The AI Escalation Receipt

Five fields in an escalation receipt

• Stop reason: Write the plain-language reason the AI output should not move forward automatically: missing evidence, sensitive data, unclear owner, or risky impact.
• Source evidence: List the exact notes, tickets, emails, screenshots, logs, or documents that are allowed to support the next decision.
• Decision owner: Name the person or role that must approve, reject, rewrite, or park the AI-assisted work before it affects anyone else.
• Safe next step: Choose one reversible action: ask for missing facts, create a human-review draft, update the checklist, or record “not enough information”.
• Audit trail: Save the receipt with date, task, source list, reviewer, decision, and unresolved questions so the team can inspect it later.

Five prompts before work resumes

• Why stop?: What made the AI-assisted task unsafe, uncertain, or incomplete?
• What facts?: Which exact sources may be used for the next review?
• Who owns it?: Who has authority to approve or reject the next action?
• What is safe?: Which reversible step can happen without external impact?
• What is logged?: What evidence proves the handoff was reviewed?

How to use this receipt

Use this receipt when an AI-assisted workflow reaches uncertainty: a missing fact, a sensitive decision, unclear ownership, or a public/customer-facing consequence. The goal is to pause safely, capture evidence, and route the next step to a responsible reviewer.