[ operations / reviews ]· synthetic demo data
Human review.
Accountability stays human.
Runs paused at a runtime risk gate wait here. Every decision is appended to the review history; nothing is overwritten.
[ decision model ]
Approve
→ Finalize the persisted validated candidate
Reject
→ Halt without releasing a recommendation
Request revision
→ Keep history, increment revision, return through the controlled revision path, require a new decision
[ queue · 1 pending ]
[ risk gate · policy CTRL-017 · revision 0 ]
Release AI-assisted review recommendation for the compliance workflow
Reason
The recommendation affects a regulated customer review process and may influence an escalation decision.
Flags
CTRL-017 four-eyesCustomer-outcome impactEscalation influence
AI recommendation (held)
Adopt AI-assisted policy retrieval and discrepancy flagging. Analysts and a second reviewer keep full authority over escalation decisions.
[ review history · audit trail ]
- risk_gateRisk gate triggeredCTRL-017 four-eyes · Customer-outcome impact · Escalation influence · policy CTRL-017
- interruptExecution paused → pending_reviewLangGraph interrupt · evaluation and release wait for a human decision