "Lusail case, lines902–909 and934", "originalUrl": "https://www.stengg.com/en/innovation/artificial-intelligence/how-ai-is-powering-the-next-generation-of-smart-cities/", "access": "full-source-read | Review: full-source-read", "relationship": "supplier-case-study", "id": "S04" }, { "title": "ST Engineering wins AI smart-city project in Lusail", "publisher": "ST Engineering", "date": "2024-10-14", "url": "https://www.stengg.com/en/newsroom/news-releases/st-engineering-drive-ai-powered-smart-city-transformation", "locator": "Contract opening;2024–2027schedule", "originalUrl": "https://www.stengg.com/en/newsroom/news-releases/st-engineering-drive-ai-powered-smart-city-transformation", "access": "full-source-read", "relationship": "supplier-contract-announcement", "id": "S05" }, { "title": "Innovation meets preservation: technology’s impact on Red Sea biodiversity conservation", "publisher": "Red Sea Global", "date": "2024-05-22", "url": "https://www.redseaglobal.com/en/media-center/blog/innovation-meets-preservation-technology-s-impact-on-red-sea-biodiversity-conservation/", "locator": "CoralNet paragraphs, lines392–393", "originalUrl": "https://www.redseaglobal.com/en/media-center/blog/innovation-meets-preservation-technology-s-impact-on-red-sea-biodiversity-conservation/", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S06" }, { "title": "Emaar Hospitality Group first to bring AI technology into UAE kitchens", "publisher": "Winnow", "date": "2019-05-14", "url": "https://www.winnowsolutions.com/resources/news/emaar-hospitality-group-first-to-bring-cutting-edge-technology-into-uae-kitchen", "locator": "Deployment and camera/scales/ML paragraphs", "originalUrl": "https://www.winnowsolutions.com/resources/news/emaar-hospitality-group-first-to-bring-cutting-edge-technology-into-uae-kitchen", "access": "full-source-read | Review: full-source-read", "relationship": "supplier-case-study", "id": "S07" }, { "title": "Emaar Properties Integrated Annual Report2024 — inaccessible current issuer URL", "publisher": "Emaar Properties", "date": "Undated", "url": "https://properties.emaar.com/wp-content/uploads/2025/03/Emaar-Properties-IR-2024_English_F.pdf", "locator": "Indexed printed45", "originalUrl": "https://properties.emaar.com/wp-content/uploads/2025/03/Emaar-Properties-IR-2024_English_F.pdf", "access": "snippet-only", "relationship": "annual-report", "id": "S08" }, { "title": "United Development Company wins three AI & Analytics Summit awards", "publisher": "United Development Company", "date": "2025-11-17", "url": "https://udcqatar.com/Media-Room/Press-Releases/Middle-East-Enterprise-AI", "locator": "CAFM ecosystem and Super App paragraphs", "originalUrl": "https://udcqatar.com/Media-Room/Press-Releases/Middle-East-Enterprise-AI", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S09" }, { "title": "Annual Report2025", "publisher": "United Development Company", "date": "Undated", "url": "https://udcqatar.com/-/media/28448e61b207417e9756f8027195f1a5.ashx", "locator": "PDF34/printed66–67, QatarCool2025highlights", "originalUrl": "https://udcqatar.com/-/media/28448e61b207417e9756f8027195f1a5.ashx", "access": "full-source-read | Review: full-source-read", "relationship": "annual-report", "id": "S10" }, { "title": "ROSHN Group launches registration for ROSHN Hackathon2025", "publisher": "ROSHN", "date": "Undated", "url": "https://www.roshn.sa/news-and-events/press-releases/article-78", "locator": "AI track and pilot-opportunity paragraphs", "originalUrl": "https://www.roshn.sa/news-and-events/press-releases/article-78", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S11" }, { "title": "Diyar sponsors Bahrain Smart Cities Summit2024", "publisher": "Diyar Al Muharraq", "date": "2024-02-11", "url": "https://www.diyar.bh/en/press/97/diyar-al-muharraq-official-platinum-sponsor-for-the-bahrain-smart-cities-summit-2024", "locator": "Conference agenda and sponsorship paragraphs", "originalUrl": "https://www.diyar.bh/en/press/97/diyar-al-muharraq-official-platinum-sponsor-for-the-bahrain-smart-cities-summit-2024", "access": "full-source-read | Review: full-source-read", "relationship": "official-implementation-account", "id": "S12" }, { "title": "Naseej AGM and new strategy", "publisher": "Naseej", "date": "2025-06-24", "url": "https://www.naseej.bh/our-news/naseej-holds-annual-general-meeting-and-announces-new-strategy", "locator": "AGM strategy paragraph", "originalUrl": "https://www.naseej.bh/our-news/naseej-holds-annual-general-meeting-and-announces-new-strategy", "access": "full-source-read", "relationship": "official-implementation-account", "id": "S13" }, { "title": "Tamdeen Entertainment unveils Planet Laser", "publisher": "Tamdeen Group", "date": "2025-06-29", "url": "https://www.tamdeen.com/news/tamdeen-entertainment-unveils-planet-laser-by-fun-tiki-at-al-khiran-mall", "locator": "AI gameplay paragraph and future launch sentence", "originalUrl": "https://www.tamdeen.com/news/tamdeen-entertainment-unveils-planet-laser-by-fun-tiki-at-al-khiran-mall", "access": "full-source-read", "relationship": "official-implementation-account", "id": "S14" }, { "title": "Mabanee Annual Report 2025", "publisher": "Primary source as linked", "date": "See original", "url": "https://www.mabanee.com/storage/storageannual-reports/annual-report-2025-eng-2070077368.pdf", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S15" }, { "title": "sustainability report register", "publisher": "Primary source as linked", "date": "See original", "url": "https://www.mabanee.com/en/sustainability/sustainability-reports", "locator": "Linked publication; see cited passage in report", "relationship": "Research or official publication", "id": "S16" } ], "body": "# Real Estate & Construction: intelligence must survive contact with operations\n\n*Mayden AI Native Index — exploratory public-evidence research. Cutoff: 10 October 2026. Ten purposefully selected developer and owner-operator groups across five GCC markets. This edition measures documented organisational practice, not investment quality, building safety or national AI maturity.*\n\nA developer becomes more interesting as an AI enterprise when intelligence changes a recurring decision: how a design is assessed, a construction discrepancy is detected, an asset is maintained or a service request is resolved. A smart-city label does not answer those questions. Neither does a digital twin, a technology partnership or an innovation award on its own. The evidence becomes useful when it identifies the operation, the responsible entity and the mechanism that connects information to action.\n\nThis sector presents an unusually difficult measurement problem. A developer may own hotels, operate a district, manage schools, invest in a cooling company and commission construction from contractors. AI used in any one of those activities cannot automatically be attributed to every other activity. A group profile can legitimately include several businesses, but it must preserve where each claim originates. Without that discipline, size and corporate complexity can masquerade as enterprise-wide adoption.\n\nThe purposeful cohort includes ROSHN Group and Red Sea Global in Saudi Arabia; Emaar Properties and Aldar Properties in the UAE; Qatari Diar and United Development Company in Qatar; Mabanee and Tamdeen Group in Kuwait; and Diyar Al Muharraq and Naseej in Bahrain. These are primarily developers and asset owner-operators, not a matched sample of general contractors. There are two headquarters per market, but no basis for estimating country adoption or ranking national ecosystems from this selection.\n\n## Five operating patterns worth examining\n\nRed Sea Global offers a specific example in environmental operations. Its marine director describes using CoralNet to analyse recurring coral surveys, training the model with human-labelled images and checking its performance against skilled analysts. That is more informative than a general claim about sustainable technology. It identifies a working process and an evaluation step. It remains a marine-monitoring case, not evidence that the entire development enterprise has become AI-native. [RSG implementation account](https://www.redseaglobal.com/en/media-center/blog/innovation-meets-preservation-technology-s-impact-on-red-sea-biodiversity-conservation/)\n\nEmaar's case is operationally narrow and historically older. Winnow's 2019 account describes AI-enabled food-waste recognition in hospitality kitchens, using images and scales to inform purchasing and production choices. The source discusses installation of waste technology across properties, but that total should not be read as a proven count of AI Vision installations. The record does not independently confirm continuous operation through October 2026. [Winnow's Emaar implementation account](https://www.winnowsolutions.com/resources/news/emaar-hospitality-group-first-to-bring-cutting-edge-technology-into-uae-kitchen)\n\nAldar's innovation report supplies several distinct internal mechanisms: specialist teams, security automation, construction-partner feedback and payment interfaces. It also labels an ESG chatbot beta and puts autonomous asset agents in a future section. Those distinctions matter as much as the positive cases. [Aldar Innovation Report 2025, pages 10, 14, 16–18, 29 and 37](https://app-aprp01saasa21mtp001.cms.optimizely.com/globalassets/aldar-innovation-report-2025---eng---final-09-april.pdf)\n\nQatari