Governance & Trust

Every agent operates inside explicit guardrails. This page shows the control framework, autonomy rules per scenario, and the current implementation status of each policy.

Needs owner

controls

Design

controls

Implemented

controls

Testing

controls

Governance controls

10 controls across consent, brand safety, financial thresholds, and audit

ID	Control	Policy	Applies To	Level	Risk	Status
GOV-001	Customer identity verification	Support agent must verify email/order number before disclosing order details	A01	Mandatory	High	Needs owner
GOV-002	Refund and replacement approval	Agent may draft but not issue refunds above threshold without human approval	A01	Mandatory	Medium	Design
GOV-003	Consent gate	Marketing agent checks consent before segmentation or send	A03/A08	Mandatory	Low	Implemented
GOV-004	Wholesale price isolation	B2B account-specific pricing is only visible to authorized users	A04	Mandatory	Medium	Needs owner
GOV-005	Claims review	Ingredient, safety, sustainability, and performance claims need approved language	A03/A07/A08	Mandatory	Low	Design
GOV-006	AI answer citations	Customer-facing answers must be grounded in approved knowledge base or live data	All	Mandatory	High	Design
GOV-007	Inventory promo block	Campaign is paused if hero SKU drops below cover threshold	A03/A06	Recommended	Low	Testing
GOV-008	PII minimization	Logs should redact raw address/payment details	All	Mandatory	Medium	Implemented
GOV-009	Human override	Every autonomous action must include rollback/owner visibility	All	Mandatory	High	Testing
GOV-010	Audit trail	Every action records agent, data source, confidence, and tool call	All	Mandatory	High	Testing