Human Oversight Controls
You decide what the platform may do without you
Five operating modes. Eight category policies. Five escalation paths. All are auditable; all are reversible in a single click.
Governance modes
Select the operating posture. The platform refuses to exceed its current mode — by design.
Currently selected
Governed Execution — AI auto-runs approved categories. All actions exceeding this posture will be queued for human review and recorded in the ledger.
Category policies
Per-category execution rules. Edit policies; the platform recompiles its action graph instantly.
- POL-DOC-01documentationauto-execute · auto-deployrisk low
- POL-FE-04frontendauto-execute · canary 5%risk ≤ medium
- POL-BE-04backendauto-execute · canary 5%risk ≤ medium
- POL-INF-09infrastructure1 approver requiredrisk ≥ medium
- POL-SEC-12security2 approvers + security on-callrisk ≥ medium
- POL-DB-07databaseDBA approval · dry-run requiredrisk ≥ medium
- POL-AUTH-03authenticationsecurity on-call + change windowrisk ≥ medium
- POL-BIL-02billingbilling lead + finance + on-callrisk any
- POL-CMP-01compliancecompliance officer requiredrisk any
Escalation workflows
Where the platform routes when a decision exceeds its envelope.
- Trigger 1Critical risk actionagent → security on-call → CISO
- Trigger 2Validation failure on prod pathagent → eng oncall → engineering lead
- Trigger 3Cross-service blast radius > 3 servicesagent → platform oncall → CTO
- Trigger 4Billing or payments mutationagent → billing lead → finance → CFO
- Trigger 5Customer data schema changeagent → DPO → compliance officer