Diar's evidence is scoped to Lusail through its development subsidiary. ST Engineering's contract announcement describes a programme scheduled through 2027; a later, undated implementation account describes deployed city operations and an operator assistant. Both belong in the record. The later account supports documented implementation, while the contract prevents a careless claim that every contractual deliverable is complete. Supplier-reported deployment remains supplier-reported. [2024 contract](https://www.stengg.com/en/newsroom/news-releases/st-engineering-drive-ai-powered-smart-city-transformation), [Lusail operating account](https://www.stengg.com/en/innovation/artificial-intelligence/how-ai-is-powering-the-next-generation-of-smart-cities/)\n\nUDC's November 2025 announcement describes an AI assistant connected to facilities services and its resident platform. The useful evidence is the workflow description, not the awards. Its annual report separately describes AI initiatives at Qatar Cool, identifies a 51% stake and places the company in a section headed Major Associates. That related-investee case is excluded from UDC points: the majority stake alone does not establish controlled or consolidated operating scope. [UDC implementation account](https://udcqatar.com/Media-Room/Press-Releases/Middle-East-Enterprise-AI), [UDC Annual Report 2025, pages 66–67](https://udcqatar.com/-/media/28448e61b207417e9756f8027195f1a5.ashx)\n\nTogether, these cases suggest that AI can enter the development enterprise through many doors. Environmental monitoring, hospitality operations, construction delivery and resident services may develop at different speeds. A single narrative about “the AI-powered developer” conceals that variation. A dimension profile and a scoped case portfolio are more useful to an executive than a score without its supporting operating detail.\n\n## Commitments, products and missing evidence\n\nROSHN's hackathon provides evidence of a specific AI-inclusive innovation programme with named partners and opportunities for pilots. It does not establish that the submitted solutions became routine operating systems. An ecosystem commitment and a completed implementation deserve different anchors. The number of submissions or size of the prize pool should not become a proxy for organisational capability. [ROSHN Hackathon 2025 announcement](https://www.roshn.sa/news-and-events/press-releases/article-78)\n\nMabanee's 2025 annual report was recovered and screened, including its operational and governance material. The bounded search did not identify a qualifying AI claim in that document. Its sustainability-report register was accessible, but the linked 2025 sustainability PDF was not successfully recovered during the research window. Those are distinct observations: one document was inspected, another remains an access gap. Neither supports a conclusion that Mabanee lacks AI capability. [Mabanee Annual Report 2025](https://www.mabanee.com/storage/storageannual-reports/annual-report-2025-eng-2070077368.pdf), [sustainability report register](https://www.mabanee.com/en/sustainability/sustainability-reports)\n\nTamdeen's Planet Laser announcement describes AI-enhanced gameplay in an entertainment offering. The article does not establish a developer's internal AI operating process, and its opening language is mixed with a forthcoming launch date. Diyar Al Muharraq's summit sponsorship includes AI in the event agenda, while the implemented city measures described are smart meters, efficient lighting and solar installations. Those technologies should not be relabelled AI. [Tamdeen announcement](https://www.tamdeen.com/news/tamdeen-entertainment-unveils-planet-laser-by-fun-tiki-at-al-khiran-mall), [Diyar announcement](https://www.diyar.bh/en/press/97/diyar-al-muharraq-official-platinum-sponsor-for-the-bahrain-smart-cities-summit-2024)\n\nNaseej's 2025 strategy announcement discusses growth and innovation without a specific operating AI commitment. Search results for similarly named Saudi technology businesses are especially hazardous: they describe a different organisation. The Bahrain developer remains in the cohort with unknown dimensions where qualifying evidence was not recovered. [Naseej strategy announcement](https://www.naseej.bh/our-news/naseej-holds-annual-general-meeting-and-announces-new-strategy)\n\nKeeping these firms visible is part of a credible benchmark. A research team could easily make an index look stronger by replacing quieter firms with companies that publish more AI content. That would change the selection process after seeing the results. Mayden's better approach is to disclose the boundary, preserve the unknowns and offer a consistent route for organisations to supply further evidence.\n\n## The enterprise boundary is a measurement decision\n\nBefore an interview begins, the organisation should confirm exactly what is being assessed. Is it the listed parent, a consolidated group, a property-management subsidiary or one development? Which operating partners are within scope? Who owns the decision being described? These questions are not administrative details. They determine whether two observations concern the same kind of enterprise.\n\nFor a contractor-delivered project, the developer may benefit from AI without owning the capability. That can still be relevant ecosystem evidence if the arrangement changes the developer's work and the responsibilities are documented. It is less persuasive as evidence of an internal workforce or knowledge capability. An outsourced inspection service does not establish that the developer can reproduce, evaluate or govern its outputs independently.\n\nOwnership percentages also need careful interpretation. They help describe the economic relationship but do not by themselves establish operational control. Where a source labels an investee differently from the researcher's initial assumption, the discrepancy should remain visible for adjudication. Company-confirmed scope and financial consolidation notes can resolve it in a proprietary assessment. The researcher should not silently choose the interpretation that produces more points.\n\nHeadquarters country is similarly limited. A Kuwait-headquartered developer may operate assets elsewhere; a UAE group may hold international businesses. Country filters should identify headquarters, with operating geography shown separately. Aggregating the resulting profiles into national performance claims would require a different sample design, appropriate denominators and a clearly defined population.\n\n## Nine questions that reveal the operating model\n\n**AI in core work.** Which recurring development or asset decision has changed? Follow one workflow from input to decision and execution. A strong answer names the project or asset population, the responsible team and the system's actual contribution. It also explains what remains manual. Ask for a recent example rather than a list of software purchased.\n\n**Workforce.** Who performs work differently because of AI? Embedded specialists, trained business users and changed approval roles are different forms of evidence. Ask which roles can evaluate an output, challenge it and work safely when a tool fails. Training hours are useful only when the relevant population and expected competence are clear.\n\n**Ecosystem.** How do suppliers, contractors and internal teams learn together? A meaningful partnership account identifies data exchanges, integration responsibilities, acceptance criteria and a route for operational feedback. A technology vendor's capability statement is not evidence that the developer uses all of those capabilities. Implementation contracts can establish commitments; operating records establish delivery.\n\n**Interfaces.** What can employees, residents or contractors actually complete through an AI interaction? Separate answering a question from initiating a maintenance job, changing a booking or paying a charge. Track where the interaction transfers to a person and whether the transaction reaches the system of record. Attractive demonstrations should not substitute for ordinary service performance.\n\n**Lineage.** Can a project decision be reconstructed from its inputs, model version and approvals? In construction, this could involve knowing which design revision and site scan informed a discrepancy. In leasing, it could involve the information underlying a recommendation. A digital twin or central data platform does not automatically provide this traceability. Ask for an example that follows the complete chain.\n\n**Experimentation.** How does the firm decide whether to expand, revise or stop a use case? Define the comparison, affected process, acceptance criteria and consequences of error. A pilot that exposes limitations can be organisationally valuable even if it never scales. The public-evidence anchor nevertheless remains a pilot anchor until routine implementation is demonstrated.\n\n**Autonomy.** Which actions can a system take without approval? Automated detection, suggested responses and completed changes are distinct. Ask who may change permissions, how exceptional conditions are handled and whether the action can be reversed. The word “agent” is not a substitute for an authority map, particularly when the system affects access, payments or physical infrastructure.\n\n**Knowledge.