AI escalation and emergency matrix
Make the boundary between routine assistance, human escalation and an emergency unmistakable.
Done when: Each trigger has an approved response, reachable human route and tested unavailable-person fallback.
Enable JavaScript for on-page entry and smart printing, or print a blank copy from your browser and complete it offline.
Routine requests the agent may handle
List bounded business requests and the approved facts or tools it may use. Exclude patient-data services.
Human-escalation triggers
Specify complaints, uncertain answers, payment disputes, unusual requests and other triggers requiring a person.
Urgent or emergency wording
Record owner-approved stop wording and appropriate emergency-service direction. The agent is not an emergency responder or clinical triage service.
Destination and coverage roles
Name the role, coverage hours and controlled directory location. Do not put private phone numbers or credentials here.
If nobody answers
Define what the caller hears, what non-sensitive message may be retained and who reviews it. Never promise a callback time that is not staffed.
Test and release decision
Use fabricated scenarios to test normal transfer, unavailable destination and ambiguous urgency. Record expected versus observed behavior and the approval decision.
Reference links and review boundary
Use the current issuer, payer, regulator, platform or qualified professional for your exact activity and location. These links informed the planning questions; they do not make this an official form, full compliance review or permission to operate.
Healthcare businesses are welcome for business education. PublicProof does not receive, store or process patient information. A completed worksheet does not establish HIPAA compliance, transport qualification, professional authority, account approval, a credential, reimbursement or revenue.
Worksheet updated October 20, 2026. Verify changing requirements before relying on them.