This scorecard turns AI programs into one executive surface: policy-bundle posture, privileged tool scope, evaluation coverage, spend and vendor controls, data-boundary clarity, and board-brief freshness.
| Lane | Owner | Status | Related findings | Focus | Next action |
|---|---|---|---|---|---|
| Policy bundle lane Policy approval has not caught up to production agent behavior. | AI Governance Council | red | 1 | Decision cards, approval posture, and board-safe control mapping | Freeze new production escalations until the policy bundle is approved. |
| Identity and tool-scope lane Tool scope is still too broad for board-ready confidence. | Identity Engineering | red | 1 | Privileged tools, reversible actions, and write-scope discipline | Reduce write scopes and add a human-review gate for CRM mutations. |
| Evaluation and runtime lane Evaluation coverage is visible but below the launch threshold. | CX Platform | yellow | 2 | Escalation evals, refusal coverage, and runtime regression proof | Expand eval coverage before the next customer-facing rollout. |
| Board brief lane The board memo is not current enough to support the next steering review. | CTO Staff | yellow | 2 | Executive narrative, savings posture, and board-pack freshness | Refresh the board brief with current score, savings, and exposure language. |