** How do insights become reusable organisational learning? A useful answer connects analysis to a decision and then to observed results. It explains who checks the output and how corrections influence future work. An AI-generated report can improve access to information, but a learning system requires a feedback process that survives changes in staff and suppliers.\n\n**Governance.** Who approves deployment and who can stop it? General corporate governance, cybersecurity and ESG reporting are important contexts, but they are not automatically AI governance. Seek the operating owner, risk review, model-change approval and incident process. Evidence of one carefully governed workflow can be more revealing than an organisation-wide policy without implementation detail.\n\nThe questions should be tested with different respondents before being used as a validated instrument. Development directors, asset managers, technology leaders and risk owners see different parts of the operation. A structured interview should route questions to the people who can substantiate the answers. Divergent answers may reveal an unclear responsibility boundary rather than a respondent error.\n\n## Outcomes need a denominator and a counterfactual\n\nA source may report shorter design cycles, automated security hours, reduced waste or faster responses. Those measures have different meanings. Some describe throughput, some estimates of avoided effort, some predictions and some realised financial effects. The evidence register preserves the original category. A projected saving should never be displayed as an achieved result, and a technology's design capacity should never become its active deployment count.\n\nFor construction, a useful evaluation identifies the project phase, baseline method, inspected elements, detected issues and follow-through. If a system flags a potential defect, the resulting value depends on review, correction and avoided rework. Counting flags alone can reward noise. Comparing the same operating process before and after introduction is a start, but changes in project mix and delivery conditions still need consideration.\n\nFor facilities and resident services, transaction completion and exception handling are more informative than interaction volume alone. A busy assistant may be popular, confused or handling repeated unresolved requests. Track the unit of work, repeat contacts, human escalation and verified resolution. The right measures depend on the service and should be established before promotional claims are produced.\n\nFor environmental applications, better measurement is not identical to better environmental outcomes. An image-analysis tool can improve monitoring capacity without demonstrating an ecological improvement. The evidence chain must show how monitoring informed management and what happened afterwards. This distinction makes a report more useful: it separates a credible operational contribution from a much broader claim that has not yet been tested.\n\n## How the public benchmark should be presented\n\nThe released profile should retain the nine discrete anchors and show unknown explicitly. A named commitment or pilot supports level 1; an implemented internal practice supports level 2; an implemented practice with a concrete mechanism, coverage, evaluation or authority can support level 3 after a separate AI-assisted source challenge. Duplicate announcements do not accumulate points. The total is a summary of documented evidence, not a percentage of enterprise maturity.\n\nAn evidence reader should let an executive inspect the company boundary, operating entity, source date, access status and coding rationale. A full mirror must be labelled as a mirror, as with the recovered Aldar annual report. Failed original links should remain in the provenance record. An accessible homepage reached through a broken PDF link is not the missing annual report. [Aldar Annual Report 2025, readable mirror](https://financialfilings.com/filings/aldar-properties-pjsc/annual-report/2026/48374610/)\n\nRead the delivered exhibits as complementary views of the same bounded record. The heatmap locates documented themes and unknowns; the points chart summarises those anchors without measuring development quality or enterprise maturity. The evidence-anchor chart distinguishes commitments and pilots from implemented practices and operating detail. Its bars count claims, so one underlying system can contribute more than one record. The source-URL-removal chart tests concentration by omitting the most influential retained URL. A lower resulting total indicates dependence on that disclosure, not a measured decline in capability. Repeated accounts can share an origin, so remaining URLs do not establish independent corroboration.\n\nFor press, the defensible story is that documented intelligence appears in specific operating processes while evidence of enterprise-wide coordination remains uneven. That is a hypothesis generated by this purposeful cohort, not a representative sector prevalence estimate. Editors should be able to follow each illustrative claim to a readable source and understand why a visually impressive counterexample was excluded.\n\nFor participating companies, the next step is a focused verification exercise. Select one important workflow, identify its owner, collect a redacted operating map, clarify authority and document how success is evaluated. Then examine whether the same organisational routines are shared across other use cases. This can turn a public-evidence profile into a valuable conversation about the enterprise without pretending that a website search has already answered every question.\n\nA durable index will reward clearer evidence and better operating practice through consistent rules. It will also record stopped pilots, revised claims and genuine uncertainty. Over repeated editions, that history can become more valuable than a single ranking: it shows how firms move from a technology announcement to an accountable capability that continues to work across projects, assets and teams.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/realestate-public-evidence.pdf", "cover": "/research/index/2026-10/realestate-public-evidence.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 94889, "pdfSha256": "4c4eab224233a8e0a0436b95db28dfc272b1c1c100cc81fc1e91fcd41ec868c7", "pdfPages": 18 }, { "slug": "state-of-ai-native-enterprise", "number": 10, "title": "The AI-native enterprise: an operating evidence atlas", "shortTitle": "The operating evidence atlas", "dek": "Eight sectors. Five markets. A closer examination of how AI enters recurring work.", "sector": null, "kind": "Measurement research", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "59 of 80 selected entity records contain qualifying AI operating evidence in this review. 21 remain unranked because no qualifying claim was located; this does not establish absent capability.", "The corpus contains 272 recorded claims, of which 224 meet the declared eligibility rules. Specific commitments, deployed practices and operating detail are coded separately.", "Decision Lineage appears in 4 of these 80 profiles, compared with 44 for AI Core. These are counts within the selected research corpus, not enterprise adoption rates." ], "executiveSources": [], "evidenceAgeContext": "Dated sources run from 2019-05-14 to 2026-10-01. 79 of 224 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": null, "narrativeExhibits": [ { "id": "N01", "displayNumber": 3, "title": "Five questions between an AI demonstration and an enterprise result", "kind": "narrative" }, { "id": "N02", "displayNumber": 4, "title": "Where to begin the operating interview", "kind": "narrative" }, { "id": "N03", "displayNumber": 5, "title": "A value account with separate, auditable lines", "kind": "narrative" }, { "id": "N04", "displayNumber": 6, "title": "Put scrutiny at the point of consequence", "kind": "narrative" }, { "id": "N05", "displayNumber": 7, "title": "Three kinds of coverage that must stay separate", "kind": "narrative" }, { "id": "N06", "displayNumber": 8, "title": "The executive challenge for the next ninety days", "kind": "narrative" } ], "exhibitRegistry": [ { "id": "E01", "displayNumber": 1, "title": "Where evidence is visible across the eight sector cohorts", "kind": "generated" }, { "id": "E02", "displayNumber": 2, "title": "The evidence record includes what did not qualify", "kind": "generated" }, { "id": "N01", "displayNumber": 3, "title": "Five questions between an AI demonstration and an enterprise result", "kind": "narrative" }, { "id": "N02", "displayNumber": 4, "title": "Where to begin the operating interview", "kind": "narrative" }, { "id": "N03", "displayNumber": 5, "title": "A value account with separate, auditable lines", "kind": "narrative" }, { "id": "N04", "displayNumber": 6, "title": "Put scrutiny at the point of consequence", "kind": "narrative" }, { "id": "N05", "displayNumber": 7, "title": "Three kinds of coverage that must stay separate", "kind": "narrative" }, { "id": "N06", "displayNumber": 8, "title": "The executive challenge for the next ninety days", "kind": "narrative" }, { "id": "E03", "displayNumber": 9, "title": "Operating detail deserves a separate test", "kind": "generated" }, { "id": "E04", "displayNumber": 10, "title": "Build a benchmark from evidence to a decision", "kind": "generated" } ], "exhibitOrder": [ "E01", "E02", "N01", "N02", "N03", "N04", "N05", "N06", "E03", "E04" ], "metrics": [ { "value": 80, "label": "selected entity records" }, { "value": 272, "label": "recorded claims" }, { "value": 224, "label": "eligible claims" }, { "value": 21, "label": "unranked records" } ], "exhibits": [ { "id": "E01", "type": "coverage", "title": "Where evidence is visible across the eight sector cohorts", "subtitle": "Number of selected entity records with a qualifying claim in each theme; cohort size shown per row", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "label": "Aviation & Logistics", "n": 10, "values": [ 7, 3, 5, 3, 1, 2, 0, 2, 2 ] }, { "label": "Banking", "n": 10, "values": [ 7, 8, 4, 7, 2, 3, 2, 4, 7 ] }, { "label": "Energy & Utilities", "n": 10, "values": [ 7, 6, 5, 3, 0, 3, 1, 3, 1 ] }, { "label": "Healthcare", "n": 10, "values": [ 6, 3, 3, 4, 0, 3, 0, 1, 4 ] }, { "label": "Manufacturing & Materials", "n": 10, "values": [ 5, 2, 2, 1, 0, 3, 0, 2, 2 ] }, { "label": "Real Estate & Construction", "n": 10, "values": [ 4, 1, 4, 3, 0, 3, 1, 1, 0 ] }, { "label": "Retail & Consumer", "n": 10, "values": [ 3, 1, 3, 5, 0, 0, 1, 1, 2 ] }, { "label": "Telecommunications", "n": 10, "values": [ 5, 5, 3, 4, 1, 2, 3, 3, 3 ] } ], "note": "Source: frozen Mayden public-evidence corpus. Counts describe located evidence, not comparable sector maturity or prevalence. Cohorts are purposeful, heterogeneous and sometimes overlap.", "displayNumber": 1 }, { "id": "E02", "type": "funnel", "title": "The evidence record includes what did not qualify", "subtitle": "Recorded claims pass explicit access, scope, deployment and review gates", "maximum": 272, "rows": [ { "label": "All recorded claims", "value": 272 }, { "label": "Eligible for scoring", "value": 224 }, { "label": "Excluded / held", "value": 48 } ], "note": "Source: release claim ledger. These counts do not measure research recall, factual accuracy or company capability. All exclusions remain inspectable in the public dataset.", "displayNumber": 2 }, { "id": "E03", "type": "stages", "title": "Operating detail deserves a separate test", "subtitle": "Eligible claim records, n = 224; disclosure anchors are not maturity stages", "maximum": 97, "rows": [ { "label": "Specific commitment / pilot", "value": 53 }, { "label": "Implemented practice", "value": 74 }, { "label": "Implemented + operating detail", "value": 97 } ], "note": "Source: reviewed coding. Detail is not independent assurance, and multiple claim records may describe a shared underlying practice.", "displayNumber": 9 }, { "id": "E04", "type": "process", "title": "Build a benchmark from evidence to a decision", "subtitle": "Mayden research architecture; conceptual framework, not an observed causal model", "rows": [ { "label": "Define", "detail": "Entity, workflow, population and period." }, { "label": "Observe", "detail": "Sources, tasks, usage and operating boundaries." }, { "label": "Challenge", "detail": "Alternative explanations, controls and exclusions." }, { "label": "Calculate", "detail": "Declared rules, sensitivity and reproducibility." }, { "label": "Decide", "detail": "Specific action, evidence gap and next test." } ], "note": "Source: Mayden methodological synthesis. Confidential company research remains a separate instrument requiring permission, respondent testing and validation.", "displayNumber": 10 } ], "cases": [], "agenda": [ { "question": "Choose one recurring workflow.", "evidence": "Its users, decisions, frequency, business importance and operating boundary." }, { "question": "Make authority and evidence inspectable.", "evidence": "Input provenance, evaluations, interventions, accountable owner and recovery path." }, { "question": "Track value with a credible comparator.", "evidence": "Quality, cost, risk and adoption denominators over a stated period." } ], "sources": [ { "title": "Large Language Model Influence on Diagnostic Reasoning: A Randomized Clinical Trial", "publisher": "JAMA Network Open", "date": "2024-10-28", "url": "https://doi.org/10.1001/jamanetworkopen.2024.40969", "locator": "Stratified randomised, blinded-rating trial", "relationship": "peer-reviewed-randomized-trial", "id": "S01" }, { "title": "GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks", "publisher": "OpenAI author preprint", "date": "2025-10-05", "url": "https://arxiv.org/html/2510.04374v1", "locator": "Expert-authored task benchmark with pairwise deliverable evaluation", "relationship": "provider-original-benchmark-preprint", "id": "S02" }, { "title": "RICS artificial intelligence in construction report 2025", "publisher": "Royal Institution of Chartered Surveyors", "date": "2025", "url": "https://www.rics.org/news-insights/artificial-intelligence-in-construction-report", "locator": "Professional-body online sentiment surveys", "relationship": "professional-body-original-survey", "id": "S03" }, { "title": "An Evaluation of DWP’s Microsoft 365 Copilot Trial", "publisher": "UK Department for Work and Pensions", "date": "2026-01-29", "url": "https://www.gov.uk/government/publications/an-evaluation-of-dwps-microsoft-copilot-365-trial/an-evaluation-of-dwps-microsoft-365-copilot-trial", "locator": "Nonrandom post-implementation surveys, adjusted regressions and interviews", "relationship": "official-original-evaluation", "id": "S04" }, { "title": "Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI", "publisher": "NBER working paper 33777", "date": "2026-03", "url": "https://www.nber.org/system/files/working_papers/w33777/w33777.pdf", "locator": "Linked adoption surveys and administrative records; observational difference-in-differences", "relationship": "empirical-working-paper-observational", "id": "S05" }, { "title": "Measuring AI agent autonomy in practice", "publisher": "Anthropic original research", "date": "2026-02-18", "url": "https://www.anthropic.com/research/measuring-agent-autonomy", "locator": "Provider telemetry with model-assisted classification", "relationship": "provider-original-observational", "id": "S06" }, { "title": "tau-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains", "publisher": "Author preprint / Sierra research", "date": "2024-06-17", "url": "https://arxiv.org/html/2406.12045v1", "locator": "Synthetic interactive agent benchmark with repeated trials", "relationship": "original-benchmark-preprint", "id": "S07" }, { "title": "SABER: Small Actions, Big Errors — Safeguarding Mutating Steps in LLM Agents", "publisher": "Amazon AGI Foundations author preprint", "date": "2025-11-26", "url": "https://arxiv.org/html/2512.07850v1", "locator": "Agent benchmark re-audit and safeguard experiments", "relationship": "original-benchmark-methods-preprint", "id": "S08" }, { "title": "The impact of AI on the workplace: Main findings from the OECD AI surveys of employers and workers", "publisher": "OECD Social, Employment and Migration Working Papers", "date": "2023-03-22", "url": "https://doi.org/10.1787/ea0a0fe1-en", "locator": "Separate employer telephone and worker online-panel surveys", "relationship": "official-original-survey", "id": "S09" } ], "body": "# The state of the AI-native enterprise\n## Follow the decision. Measure the operating change.\n\n*Mayden research perspective for the eight-sector evidence edition · October 2026*\n\nAn enterprise becomes meaningfully different when AI changes how work moves, who makes decisions and how the organisation learns from results. Buying access to a model can begin that change. A convincing demonstration can make it imaginable. Neither tells a board whether the business now operates differently.\n\nMayden proposes a practical definition: an AI-native enterprise repeatedly combines AI, people, data and decision authority inside its operating processes, with measurable results and a working capacity to correct failure. This is a research proposition to test, not a validated threshold that companies have already crossed.\n\nThe immediate executive question is concrete: **Which important decision works differently today, and what evidence would persuade someone outside the project team?** Answering it connects technology investment to business operations. It also exposes the difference between a useful isolated application and an organisational capability that survives a change of model, supplier or leadership.\n\nThis edition approaches that question across banking, telecommunications, energy and utilities, logistics and aviation, manufacturing and materials, retail and consumer businesses, healthcare, and real estate and construction. Its public evidence is a starting point for inquiry. It cannot observe every internal practice, and it does not estimate regional adoption or certify the most capable enterprises.\n\n## Exhibit 03 — Five questions between an AI demonstration and an enterprise result\n\n| Evidence layer | Executive question | Evidence that moves the discussion forward |\n|---|---|---|\n| Task capability | Can the system do the defined job? | Representative tests, quality criteria and failure cases |\n| Operating use | Does it handle real work? | Production dates, eligible workload and actual usage |\n| Decision authority | What may it decide or change? | Permissions, accountable owner and escalation rules |\n| Business consequence | Did the relevant outcome improve? | Baseline, comparison, full costs and observation period |\n| Organisational learning | Can the enterprise detect and correct deterioration? | Monitoring, incidents, changed procedures and retesting |\n\n*Mayden analytical framework. These are different evidence questions, not five measured maturity bands.