Approvals
High-impact actions remain human-owned and approval-aware.
Governance
Shared platform controls apply across the Team. Staff role pages publish role-specific approvals and boundaries.
Approval gate walkthrough
The control model is visible: proposed action, risk, evidence, human decision, and read-only Trace & Evidence.
Approval required
External message delivery is blocked until the human owner decides.
Evidence packet
Trace rule
Trace & Evidence explains the recorded decision path. It does not execute the send again.
Recorded timeline
High-impact actions remain human-owned and approval-aware.
Access follows role and workspace boundaries.
Actions, decisions, and visible results remain reviewable.
Recorded history supports investigation without re-execution.
Human owners can pause or stop runtime activity.
Staff roles, evidence requirements, and approval gates reduce unreviewed AI output risk.
Governance evidence
These screenshots and diagrams support control, traceability, and human-review claims.
A captured product evidence flow showing how a request becomes a reviewable operating output.
A natural-language request routed to Recovery AI Staff with a reviewable response and action context.
A trace view connecting the request, Staff identity, intent, outcome, and recorded evidence.
Security review notes
Public governance materials focus on approval gates, RBAC, workspace isolation, Trace & Evidence, recorded evidence, and stop controls. Formal legal or certification commitments belong in reviewed procurement materials.
Approval gates
RBAC
Workspace isolation
Audit evidence
Trace & Evidence
Stop controls
Integration permission preview