Ask for operational specificity
- Which exact call windows and job types are included?
- What information is required before a lead is qualified?
- Which calls transfer to a person, and what happens if nobody answers?
- What language handles safety-sensitive requests?
- Which changes can staff request, and how are they tested?
Verify integration claims
Ask the provider to name the calendar, CRM, or dispatch product and demonstrate the exact production path. A mock booking screen is not the same as a confirmed write. If direct booking is unavailable, require honest appointment-request language.
Understand data and call processing
Ask whether audio is recorded, whether speech is transcribed in real time, which vendors receive it, what is retained, whether vendor training is disabled, who can access records, and how deletion works. Caller disclosure and refusal must match the actual technical flow.
Compare the complete commercial offer
Review implementation fees, monthly fees, included usage, rounding, transfers, spam, overages, telephone numbers, integrations, support, trial conversion, cancellation, and data transition. The signed Order Form should control rather than a marketing-page headline.
Define evidence of success
Choose measures before launch: qualified opportunities captured, booking accuracy, successful transfers, follow-up time, corrections, caller complaints, and completed jobs where attribution is reliable. Avoid treating every answered call as revenue.
Straight answers
Frequently asked questions
Is the most human-sounding agent automatically best?
No. Voice quality matters, but workflow accuracy, latency, escalation, integration reliability, and support also determine usefulness.
Should we require a live demo?
Yes, but also ask to see failure and handoff scenarios—not only a scripted successful booking.
What contract documents should we expect?
A signed service agreement and Order Form should define scope, pricing, responsibilities, data handling, limitations, and termination.
What is a sensible first deployment?
A narrow after-hours or overflow workflow with human fallback, measured outcomes, and a tested rollback path.