*\n\nA board can use this sequence without waiting for a perfect enterprise score. Choose a workflow, ask for the evidence at each layer, and note the first unanswered question. That point often identifies the next useful investment: better data, a clearer owner, broader rollout, or a more credible evaluation.\n\n## 1. A capable model does not guarantee a capable workflow\n\nThe central unit of investigation should be a recurring business process with a clear beginning and end. “Our staff use AI” leaves too much unspecified. “The system proposes a replenishment order; a planner checks exceptions; an approved order updates inventory” describes work that can be inspected.\n\nResearch illustrates why this distinction matters. In a randomised study of 50 US physicians, chatbot access did not significantly improve the primary diagnostic-reasoning score on six clinical vignettes. The study concerns a bounded test setting, not patient outcomes, but it shows why performance of the combined workflow deserves its own evaluation. [Goh and colleagues, JAMA Network Open](https://doi.org/10.1001/jamanetworkopen.2024.40969).\n\nTask benchmarks answer another useful question. GDPval evaluates professional work products across selected occupations, while explicitly limiting claims about tacit context and interactive work. It offers ideas for realistic test materials; it does not measure whether a named enterprise has integrated AI successfully. [GDPval, version 1](https://arxiv.org/html/2510.04374v1).\n\nFor an executive, the operating test has four parts. Identify the eligible work, observe the human and system together, include the exceptions, and examine the final accepted output. An impressive draft that requires extensive repair is different from an accepted result. A fast decision that creates downstream rework may move cost rather than remove it.\n\nThe same logic applies to governance. A policy is useful evidence of intent. An operating control has a trigger, a responsible person and a consequence. The relevant question becomes: **Show one occasion when this control changed what the system or an employee did.**\n\n## 2. Eight sectors need different proof\n\nA common measurement backbone makes comparisons understandable. Sector-specific operating questions make them useful. A bank and an industrial operator may both need reliable data, accountable owners and trained staff; their material decisions and failure costs differ.\n\n## Exhibit 04 — Where to begin the operating interview\n\n| Sector | Start with this workflow | Ask for this denominator | Examine this exception |\n|---|---|---|---|\n| Banking | Fraud alert investigation or service resolution | Eligible alerts or cases, by type | Legitimate customer incorrectly blocked |\n| Telecommunications | Network fault detection and resolution | Covered sites, incidents or service lines | Escalation during a severe outage |\n| Energy and utilities | Equipment maintenance or process optimisation | Eligible assets and operating hours | Recommendation outside safe operating limits |\n| Logistics and aviation | Routing, disruption response or cargo handling | Eligible journeys, shipments or disruption cases | Late data or an infeasible alternative |\n| Manufacturing and materials | Defect inspection or production scheduling | Products, lines and inspected units | Unseen defect or changed material |\n| Retail and consumer | Forecasting, replenishment or returns | Eligible products, stores or transactions | Promotion, stockout or disputed refund |\n| Healthcare | Documentation, scheduling or clinical support | Encounters and sites within the approved scope | Missing information or disputed recommendation |\n| Real estate and construction | Design review, site monitoring or asset operations | Projects, buildings or inspected work packages | Conflicting design versions or unsafe site condition |\n\n*Illustrative evaluation agenda developed by Mayden; no row asserts a measured sector result or recommends autonomous clinical or safety decisions.*\n\nBegin with one process that matters economically or operationally, then ask whether the practice travels. A deployment at one flagship site may be valuable without being enterprise-wide. A parent group may announce a programme that reaches only one subsidiary. A technology service sold to customers may demonstrate commercial capability without showing the supplier's own internal transformation.\n\nThose distinctions should appear beside company findings. Otherwise, readers can mistake a broad corporate name for a broad deployment.\n\nIndustry surveys also require careful scope. RICS separates pilot activity from regular process use in its construction survey; its respondents are professionals, not a census of enterprises. Our local evidence collection needs similarly precise units. [RICS construction research](https://www.rics.org/news-insights/artificial-intelligence-in-construction-report).\n\n## 3. Measure the value that survives the whole workflow\n\nTime saved is an intermediate result. It can improve response speed, increase output, reduce strain or create room for better work. Translating it into financial value requires another observation: what happened to the released capacity?\n\nIn its Copilot evaluation, the UK Department for Work and Pensions reports user benefits alongside verification overhead. The evaluation relies on self-reported outcomes, nonrandom licence allocation and no pre-trial baseline. Its methodological candour is a useful model for separating a promising experience from a causal estimate. [DWP evaluation](https://www.gov.uk/government/publications/an-evaluation-of-dwps-microsoft-copilot-365-trial/an-evaluation-of-dwps-microsoft-365-copilot-trial).\n\nA Danish study linking adoption surveys to administrative records likewise distinguishes changes in work from recorded labour-market outcomes. Its March 2026 revision finds substantial task reorganisation alongside earnings and hours estimates near zero over the observed period. Different outcome layers can move at different speeds. [Humlum and Vestergaard](https://www.nber.org/system/files/working_papers/w33777/w33777.pdf).\n\n## Exhibit 05 — A value account with separate, auditable lines\n\n| Line in the account | What to record |\n|---|---|\n| Gross task effect | Time, quality, throughput or loss change on the defined task |\n| Additional work | Review, correction, escalation, integration and training effort |\n| Operating cost | Model use, software, infrastructure, support and monitoring |\n| Capacity use | More output, shorter queues, different work or reduced paid hours |\n| Business outcome | Margin, service, reliability or another chosen outcome |\n| Attribution limits | Demand changes, staffing changes and other concurrent initiatives |\n\n*Accounting structure, not a forecast. Do not add unlike units or convert every saved minute into cash.*\n\nThe chief financial officer and process owner should agree on the last two lines before a pilot starts. Choose one primary outcome and a small set of guardrails. Record the baseline and comparison design. Preserve unsuccessful runs. Where randomisation is practical, use it; where it is not, make the identification limits visible.\n\nA programme can be worth continuing before it generates a proven financial return. The investment decision becomes stronger when the organisation can say what it has learned, what remains uncertain and which next observation will resolve the uncertainty.\n\n## 4. Agentic systems make authority a measurement problem\n\nAn agent that drafts a recommendation and an agent that changes a customer record may use the same underlying model. Their operating implications differ. Executives need a map of permitted actions, dependencies and recovery routes.\n\nProvider telemetry reinforces the need to examine deployment context. Anthropic's original agent-use study distinguishes individual API tool calls from complete sessions and acknowledges that observed traffic can include evaluations. Its findings are evidence about one provider's activity, not a count of autonomous enterprises. [Anthropic research](https://www.anthropic.com/research/measuring-agent-autonomy).\n\nInteractive benchmarks offer another lesson. Tau-bench tests repeated task completion and notes that a correct final database state can still conceal a policy violation. Success therefore needs more than one indicator. [Tau-bench, version 1](https://arxiv.org/html/2406.12045v1).\n\n## Exhibit 06 — Put scrutiny at the point of consequence\n\n| Action | Evidence to request |\n|---|---|\n| Read or retrieve | Allowed sources, access boundaries and stale-data handling |\n| Recommend | Acceptance criteria, reviewer competence and rejection records |\n| Change a record | Authorisation, validation and an attributable change log |\n| Commit resources | Limits, approvals, reconciliation and exception ownership |\n| Affect a person or physical process | Domain-specific controls, intervention route and tested recovery |\n\n*Illustrative action categories; greater autonomy is not automatically a better operating design.*\n\nAsk the process owner to walk through one successful case and one failure. Where did the system stop? Who noticed? Could the affected state be restored? Did the team repair only the immediate result, or change the process that produced it?