The city’s change advisory board uses AI to flag change requests that omit a rollback plan before humans vote. What automation pattern is this?

AI-assisted change approvals that leave humans accountable fit how a municipal change advisory board (CAB) should use automation: the model flags change requests that omit mandatory artifacts such as a rollback plan, then board members still vote with that signal in view. The pattern works because AI is strong at checklist completeness and consistency checks across high-volume civic change tickets, while production risk acceptance—especially for resident-facing or OT-adjacent systems—remains a human governance decision. Flagging missing rollback documentation before the vote reduces rubber-stamp approvals without transferring authority to the model. AI-summarized change approvals that still leave humans accountable misses the question because it describes AI condensing the change request for faster reading, not flagging a missing rollback plan to inform the CAB vote, so it lacks the pre-vote completeness check for approval. AI-prioritized change approvals that still leave humans accountable are wrong because they focus on ordering or ranking changes by risk or urgency, not on detecting an omitted rollback plan before the board votes. AI-routed change approvals that still leave humans accountable fail because routing directs the request to the right approvers or workflow path, while this pattern requires flagging the missing rollback plan BEFORE humans vote. Exam caveat: an AI flag is a risk signal, not a veto or an approval; treating model silence as proof of safety is a common exam trap. Operational check: map required CAB fields (rollback, test evidence, impact, backout owner) into the change form; log every AI completeness finding with the human vote; and periodically audit auto-flag accuracy against changes that later needed emergency rollback.