What an AI receptionist actually does
An AI receptionist is software that answers your business phone with a synthetic voice. It greets callers, answers common questions from information you give it, takes messages and, in many products, books appointments into your calendar or sends details to your team. Some answer every call, some only after hours, and some only when you don't pick up.
It isn't a person. It works well for predictable calls (hours, service area, "can someone come out Tuesday?") and less well for calls that need judgment, empathy or a quote. If you're still choosing between this, a missed-call text-back and a live answering service, start with our comparison of the three fixes. This guide is about whether to buy and what it should cost you.
Step 1: count your missed calls for a week
Most owners guess how many calls they miss. A week of counting replaces the guess with a number. Use your phone carrier's call log, your VoIP system's report, or a tally sheet by the phone. For each missed call, note:
- Was it a new customer, an existing customer, a supplier or spam?
- Did they leave a voicemail?
- Did you reach them when you called back, and did it turn into work?
- What time was it: during jobs, at lunch, after hours or at weekends?
The new-customer calls that never turned into a conversation are the ones an AI receptionist could recover. The time pattern tells you whether you need all-day coverage or just after hours.
Step 2: the break-even math
Use your own numbers. The formula is simple:
Monthly value recovered = missed new-customer calls per month × the share you'd book if answered × average gross profit per job
Worth it if monthly value recovered is clearly more than the monthly fee plus your setup time
An illustration, not a benchmark: say your week of counting finds 5 missed new-customer calls, so roughly 20 a month. If you'd normally book 1 in 4 answered calls, that's 5 jobs. At $300 gross profit per job, that's $1,500 a month in work you're currently not getting. Even if the AI recovers only a fraction of it, most monthly fees are covered. If your count finds 2 missed calls a month, the math rarely works, and a simple text-back may be enough.
Use gross profit, not the invoice total. And be honest about the booking rate: a caller who gets a robot may book less often than one who gets you.
What drives the cost
AI receptionist pricing varies widely and changes often, so we don't quote ranges here. These are the things that move the price, so you can compare quotes properly:
- Pricing model. Flat monthly fee with included minutes, per-minute or per-call billing, or a mix. A busy season can change the bill a lot on usage pricing.
- Coverage. After-hours only, overflow only (when you don't answer) or every call.
- Booking and integrations. Taking a message is simpler than booking into your calendar or field-service software, and integrations often cost more.
- Setup. Whether you configure it yourself or pay for setup, and how much of your time training it takes.
- Contract terms. Monthly versus annual, minimums, and what it costs to leave.
Step 3: run a 30-day trial with a pass mark
- Week 1: count missed calls (above) and write down your pass mark, for example "at least 3 booked jobs I would otherwise have lost."
- Setup: give the tool your hours, service area, services, prices you're happy to share and the questions it must hand to a person. Call it yourself ten times with real-sounding questions.
- Weeks 2 to 4: turn it on for overflow or after-hours first. Read every transcript or summary for the first two weeks.
- Day 30: count booked jobs that came through it, complaints or confused callers, and your time spent fixing things. Compare against your pass mark and decide.
When not to buy one
- You miss very few new-customer calls. Fix voicemail and call-back habits first.
- Your calls are mostly emergencies or sensitive (legal intake, health concerns) where a wrong answer is costly. Get a person, or use AI only to take messages.
- Your callers expect to haggle or get a quote on the spot, and you can't give the AI rules for that.
- You won't have time to read transcripts in the first month. An unsupervised AI receptionist is a risk to your reputation.
Ask any vendor whether the product makes outbound calls, such as reminders or follow-ups. The FCC has confirmed that AI-generated voices in robocalls fall under the Telephone Consumer Protection Act, so those calls need the right consent (FCC).
The phone is often only one of several leaks. Your business's AI plan looks at calls alongside quoting, follow-up and admin, so you can see whether a receptionist is the best first spend or the third.