\n\nEvaluation itself needs inspection. The SABER authors identify annotation problems in an established agent benchmark. Their results table also includes cases where the proposed safeguard lowers performance, despite broader positive wording in the narrative. The lesson for Mayden is to retain exceptions and check the underlying comparison. [SABER, version 1, sections 5–8](https://arxiv.org/html/2512.07850v1).\n\n## 5. Workforce evidence should describe changed work\n\nTraining participation is an input. It does not establish whether someone can judge a difficult output, handle an exception or redesign a process. A workforce assessment should distinguish access, practice, competence and responsibility.\n\nA useful interview asks an operator to describe the last task completed with AI, show how the output was checked and explain what happens when the answer is uncertain. A manager should be able to describe how roles or workload changed. A learning lead should be able to show how the training responds to observed mistakes.\n\nThe OECD's original employer and worker surveys demonstrate the value of sector-specific wording and more than one organisational perspective. They also predate mass generative-AI adoption, so their reported attitudes are historical evidence rather than current regional norms. [OECD workplace study](https://doi.org/10.1787/ea0a0fe1-en).\n\nFor Mayden's future GCC fieldwork, bilingual interviews should test whether terms such as deployment, approval and active use mean the same thing to respondents. The English label alone is not a stable measure. Interviewers should offer equivalent examples, allow “I do not know”, and record whether the respondent speaks for a team, site, business unit or group.\n\nThe report back to a company should identify a next action its managers can own: an exception-handling exercise, a better usage denominator, a revised approval route or a repeatable evaluation. That makes participation valuable before the broader benchmark has statistical norms.\n\n## 6. Read the public index as an evidence map\n\nMayden's public edition records the strength of located documentation across nine dimensions. The highest accepted anchor in each dimension contributes to a fixed-denominator score. Unknown evidence contributes no documented points, while remaining explicitly unknown. The result is a comparison of documented evidence in the research corpus.\n\nIt is not a percentage of enterprise transformation. A score twice as large does not mean twice the capability. A sophisticated private company may disclose little. A highly communicative group may provide detailed examples from a narrow operating area. Both possibilities matter when reading a ranking.\n\n## Exhibit 07 — Three kinds of coverage that must stay separate\n\n| Coverage | What it answers | What it cannot establish |\n|---|---|---|\n| Search coverage | Which routes, languages and document types were checked? | That all relevant information was found |\n| Evidence coverage | In how many dimensions was qualifying documentation located? | That the practices cover the whole organisation |\n| Operating coverage | What share of the eligible business uses the practice? | Its quality or causal business impact |\n\n*Mayden measurement distinction. A source can support one coverage statement without supporting the others.*\n\nReaders should be able to open a company result and follow the score to its contributing claim, source and review decision. If a source is a supplier case study, that origin should remain visible. Several pages repeating one announcement do not become several independent confirmations.\n\nSensitivity views help readers see dependence on choices: remove commitment-only evidence, cap detailed anchors, or vary one dimension's weight. These are alternative assumptions applied to the same corpus. They are not confidence intervals, and stability across them cannot validate the underlying construct.\n\n## 7. The next research step is a company-level learning loop\n\nThe public edition should lead into voluntary, evidence-backed company research. The two products should retain separate identities: public documentation provides an external view; participating organisations can supply operational evidence unavailable to an outside reader.\n\nStart the interview with one material workflow and follow it deeply. Then broaden to the organisational arrangements supporting it: sponsorship, data access, skills, decision rights, evaluation and improvement. Ask for examples before asking respondents to rate themselves.\n\nA useful company report has three parts. First, show the evidence supplied and its scope. Second, identify the important unanswered questions. Third, propose a small set of operating experiments with owners, measures and review dates. Keep sensitive operational information out of the public comparison unless the company has authorised its use.\n\n## Exhibit 08 — The executive challenge for the next ninety days\n\n| Period | Decision to make | Concrete output |\n|---|---|---|\n| First month | Which workflow is important and measurable? | Scope, owner, baseline and eligible-work denominator |\n| Second month | What should change, and what could fail? | Bounded trial, authority map and recorded exceptions |\n| Third month | What evidence justifies continuation or scale? | Outcome account, deployment decision and next evaluation |\n\n*Proposed operating agenda, not a claim about completed Mayden fieldwork.*\n\nA credible AI-native benchmark will be built through repeated observation, challenge and correction. Its value to an executive is the quality of the questions it makes unavoidable: what changed, where it changed, who remains accountable, and how the business knows the change worked.\n\n### Research note\n\nThis report is an original Mayden interpretation informed by a scoped, AI-assisted evidence review. It combines distinct study designs to develop questions, not to pool effect sizes. International experiments, surveys, provider telemetry and simulations are labelled at their point of use. They do not supply GCC adoption estimates or validate Mayden's index. Company comparisons belong to the accompanying frozen release and retain their documented scope, eligibility and source limitations.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/state-of-ai-native-enterprise.pdf", "cover": "/research/index/2026-10/state-of-ai-native-enterprise.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 78403, "pdfSha256": "db94d6c0e02f1ec9dd09fc5100e367c0a671c3a05f2661efcb7e2ee195e756db", "pdfPages": 15 }, { "slug": "research-depth", "number": 11, "title": "How much should you trust an AI index?", "shortTitle": "How much should you trust an AI index?", "dek": "Evidence quality, disclosure bias and what changes when scoring assumptions change.", "sector": null, "kind": "Measurement research", "edition": "PE-2026.10-v4", "date": "2026-10-10", "executive": [ "59 of 80 selected entity records contain qualifying AI operating evidence in this review. 21 remain unranked because no qualifying claim was located; this does not establish absent capability.", "The corpus contains 272 recorded claims, of which 224 meet the declared eligibility rules. Specific commitments, deployed practices and operating detail are coded separately.", "Decision Lineage appears in 4 of these 80 profiles, compared with 44 for AI Core. These are counts within the selected research corpus, not enterprise adoption rates." ], "executiveSources": [], "evidenceAgeContext": "Dated sources run from 2019-05-14 to 2026-10-01. 79 of 224 eligible claim records have undated sources. Publication dates do not confirm continuing operation or a common observation period; undated does not mean stale.", "historicalNotice": "Two frozen comparisons: the historical chapter uses 20 records from PE-2026.10-v1; the current-corpus chapter uses 80 records from PE-2026.10-v3. The accompanying charts and metrics use PE-2026.10-v4 and its 80 selected entity records. The editorial revision retains the v3 coding. These are different research snapshots, not a time series of company performance.", "narrativeExhibits": [], "exhibitRegistry": [ { "id": "E01", "displayNumber": 1, "title": "Where evidence is visible across the eight sector cohorts", "kind": "generated" }, { "id": "E02", "displayNumber": 2, "title": "The evidence record includes what did not qualify", "kind": "generated" }, { "id": "E03", "displayNumber": 3, "title": "Operating detail deserves a separate test", "kind": "generated" }, { "id": "E04", "displayNumber": 4, "title": "Build a benchmark from evidence to a decision", "kind": "generated" } ], "exhibitOrder": [ "E01", "E02", "E03", "E04" ], "metrics": [ { "value": 80, "label": "selected entity records" }, { "value": 272, "label": "recorded claims" }, { "value": 224, "label": "eligible claims" }, { "value": 21, "label": "unranked records" } ], "exhibits": [ { "id": "E01", "type": "coverage", "title": "Where evidence is visible across the eight sector cohorts", "subtitle": "Number of selected entity records with a qualifying claim in each theme; cohort size shown per row", "columns": [ "AI Core", "Augmented Workforce", "Ecosystem", "Interfaces & Automation", "Decision Lineage", "Experimentation", "Autonomy & Operating Model", "Knowledge", "Governance" ], "rows": [ { "label": "Aviation & Logistics", "n": 10, "values": [ 7, 3, 5, 3, 1, 2, 