Straight answer: an AI receptionist can pay for itself in the first week if you get a lot of after-hours calls you currently lose to voicemail, but it will quietly cost you your best leads if you use it as a permanent front door for anything urgent or expensive. I tested one for 30 days on a real client's line and the numbers surprised me in both directions.
Why I even bothered testing this
A client of mine runs a small boiler and heating repair business in Kent, three engineers, one office manager who also does the school run. Her biggest complaint wasn't marketing. It was the phone. She reckoned she was missing somewhere between four and six calls a day, mostly between 5pm and 9am, and every single one of those could be a £180 callout that walked straight to a competitor instead.
So we set up an AI phone agent for her main line for a full calendar month. I wanted to see it as a real business decision would see it, not as a demo on a stage. Cost, missed calls recovered, and what customers said when they realised they weren't talking to a person.
The numbers, honestly laid out
Over the 30 days the AI receptionist handled 214 calls that would otherwise have gone to voicemail or rung out. Of those:
- 89 were booked appointments or callback requests it captured correctly with name, address and job description
- 41 were people who hung up within 10 seconds of hearing it wasn't a human
- 34 were duplicate calls (same person ringing twice because the first call didn't feel finished)
- the remaining 50 were spam, wrong numbers, or suppliers chasing invoices
The monthly cost of the service was £129 plus a per-minute charge that came to roughly £43 that month, so call it £172 total. She converted 21 of those 89 captured leads into paid jobs averaging £165 each. That's £3,465 in revenue from a £172 spend. On paper that's a fantastic return, and if you stopped reading here you'd think every small business should do this tomorrow.
But that 41 out of 214 who hung up is the number nobody selling these systems puts in their pitch deck. That's nearly one in five people who rang a heating engineer, heard a synthetic voice, and left without saying what they needed. Some of those were probably spam or timewasters. Some almost certainly weren't. We have no way of knowing how many of those 41 rang a human-answered competitor five minutes later.
The uncomfortable bit vendors gloss over
Here's what I noticed listening to the call transcripts every week. The AI receptionist worked brilliantly for low-stakes, low-emotion enquiries: "can someone come and service my boiler next Tuesday," "what do you charge for a callout," "do you cover this postcode." It worked terribly for anything with urgency or fear attached to it. A woman with no heating and two small kids in a cold house on a Sunday night does not want to have a pleasant, articulate conversation with software. She wants a human to say "we'll get someone to you" in a voice that sounds like it means it.
That's the trade-off nobody wants to say out loud when they're trying to sell you a phone-answering subscription. These systems are good at capturing information. They are not good at reassurance. If your business sells anything where the caller is stressed, in pain, embarrassed, or scared, an AI voice at 2am can lose you the sale even while perfectly capturing the details on paper. If your business is bookings, orders, or basic scheduling, it barely matters. The caller wants an outcome, not a feeling.
What made it work (or not)
Three things separated the good weeks from the bad ones:
- The greeting had to say "AI" honestly in the first sentence. When we tried a version that pretended to be a regular receptionist, callers who twigged got annoyed and hung up mid-sentence far more often than when it was upfront. Being straight about what it was reduced the hang-up rate over the second half of the month.
- Someone had to check the transcripts daily, not weekly. The first ten days I let it run and only reviewed it every Friday, and two genuine emergency callouts sat unanswered in a dashboard nobody was looking at until Monday morning. That's the trap. People set these up, feel relieved they've solved the phone problem, and then stop checking the thing that's now their only line of defence against a missed job.
- It needed a human escape hatch within the first 15 seconds. Once we added "press 1 or say 'speak to a person' at any time," the number of frustrated hang-ups dropped noticeably, even on days with no engineer free to take the call. People just needed to know the option existed.
How to set one up if you decide to try it
You don't need a developer for this. Here's roughly how I did it for the client, spread over a single working day:
- Pull your last 90 days of call logs from your phone provider or CRM and work out what percentage of calls you're currently missing, and when. Most small businesses are shocked it's higher than they think, often 20 to 30 percent outside of core hours.
- Write down the five things a caller most commonly needs: booking, pricing, availability, an urgent request, or a complaint. Script how you'd want each one handled, including exactly where the AI should hand off to a human.
- Pick a provider that lets you review call transcripts and recordings, not just a summary. You need to hear the actual tone, not a bullet-point recap someone else wrote.
