Leave the Committee meeting knowing what can be relied on, what still needs proof and what must happen next.
A full-year Committee operating system for AI value, trust, visibility, human accountability, system assurance and decision closure.
Leave the Committee meeting knowing whether AI is creating enough value, remaining trustworthy and staying within human control.
The Pack expands the open 10-question Challenge Card into a full-year 50-question operating discipline. The updated bank is organised across ten Board lenses and links every challenge question to expected evidence, warning signs, ownership, decision state and a trigger for reconsideration.
The Committee test: Are we clear on the value being pursued, what AI is actually in scope, what evidence supports trust, where human accountability sits, and what decision or action cannot wait?
Stop treating adoption as proof of value.
See the business outcome, success measure, supplier contribution, scale/stop threshold and value assumption still unproven.
Stop relying on a self-declared inventory.
Reconcile management's register to procurement, SaaS, cloud, API, browser/security and supplier evidence.
Stop accepting 'human in the loop' as an answer.
See who can understand, challenge, override and stop the AI, and whether they have enough time and competence to do it.
Stop using governance maturity as a proxy for system behaviour.
Ask for the model, GenAI, agent or supplier evidence appropriate to the actual technology and decision.
Stop letting approval become permanent by default.
Define the drift, incident, provider, regulatory, data or autonomy change that reopens the position.
Stop governing everyone except the governing body.
Set rules for director AI literacy, Board-paper AI use, confidential information, verification and Committee allocation.
The operating discipline you already built.
- Committee mandate, reserved matters and decision rights
- 12-month oversight calendar and agenda
- 50-question recurring challenge bank
- Evidence catalogue and assurance map
- Decision dashboard and escalation thresholds
- Decision record, challenge log and action tracker
- 90-day implementation plan
The evidence and Board questions the old Pack could not see clearly enough.
- AI Value & Trust Scorecard
- Human-AI Decision Boundary Map
- AI Inventory Completeness Test
- AI System Assurance Evidence Standard
- ML, GenAI, agentic and embedded-AI overlays
- Continuous assurance trigger library
- Board AI Operating Protocol
- Workforce and competence evidence lane
One governance question does not fit every kind of AI.
Performance, fairness, robustness and drift.
Ask for validation evidence, affected-population performance, sensitivity/robustness results and monitoring thresholds.
Grounding, hallucination, leakage and user reliance.
Ask for use-case evaluations, prompt/grounding controls, abuse testing, data leakage evidence and human verification.
Identity, permissions, action limits and stop rights.
Ask what the agent can do, which tools it can invoke, where approval gates sit, how actions are logged and who can stop it.
Supplier change, concentration, rights and exit.
Ask what capability changed, what data/model dependency exists, whether notification rights work and what a credible substitute looks like.
Use the upgraded working tool in the next Committee cycle.
The accompanying browser tool contains the ten-lens 50-question bank, evidence prompts, technology overlays, continuous triggers, decision records and local export.