0, 2, 2 ] }, { "label": "Banking", "n": 10, "values": [ 7, 8, 4, 7, 2, 3, 2, 4, 7 ] }, { "label": "Energy & Utilities", "n": 10, "values": [ 7, 6, 5, 3, 0, 3, 1, 3, 1 ] }, { "label": "Healthcare", "n": 10, "values": [ 6, 3, 3, 4, 0, 3, 0, 1, 4 ] }, { "label": "Manufacturing & Materials", "n": 10, "values": [ 5, 2, 2, 1, 0, 3, 0, 2, 2 ] }, { "label": "Real Estate & Construction", "n": 10, "values": [ 4, 1, 4, 3, 0, 3, 1, 1, 0 ] }, { "label": "Retail & Consumer", "n": 10, "values": [ 3, 1, 3, 5, 0, 0, 1, 1, 2 ] }, { "label": "Telecommunications", "n": 10, "values": [ 5, 5, 3, 4, 1, 2, 3, 3, 3 ] } ], "note": "Source: frozen Mayden public-evidence corpus. Counts describe located evidence, not comparable sector maturity or prevalence. Cohorts are purposeful, heterogeneous and sometimes overlap.", "displayNumber": 1 }, { "id": "E02", "type": "funnel", "title": "The evidence record includes what did not qualify", "subtitle": "Recorded claims pass explicit access, scope, deployment and review gates", "maximum": 272, "rows": [ { "label": "All recorded claims", "value": 272 }, { "label": "Eligible for scoring", "value": 224 }, { "label": "Excluded / held", "value": 48 } ], "note": "Source: release claim ledger. These counts do not measure research recall, factual accuracy or company capability. All exclusions remain inspectable in the public dataset.", "displayNumber": 2 }, { "id": "E03", "type": "stages", "title": "Operating detail deserves a separate test", "subtitle": "Eligible claim records, n = 224; disclosure anchors are not maturity stages", "maximum": 97, "rows": [ { "label": "Specific commitment / pilot", "value": 53 }, { "label": "Implemented practice", "value": 74 }, { "label": "Implemented + operating detail", "value": 97 } ], "note": "Source: reviewed coding. Detail is not independent assurance, and multiple claim records may describe a shared underlying practice.", "displayNumber": 3 }, { "id": "E04", "type": "process", "title": "Build a benchmark from evidence to a decision", "subtitle": "Mayden research architecture; conceptual framework, not an observed causal model", "rows": [ { "label": "Define", "detail": "Entity, workflow, population and period." }, { "label": "Observe", "detail": "Sources, tasks, usage and operating boundaries." }, { "label": "Challenge", "detail": "Alternative explanations, controls and exclusions." }, { "label": "Calculate", "detail": "Declared rules, sensitivity and reproducibility." }, { "label": "Decide", "detail": "Specific action, evidence gap and next test." } ], "note": "Source: Mayden methodological synthesis. Confidential company research remains a separate instrument requiring permission, respondent testing and validation.", "displayNumber": 4 } ], "cases": [], "agenda": [ { "question": "Choose one recurring workflow.", "evidence": "Its users, decisions, frequency, business importance and operating boundary." }, { "question": "Make authority and evidence inspectable.", "evidence": "Input provenance, evaluations, interventions, accountable owner and recovery path." }, { "question": "Track value with a credible comparator.", "evidence": "Quality, cost, risk and adoption denominators over a stated period." } ], "sources": [ { "title": "Conventional Wisdom on Measurement: A Structural Equation Perspective", "publisher": "Psychological Bulletin", "date": "1991-09", "url": "https://doi.org/10.1037/0033-2909.110.2.305", "locator": "Conceptual measurement-model analysis", "relationship": "peer-reviewed-methodological", "id": "S01" }, { "title": "Weights and importance in composite indicators: Closing the gap", "publisher": "Ecological Indicators / European Commission JRC", "date": "2017-09", "url": "https://doi.org/10.1016/j.ecolind.2017.03.056", "locator": "Methodological analysis with three case studies", "relationship": "peer-reviewed-methodological", "id": "S02" }, { "title": "PRISMA-S: an extension to the PRISMA Statement for Reporting Literature Searches in Systematic Reviews", "publisher": "Systematic Reviews", "date": "2021-01-26", "url": "https://doi.org/10.1186/s13643-020-01542-z", "locator": "Reporting-guideline development", "relationship": "reporting-guideline", "id": "S03" }, { "title": "PROV-DM: The PROV Data Model", "publisher": "W3C", "date": "2013-04-30", "url": "https://www.w3.org/TR/prov-dm/", "locator": "Normative data-model specification", "relationship": "technical-standard", "id": "S04" }, { "title": "How Organizations Use AI: Evidence from ChatGPT", "publisher": "arXiv author preprint", "date": "2026-08-12", "url": "https://arxiv.org/abs/2608.12236v1", "locator": "Observational linked enterprise-account and usage analysis", "relationship": "empirical-preprint-observational", "id": "S05" }, { "title": "Generative AI and Sales Productivity: Field Experiments in Online Retail", "publisher": "arXiv author preprint", "date": "2026-06-29", "url": "https://arxiv.org/abs/2510.12049v6", "locator": "Series of randomized field experiments", "relationship": "empirical-preprint-experimental", "id": "S06" }, { "title": "COSMIN methodology for assessing the content validity of PROMs: User manual", "publisher": "COSMIN", "date": "2018-02", "url": "https://www.cosmin.nl/wp-content/uploads/COSMIN-methodology-for-content-validity-user-manual-v1.pdf", "locator": "Methodological manual", "relationship": "methodological-guideline", "id": "S07" }, { "title": "Large Firms With at Least 20 Employees Biggest AI Users", "publisher": "U.S. Census Bureau", "date": "2026-05-26", "url": "https://www.census.gov/library/stories/2026/05/ai-use-businesses.html", "locator": "Original statistical-agency survey analysis", "relationship": "original-survey-analysis", "id": "S08" }, { "title": "The PRISMA 2020 statement: an updated guideline for reporting systematic reviews", "publisher": "The BMJ", "date": "2021-03-29", "url": "https://doi.org/10.1136/bmj.n71", "locator": "Reporting-guideline development", "relationship": "reporting-guideline", "id": "S09" } ], "body": "# What would make an AI-native benchmark scientifically credible?\n\nResearch development note · 10 October 2026\n\nMayden's ambition is to establish a durable benchmark for how enterprises turn AI into useful, accountable work. The current public corpus contains eighty selected records across eight sectors and five headquarters markets: Saudi Arabia, UAE, Qatar, Kuwait and Bahrain. The frozen first edition (PE-2026.10-v1), with twenty banking and telecom records, remains a historical worked example below. It measures located public evidence. The next stage must establish what the evidence can tell us about an organisation, where it remains incomplete and how a company could demonstrate progress.\n\nA useful benchmark gives executives better questions before it gives them a position. Which workflows have changed? Who uses the systems? What happens when they fail? What decisions may they execute? What evidence shows that the change helped? Public announcements can suggest answers. They rarely supply the complete operating record.\n\n## Start with the object being measured\n\nThree objects need separate treatment: an enterprise's actual operating capability, what its people report about that capability, and what outsiders can locate in public documents. They are related but not interchangeable. A quiet company can have strong capabilities. An articulate company can describe a narrow deployment in impressive detail.\n\nThe public index is a designed composite of disclosed evidence. It is not yet a validated psychological or organisational scale. Measurement research distinguishes indicators that define a construct from indicators that reflect an underlying trait; the appropriate tests depend on that choice. High internal consistency would not, by itself, establish that the nine themes measure AI-native enterprise capability. [Bollen and Lennox](https://doi.org/10.1037/0033-2909.110.2.305)\n\nThe programme therefore separates two linked instruments. The public edition maps documented practices and their evidential limits. A confidential company study investigates operating capability using interviews, artefacts, multiple respondents and, where appropriate, usage records. Neither instrument should silently stand in for the other.\n\n## Keep the number interpretable\n\nThe current public calculation takes the highest eligible evidence anchor in each of nine themes and divides the documented points by 27. Expressing the result on a 0–100 scale makes the arithmetic readable. It does not turn it into a percentage of AI nativeness.\n\nFixed anchors have advantages: another company's new announcement does not mechanically rescale everybody else, and repeated announcements within a theme do not add points. But a maximum has limitations. One detailed practice can fill a theme, even when enterprise-wide coverage is unknown. The arithmetic also treats the difference between a commitment and an implementation as equal to the difference between an implementation and a more detailed implementation account. That is a design assumption.\n\n**Historical worked example: PE-2026.10-v1, twenty records.** A frozen-data sensitivity exercise illustrates the consequence. Across the twenty frozen v1 records, removing commitment-only points while retaining the denominator changes five sector positions. Capping the higher implementation anchor at two points, with the fixed denominator, changes nine. Eighteen checks that individually halve or double one theme's weight change between one and eight positions each. These are deterministic alternative assumptions, including changes in ties. They are not confidence intervals or estimates of the probability that a company belongs in a position.