Approvals and audit
AI Approvals
Command Center → AI Approvals lists every action that needs a human:
- Approve — re-runs the action through all governance gates and executes it
- Reject — records the decision as blocked
AI Activity
Every run is explained in four parts: WHAT happened, WHY (the AI decision), GUARDRAIL (policy verdict), RESULT. Action-capable employees record an observation even when no rule matches, so nothing is silently dropped.
Governance chain
propose → schema → policy → authorize → execute → audit
No AI action bypasses policy or your audit trail. A per-conversation action cap (default 3) prevents loops, and outgoing messages never re-trigger employees.