AI Receptionist QA Checklist

A practical launch checklist for teams deploying AI receptionists, phone agents, appointment-booking assistants, and after-hours answering workflows.

Launch Questions

• What caller intents can the AI receptionist handle without human review?
• Which intents must always route to a human?
• What happens when caller audio is unclear or intent confidence is low?
• What happens when the caller interrupts the assistant with a correction?
• What details can the assistant collect, and what details are forbidden?
• What does the assistant say when a caller asks whether it is AI?

Core QA Scenarios

• Supported booking with complete details.
• Booking request with missing address, date, service type, or contact preference.
• Calendar or CRM unavailable.
• Caller asks for pricing, refund, deposit, payment, medical, legal, or emergency help.
• Caller explicitly asks for a human.
• Caller reports that a previous AI call routed incorrectly.
• Transcript or intent classification appears wrong.
• Caller is angry, confused, silent, or repeating a failed request.

Pass Criteria

• The assistant does not invent unavailable appointments, prices, policies, or services.
• The assistant asks only for approved fields.
• The assistant uses a human handoff when the caller requests one.
• The assistant does not collect payment credentials, one-time codes, passwords, PHI, or private account records.
• The support team has a macro for explaining and reviewing failed routing.
• Every production change has an owner, date, reason, and approval status.