What an AI receptionist should never promise a trade customer
Keep prices, attendance, diagnosis, bookings, payments, and safety-critical decisions with the tradesperson unless explicit authority and controls exist.
An AI receptionist should not promise a price, attendance time, diagnosis, booking, payment outcome, or technical fix unless the business has explicitly authorised that action and built controls around it. Max & Mia's MVP does not include those commitments.
Price
“How much will it cost?” often depends on facts unavailable during the first call. The safe handover records what the caller reported and leaves the quote with the owner. Even a familiar job may have access, parts, travel, or scope differences.
Attendance and booking
Capturing “home after four” is not the same as booking a visit after four. The job card should distinguish customer availability from an agreed appointment.
Diagnosis and safety
A caller can describe sparks, water, smell, noise, or loss of service. The receptionist can preserve those observations. It should not turn them into a technical conclusion or instructions that belong with a qualified person and the business's approved safety process.
Supported work
Business context can state services and exclusions. When fit remains uncertain, mark it for owner review. A confident tone must not convert an unclear request into accepted work.
Make boundaries testable
Put restricted requests into test calls. Ask for a fixed quote, same-day guarantee, diagnosis, and unsupported service. Check the spoken response and final job card. The system should not smuggle a promise into a summary after declining it on the call.
Boundaries also need a failure path. When the service lacks information or authority, an honest close and review state are more useful than improvised certainty.
See how the call-handling mechanism works, then run the restricted-request case in the test guide.