# Coding calibration workbook
Version 0.1 | 10 October 2026 | Prepared examples, not empirical reliability results

These examples preserve PE-2026.10-v3 decisions. Positive means qualifying disclosure, not verified capability. A boundary example may remain eligible: it identifies the limit of the accepted interpretation. Every ID, URL, locator, rationale and claim fingerprint is retained in coding-calibration.json.

## ai-core
A named in-progress integration is not routine production. An inaccessible general ambition cannot qualify.

- Positive: LOG-001; retained anchor 2. Emirates Airport Services reports a launched AI system for aircraft-turnaround analysis and resource optimization.
- Boundary: LOG-021; retained anchor 1. Milaha reports beginning AI integration at its logistics centre in 2025 as it moves toward more automated warehousing.
- Excluded claim: LOG-037. 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.

## workforce
Onboarding or course participation supplies disclosure detail; it does not demonstrate active use or competence.

- Positive: LOG-004; retained anchor 3. More than 2,000 Group IT staff had joined its AI gateway by August 2026.
- Boundary: LOG-003; retained anchor 1. The Group announced enterprise ChatGPT access and AI-literacy programmes for employees.
- Excluded claim: LOG-024. Excluded: the retrieved passage does not establish AI-specific employee curriculum or completed participation. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained.

## ecosystem
Retain specific partner delivery commitments at 1; recover exact source content before admitting indexed-only evidence.

- Positive: LOG-019; retained anchor 2. Qatar Airways Cargo staff contributed to the supplier implementation and refinement of CARGOSTACK Optimiser.
- Boundary: LOG-013; retained anchor 1. Bahri and Aventra announced a maritime data-orchestration platform intended to support machine learning and AI.
- Excluded claim: bank-nbb-01. 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 independent source-review pass. Complete readable source support not established; indexed-only evidence cannot score.

## interfaces
Technical detail cannot promote a pilot to implementation; distinguish service workflow integration from a proposed customer channel.

- Positive: LOG-016; retained anchor 3. The AI-driven TOS+ digital twin at Jafza uses real-time data for terminal actions including gate opening and maintenance alerts.
- Boundary: BAPCOU02; retained anchor 1. The implemented model lets engineers test scenarios and choose network setpoints.
- Excluded claim: LOG-027. Excluded: proposed customer channels alone do not demonstrate internal AI operating capability. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained.

## lineage
A described logging architecture can qualify as a company assertion, but is not an inspected decision trace. No eligible level-1 lineage example exists in this frozen corpus.

- Positive: LOG-005; retained anchor 3. ARRIVE records intent, scope, risk, evidence and decisions in AI-assisted software delivery.
- Boundary: bank-emirates-nbd-03; retained anchor 3. Leap supplies agent audit logging, observability, prompt management and model-governance controls.
- Excluded claim: HLT-023. Infrastructure residency rationale does not establish data lineage or traceability. Preserve raw exclusion. Exclusion: New claim has no explicit accepted review decision. Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Excluded by independent source-review pass.

## experimentation
A trial can qualify at 1; described model iteration can qualify at 3 without proving independent validation or a causal effect. No excluded experimentation claim exists in this frozen corpus.

- Positive: LOG-018; retained anchor 3. The cargo team and Wiremind tested and iterated multiple machine-learning approaches during model validation.
- Boundary: LOG-002; retained anchor 1. Emirates Engineering is trialling AI and computer vision to identify engine parts.
- Excluded interpretation of LOG-018: independent causal validation of enterprise-wide performance. The released claim remains eligible; no excluded experimentation claim is invented.

## autonomy
A live decision mechanism must retain its stated authority boundary; projected workload reduction cannot establish deployed authority.

- Positive: bank-cbq-04; retained anchor 2. The investor presentation identifies AI-optimised credit-limit increases as live.
- Boundary: TEL-012; retained anchor 1. e& UAE and TM Forum published a roadmap towards Level 4/5 autonomous networks with human governance and policy controls.
- Excluded claim: HLT-015. Potential workload automation does not establish implemented decision authority. Preserve raw exclusion. Exclusion: New claim has no explicit accepted review decision. Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained. Excluded by independent source-review pass.

## knowledge
Named guidance under development is a commitment; network observability alone does not demonstrate organisational knowledge management.

- Positive: LOG-034; retained anchor 2. FalconEye combines flight, passenger, crew and other operating information for executive questions.
- Boundary: MAN-ALBA06; retained anchor 1. Alba describes a Code of Conduct chatbot designed to guide employees on conduct-policy questions.
- Excluded claim: TEL-035. Network observability does not establish organizational knowledge-management capability. Exclusion: No qualifying AI-specific operating anchor retained.

## governance
Applied AI-specific responsibility differs from an unfinished framework or ordinary security and record access.

- Positive: LOG-006; retained anchor 3. ARRIVE embeds governance and human accountability in Group IT AI-assisted engineering.
- Boundary: bank-al-rajhi-06; retained anchor 1. Al Rajhi states its formal AI governance framework remains under development.
- Excluded claim: LOG-010. Excluded: training-record access and ordinary system security do not establish AI-specific governance. Exclusion: Explicit extraction-stage exclusion retained. No qualifying AI-specific operating anchor retained.

## Training and evaluation must be separate
Use these disclosed answers to teach the codebook. Then assess reliability on a different, blinded sample selected before coding. Retain every disagreement and resolution. Reusing these worked answers as the reliability test would measure recall and agreement with the existing release, not independent reproducibility.