Skip to content

1. Executive Summary Top-level signals and decisions for the Board

Headlines
  • S-01 (Credit decision assistant) remained stable; no material deterioration in key risk metrics after last-quarter parameter adjustments.
  • S-03 (Vendor risk screening) is ready for controlled production launch, subject to board approval of conditions in Section 7.
  • One medium-severity near miss (I-01) was intercepted by human review; guardrails and reviewer guidance were strengthened.
Board decisions requested
  • Approve deployment of S-03 into production with safeguards and monitoring thresholds.
  • Confirm decommissioning of legacy classifier S-07 and retention of logs per retention standard.
Decision items are listed in Section 7 with recommended actions and deadlines.

2. AI System Inventory Snapshot Material AI systems in scope for the period

Inventory (illustrative) — systems classified as “material” under internal standard AIG-01
ID System / Use Case Business Owner Model / Provider Infrastructure Data Sensitivity Risk Tier Status
S-01 Credit decision assistant — retail limits CRO Internal model + frontier API (guardrailed) GPU-backed cloud High Tier 2 Production
S-02 Customer support triage — inbound messages COO Hosted LLM + internal guardrails Cloud CPU Medium Tier 3 Pilot
S-03 Vendor risk screening — high-risk vendors CRO Composite (internal + third-party) GPU-backed cloud Medium Tier 2 Pre-production
Scope note: In a live ledger, each system entry links to a system profile (purpose, data boundary, owners, controls, test results, evidence IDs).

3. Significant Changes This Period New, changed, and decommissioned systems

Changes (illustrative) — each entry would carry evidence references and approvals
Change Type System ID Description Rationale Governance Actions Effective Date
New S-03 AI-based vendor screening for high-risk vendors. Reduce manual review time, increase consistency, improve coverage. AI Governance Committee review; recommended for Board approval with monitoring conditions. 2026-01-15
Parameter change S-01 Max auto-approved credit limit increase adjusted from +10% to +15% for low-risk customers. Align with peer practice while staying within board-approved appetite. Impact analysis reviewed; change endorsed subject to weekly drift and delinquency monitoring. 2026-01-28
Decommissioned S-07 Legacy NLP classifier in complaints routing decommissioned. Consolidated into S-02 with stronger monitoring and vendor support. Decommission plan approved; logs archived per retention standard (illustrative: 7 years). 2026-02-03

4. Incidents, Exceptions, and Near Misses Material AI-related events during the period

Events (illustrative) — near misses remain board-visible for learning and control hardening
Ref Date System ID Severity Description Root Cause Actions Taken Status
I-01 2026-01-07 S-01 Medium Suggested credit-term adjustment would have breached internal affordability guidance; intercepted by human review before customer impact. Prompt/configuration gap; missing explicit affordability constraint. Guardrail added; prompts updated; reviewer guidance tightened; incident template added to playbook. Closed
Regulatory note (illustrative): No events met external notification thresholds this period under current guidance.

5. Controls & Testing Key controls, owners, cadence, and results

Controls (illustrative) — mapped to internal standard AIG-01 and sector guidance
Control ID Control Description Systems in Scope Owner Testing Frequency Last Test Date Result
C-01 Human-in-the-loop review for all high-impact credit decisions. S-01 Risk Each batch 2026-02-20 Effective — no unreviewed high-impact decisions identified.
C-02 Pre-deployment configuration and prompt review for Tier 1–2 systems. S-01, S-03 AI Governance Per deployment 2026-02-05 Minor findings — documentation tightened; remediated within the period.
C-03 Quarterly performance and drift review with challenge from Risk. S-01 Model Risk Quarterly 2026-02-12 Effective — within tolerance; no recalibration required.
Control gap note (illustrative): Extend quarterly drift review to S-03 post go-live and formalize evidence capture requirements for C-02.

6. Provider & Infrastructure View How external providers and compute are used

Hosted / frontier services
  • Total API calls: 1.8M (illustrative) • +12% vs prior quarter
  • High-risk use cases: 2 (S-01, S-03)
  • Incidents: 1 (see I-01)
Contracting note (illustrative): data-use, retention, and audit clauses aligned to internal standard AIG-03; access is restricted and centrally logged.
GPU-backed workloads
  • GPU hours: 420 (illustrative) • +8% vs prior quarter
  • New models deployed: 1 (S-03)
  • Governance reviews: 3 completed pre go-live
Security posture (illustrative): restricted cloud environment with network segmentation, identity controls, and immutable logging.
Executive access & device posture (illustrative): board/exec access to AI governance artefacts is restricted to managed devices with enforced encryption and strong authentication; exceptions are time-bound and approved.

7. Board Actions & Approvals Decision items for board sign-off

Board actions (illustrative) — decisions can be recorded here or in board minutes referencing the item IDs
Item Description Request Recommended Decision Deadline
BA-01 Approve deployment of S-03 into production with monitoring thresholds and human-review steps (see Sections 3 and 5). Approve / Amend / Reject Approve with conditions — quarterly review of FP/FN rates; escalation triggers if thresholds breached. 2026-03-10
BA-02 Confirm decommissioning of S-07 and archival of logs and documentation for audit and supervisory review. Confirm Approve — no remaining dependency; replacement S-02 in place. 2026-03-10
Implementation note (live deployment): each board action links to an evidence bundle (approvals, test results, risk sign-offs) and records an “Owner / Due / Status” trail.

8. Annex (Illustrative) What a live Trust Ledger attaches or links

  • Tier 1–2 system profiles (scope, data boundary, safeguards, owners, KPIs, evidence IDs).
  • Full incident/near-miss reports (root cause, corrective actions, control updates, evidence).
  • Control mapping to standards and regulations (policy → control → test → evidence chain).
  • Glossary for non-technical directors (terms, thresholds, materiality rules).
This online sample is intentionally concise; production deployments can be configured for your governance model, audit requirements, and board cadence.