What to ask before you hand your phones to an AI receptionist

Most of the difference between AI phone products isn't the voice. It's how much work lands back on you after you sign. These are the questions that surface that early.

About the setup

• Who writes the call flow — your team, or mine? If the answer involves you writing prompts, you've bought software, not a service.
• Who connects it to my calendar and CRM? Integration work is where do-it-yourself platforms quietly hand you a project.
• What do you need from me, and how long will it take? A serious answer is measured in one conversation, not a series of onboarding tasks.

About the calls it will actually get

• What happens on a call it can't handle? You want a specific rule you control — transfer, escalate, or take a detailed message — not a vague reassurance.
• How does it decide something is an emergency? Then ask who gets woken up, and confirm you set that, not them.
• Was it tested before it went live? Ask what scenarios were run. "We tested it" and "we ran your ten worst calls through it" are different answers.

About six months from now

• What happens when I add a service or change my hours? If a change request goes into a queue or costs extra, that's a maintenance burden with a delay attached.
• Am I locked in? Month-to-month keeps the incentive on the provider to keep it working.
• What does it cost, all in? Watch for a low headline price with setup, integration, and change fees underneath it.

The tell

Ask any provider to describe what you'll be doing in week two. If the answer includes logging into anything, you're going to end up maintaining it. If the answer is "answering the jobs we booked for you," that's the service you were trying to buy.