Design approvals and escalation payloads
Course overview · 4 min reading + 12 min practice, estimated
Principles and method
A human escalation should explain the task, evidence, uncertainty, proposed action and consequences. The reviewer must see enough context to decide without reconstructing the entire run. Approval should bind to a specific action and payload, with current-state validation immediately before execution. If the material content changes, seek a new decision. Define timeout and no-response behaviour. A safe default may be to pause rather than infer consent. Avoid overwhelming reviewers with low-value requests that encourage automatic approval of consequential actions.
Worked example
An agent drafts a candidate message. The review shows recipient, content, source facts and why contact is appropriate. A changed recipient or material message edit invalidates the earlier approval.
Put it into practice
Design an escalation card and three outcomes: approve, revise and stop.
Use fictional information and keep your work in your own notes.
Compare your approach: self-review guidance
Include what the agent does after no response and how it records the decision. A reviewer should be able to reject the action without breaking the rest of the workflow.
Sources and further reading
Original Academy teaching and fictional examples. These references provide context, not endorsement. Edition 2026.09; updated 2026-09-24.
- OWASP: Top 10 for large language model applications
Security background for applications that use language models.
- NIST: AI Risk Management Framework
Voluntary framework for organising AI risks and controls.
- GOV.UK: Responsible AI in recruitment
UK guidance on procuring and deploying recruitment AI.