Questions to Ask Before You Buy an AI Receptionist

Start with your phone, not their demo. For a week, keep a rough tally on a sheet of paper. When do calls go unanswered: lunch, mornings, after closing? Who is calling: new customers, regulars, people asking where you are? Which calls could anyone answer, and which ones need you?

This takes about five minutes a day. It also protects you from buying a fix for a problem you don't have. If most of your missed calls are at 6pm after you've locked up, you need something very different from a shop that's buried at noon.

This is the first question to ask, and the best one to test yourself. Every AI receptionist gets asked something it wasn't set up for. What matters is what it does next. Does it say "let me get someone to call you back" and take a name and number? Or does it make something up, or loop the caller around until they hang up?

Ask for a demo line and call it yourself. Ask something odd. Ask something it shouldn't answer. Talk over it. Mumble. A caller on a bad cell phone in a noisy parking lot is your real customer, not a clean demo.

A caller who says "let me talk to a person" should get one, or get a clear promise of a call back. Ask the seller whether it can pass a live call to your cell or the front desk, what happens if nobody picks up when it tries, and whether you can decide which kinds of calls always go straight to a human.

Emergencies, angry customers and anything involving a big job should have an easy path to you. If the answer is "it can take a message," that may be fine for after hours. It isn't fine for everything.

Local businesses get hurt here. A machine that guesses at a price, promises a Tuesday appointment you can't keep, or tells a customer you do something you don't, creates a problem you then have to fix with a real person.

Ask whether you can write exactly what it says about prices, hours and services, whether you can tell it what it must never promise, and whether it books into your real calendar or just collects a request for you to confirm. For anything that takes a look before you can quote, the right answer from the machine is "someone will call you back," not a number.

A good setup leaves you with something useful. Ask what you get after every call: a short summary, the caller's name and number, what they wanted, and how soon. Ask whether you can listen to the call or read a transcript.

Check where that information goes. Does it land in your text messages or email, or does it sit in a dashboard you will log into twice and forget? If you have to hunt for it, the calls you "saved" will still go cold.

Pricing in this space gets confusing fast. Ask for the whole picture in writing. Is there a setup fee? Is it a flat monthly price, or charged by the minute or by the call? What happens to the bill in a busy month? Does the price change after a trial or a first-year deal?

Then ask what you are paying for compared with other ways to cover your phone, such as a person at the desk for peak hours or a live answering service. A machine is usually cheaper for simple, repeat questions. A person is usually better for upset or complicated calls. Which one fits depends on your calls, which is why that week of notes matters.

Phone calls contain customer names, numbers, addresses and details about their vehicles, homes or health. Ask where the recordings and notes are kept and who can see them. Ask whether calls are recorded and how the caller is told, since rules on recording calls differ from state to state. If you aren't sure about your state, check before you turn it on.

Then ask about leaving. Is there a contract, and for how long? Can you cancel with a click or a phone call? Can you keep your own phone number? If you cancel, can you take your call records with you? If your business number is tied up in their system, quitting gets painful. Keep your number in your own name.

A few things are worth slowing down for. Big promises about money: any claim like "you will win back X in lost jobs" is the seller's estimate, not a guarantee, so ask for the claim in writing and ask how they measured it.

A demo that only works on easy questions is another flag. If they won't let you test it your way, that tells you something. So is no trial, because a seller who believes in the product should let you try it for a short period before locking in. And pressure to sign today is a reason to wait, since a real fix will still be there next week.

Start small and stay in the loop. First, pick one job. After hours, or overflow when you're already on another call, is the easiest start, because the alternative is a call nobody answers. Second, write the script yourself and decide what it can say and what it hands to you.

Third, call it yourself, then have a friend call it, before a real customer does. Fourth, read the call notes for the first few weeks so you see quickly where callers get stuck. Fifth, tell whoever answers your phone what's changing. The goal is fewer interruptions for them, not a replacement they have to compete with.