- Set the greeting to disclose it's an AI agent within the first sentence, and build in a spoken opt-out to reach a person.
- Review transcripts daily for the first two weeks, then move to every other day once you trust the pattern.
If any of that feels fiddly, this is exactly the kind of small, contained project worth a few hours with someone who's set these up before rather than a month of trial and error. I cover what that kind of help typically costs and what's fair to pay in this breakdown of AI consultant costs, because prices for phone agent setup range wildly and it's worth knowing what a fair quote looks like before you sign anything.
Working out if it's worth it for your business
Before you spend a penny, do the maths on what a missed call costs you. Take your average job or sale value, multiply it by your close rate on inbound calls, then multiply that by the number of calls you think you're missing each month. For a business with a £150 average job, a 25 percent close rate, and 40 missed calls a month, that's £1,500 walking out the door every single month, before you've spent anything on the AI agent to catch it.
I built a rough version of this sum for the client on a spreadsheet in about ten minutes, but if you want something you can hand to a business owner without them needing to open Excel, this is exactly the kind of thing an interactive calculator on your own website does far better than a static blog post. I wrote about building these for visitors in this piece on interactive calculators, and a "missed call cost calculator" is a useful lead magnet for anyone in a trade or service business, not just a nice-to-have.
Who should skip this entirely
If you sell anything high-trust, high-emotion, or high-ticket where the first phone call is where the relationship starts, be careful. Funeral directors, family solicitors, anyone doing emergency callouts where the caller is frightened, anyone selling something over £1,000 where the voice on the other end is doing real persuasion work. I wouldn't put an AI voice on the front of any of those without a very fast, very obvious human handoff, and even then I'd think twice.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
If you're a tradesperson doing routine bookings, a clinic doing appointment scheduling, a restaurant taking reservations, a salon confirming slots, or any business where the call is transactional rather than emotional, it's a sensible tool and the 30-day test above shows real, provable return.
What I'd do differently next time
I'd set the daily transcript review from day one instead of day ten. Those two missed emergency calls in week two were entirely avoidable and they're the sort of thing that makes an owner distrust the whole system, fairly. I'd also test a version where the AI only picks up outside business hours and after three rings during the day, rather than answering every single call. Half the value here is catching the calls a human can't get to, not replacing a person who's sitting right there.
None of this is complicated technology in the way people assume. If you've read anything I've written before about being non-technical and still using AI daily, this is the same rule applying to phones instead of writing or research. You don't need to understand how the voice model works. You need to understand your customers well enough to know where a synthetic voice will help them and where it will annoy them, and be honest enough to admit the difference.
And if you're hunting around for the actual tools rather than the marketing pages around them, I'd rather point you at a proper directory than a single sponsored link. I've collected a fair number of these lesser-known but useful tools in this list of 50 websites you probably haven't heard of, several of which are worth a look before you commit to a big-name phone agent provider on price alone.
Frequently asked questions
How much does an AI phone receptionist cost for a small business?
Expect somewhere between £50 and £300 a month as a base subscription, plus a per-minute usage charge often between 15p and 50p a minute depending on the provider and call volume. For the client in this test, total cost including usage came to roughly £172 for a month of real calls.
Will customers know they're talking to AI?
Often yes, and the businesses that disclose it upfront in the greeting see fewer frustrated hang-ups than the ones that try to make it sound like a human receptionist. Being straightforward about it, and offering an instant way to reach a real person, matters more than making the voice sound convincing.
Is an AI receptionist better than voicemail?
For capturing basic booking and enquiry information, yes, clearly better, because it collects details in real time rather than leaving a caller to hang up on an answerphone. For anything urgent or emotionally charged, a fast human callback still beats both options, so the best setups use AI as a first catch with a quick human follow-up rather than a full replacement.
What kind of business should not use one?
Anywhere the phone call is doing real trust-building work before a big or sensitive decision, such as legal services, funeral care, medical practices, or high-ticket sales, should be very cautious about putting an AI voice on the front line without an immediate, obvious route to a human.
Related: looking for an AI consultant for small business? Here is how I can help.
Related reading: The AI Subscription Creep Nobody Warned Small Business Owners About and How to Get Your Small Business Mentioned by ChatGPT (Not Just Ranked by Google).
