Record every instance where a human operator overrides, rejects, or challenges an AI output. Review monthly. If the override rate for any system exceeds [20%], escalate to the Executive Accountable Official as a potential systemic issue requiring investigation.
Override Log
| Log ID | Date | AI System | AI Output / Decision | Challenged By | Reason for Override | Human Reviewer | Override Decision | Final Outcome | Documented in Risk Register? |
| OVR-001 | [DD/MM/YYYY] | [System name] | [Brief description of the AI output being challenged — e.g. "Estimated labour cost: $X"] | [Name / Role] | [e.g. AI output inconsistent with site conditions known to estimator] | [Name / Role] | [Accepted override / Upheld AI output / Escalated] | [Final value or decision used] | [ Yes / No ] |
| | | | | | | | | | |
| | | | | | | | | | |
Monthly Summary
Month / Period
[Month YYYY]
Total AI outputs reviewed
[Number]
Total overrides
[Number] · Override rate: [X]%
Pattern or systemic issue identified?
[ None | Yes — describe and reference escalation action ]