Use case 1
AI Agent Action Governance
Let agents assist without giving them uncontrolled authority.
AI agents can move from reading to drafting, recommending, requesting approval, and executing actions. Each step may require different authority.
- Example requester
- An enterprise AI agent assisting a service team.
- Example request
- Access selected customer information and send an external message.
- ConsentKey evaluation
- ConsentKey can evaluate identity, purpose, data sensitivity, disclosure rules, action type, approval requirements, and time limits.
- Possible decision
- Partially approved with human review before external sending.
- Minimum permitted result
- Selected fields released for drafting only, with no external disclosure until a human approves.
- Decision receipt
- Requester, purpose, permitted fields, denied external send, reviewer requirement, time limit, and decision timestamp.
- Organizational value
- Helps organizations let AI agents assist while keeping delegated authority explicit, scoped, and reviewable.
- Public-safe mock scenario
- A support agent drafts a response using only account status and recent case context. The message remains a draft until a manager approves external delivery.