\n\n**Current-corpus comparison: frozen PE-2026.10-v3, eighty records.** Repeating the same deterministic checks changes 16 sector positions when level-1 points are omitted and 20 when detailed anchors are capped at two. The eighteen individual weight changes alter between 1 and 23 positions each. This calculation includes changed ties and unranked states; it compares each scenario with its own frozen baseline. These counts describe method dependence in the expanded corpus, not company improvement since v1. The fourth-edition editorial revision retains the company coding and baseline scores.\n\nThe practical implication is to foreground the nine-theme profile, raw points and evidence coverage, with positions as a secondary view. Methodology research also shows that nominal weights and statistical importance need not coincide when indicators vary and correlate. Equal weights are a transparent starting choice, not a discovery that each theme contributes equally to performance. [Becker and colleagues](https://doi.org/10.1016/j.ecolind.2017.03.056)\n\n## Separate three kinds of coverage\n\nAn executive should be able to distinguish research coverage, evidence coverage and operational coverage.\n\nResearch coverage describes work completed against a declared search plan. It includes source types checked, languages searched, reporting periods considered and documents retrieved. Evidence coverage describes how many themes have qualifying claims. Operational coverage describes the proportion of relevant people, workflows, transactions or decisions affected by a practice.\n\nOnly the last speaks directly to penetration inside the firm. A report discussing all nine themes does not prove enterprise-wide use. Conversely, an inaccessible annual report is a research limitation, not evidence that an operating practice is absent.\n\nThe public research-input ledger publishes recorded search routes and access gaps. A future standardised search matrix should add planned-versus-completed routes alongside each company profile. The matrix would make failed retrievals, unresolved scope questions and under-searched themes visible. It would report completion against the plan without calling that an estimate of the percentage of all relevant evidence found. PRISMA-S offers useful reporting disciplines for documenting searches; adopting those disciplines does not make this corporate review an exhaustive systematic review. [PRISMA-S](https://doi.org/10.1186/s13643-020-01542-z)\n\n## Trace every conclusion to its support\n\nA credible claim needs more than a link. Readers should see the relevant page or passage, the corporate entity it concerns, the operating period, whether the practice is a plan or in use, and what remains unestablished. A group announcement should not automatically become evidence for every subsidiary. A vendor's customer story remains a commercially connected account. A mirrored annual report is another route to the same document, not another independent observation.\n\nThe evidence ledger records these distinctions and links each transformation: fetched source, extracted passage, atomic claim, review decision and published contribution. W3C's provenance model provides a useful vocabulary for entities, activities, agents and derivations. Traceability helps assessment; it does not certify truth. [W3C PROV-DM](https://www.w3.org/TR/prov-dm/)\n\nThe public interface exposes contributing claims beside theme results. That view should include qualifying, contradictory and excluded material, with the reason for each decision. “Company-reported” and “independently corroborated” should be explicit labels, never an unexplained confidence percentage.\n\n## Measure work, people and boundaries together\n\nRecent empirical research supports studying AI at the level of tasks, roles and operating conditions. It does not establish one universal enterprise productivity multiplier.\n\nA 2026 observational preprint links ChatGPT Enterprise account and usage records with worker roles and company information. Its author abstract describes differences in adoption speed, breadth and purpose. The selected customers of one product cannot represent all enterprise AI, and observed usage does not demonstrate causal business impact. The study nevertheless motivates a useful distinction between available accounts, active use, tasks performed and roles reached. [Chatterji and colleagues, version 1](https://arxiv.org/abs/2608.12236v1)\n\nAnother preprint describes randomized field experiments across seven online-retail workflows, with effects varying by application, including some with no effect. Its author abstract supports a workflow-specific evaluation question, not a transferable enterprise-wide return estimate. The inspected version concerns one platform and sales outcomes. [Fang and colleagues, version 6](https://arxiv.org/abs/2510.12049v6)\n\nFor Mayden, the corresponding interview should ask for a concrete example, a relevant denominator and an observable boundary. “We rolled out AI” becomes: which process, to whom, over what period, how frequently, with which review or fallback, compared with what baseline? The answers should remain separate from unrelated marketing claims or future forecasts.\n\n## Test the questionnaire before making it authoritative\n\nA company instrument must be understandable to the people answering it. That requires more than expert approval of attractive terminology. An executive, an operations manager and a technology owner may interpret “autonomy”, “deployment” or “governance” differently.\n\nContent-validity guidance in health measurement distinguishes relevance, comprehensiveness and comprehensibility, including direct testing with intended respondents. These principles can inform organisational questionnaire development, while the clinical thresholds and grading rules require adaptation. [COSMIN content-validity manual](https://www.cosmin.nl/wp-content/uploads/COSMIN-methodology-for-content-validity-user-manual-v1.pdf)\n\nThe next company pilot should begin with English and Arabic cognitive interviews in the UAE and Saudi Arabia, across more than one sector and respondent role. Participants should explain how they interpret each question and how they chose an answer. Unknown, no and not applicable need distinct response options. A limited feasibility pilot can reveal misunderstood questions, onerous evidence requests and missing topics. It cannot establish population norms.\n\nSubsequent validation should examine respondent agreement, stability over an appropriate interval, relationships with independently measured operating practices and differences across language, sector and company scope. The required sample depends on those analyses and the desired precision. There is no defensible universal pilot size that makes the instrument scientific.\n\n## Preserve comparability as AI changes\n\nA durable index needs stable definitions and an honest account of revisions. The U.S. Census Bureau's 2026 business-AI analysis explains that its survey question broadened in November 2025 from production-related use to use in any business function. Such changes demonstrate why a trend can reflect question design as well as business behaviour. U.S. survey estimates are not GCC adoption norms. [U.S. Census Bureau](https://www.census.gov/library/stories/2026/05/ai-use-businesses.html)\n\nThe programme needs a stable core and version emerging-practice modules separately. Each update should distinguish newly located evidence, corrected coding, changed organisational scope, revised methodology and documented operational change. Historical releases should remain reproducible. A better search next year must not be described as proof that the company improved.\n\nThis development note is a scoped methodological review with selected empirical examples. It is not a systematic literature review, new meta-analysis or validation of Mayden's instrument. Reporting guidance such as PRISMA provides useful transparency standards; using the name is not a substitute for executing and documenting the corresponding research design. [PRISMA 2020](https://doi.org/10.1136/bmj.n71)\n\nThe near-term publication can be valuable on these terms: a transparent account of what selected companies disclose, where evidence is strong enough to inspect, and which important questions remain unanswered. A stronger enterprise benchmark will follow from the quality of those questions, the evidence collected against them and the discipline with which Mayden separates observation from inference.\n\n**Review transparency.** Extraction and source challenge in this edition were AI-assisted. Separate AI readers do not constitute human expert sign-off or independent assurance. Company claims, research findings and Mayden interpretations retain their distinct scope.", "pdf": "/research/index/2026-10/research-depth.pdf", "cover": "/research/index/2026-10/research-depth.svg", "basis": "Public-evidence review. Selected entity scopes. No company participation, population inference or independent assurance implied.", "bytes": 67160, "pdfSha256": "d67182f16aba7316606608952ca474194a5d845df85a180858aeb11741f5cf9e", "pdfPages": 12 } ] }