The short version: I handed my business phone over to an AI receptionist for a month, logged every single call, and it caught more leads than voicemail ever did, but it also nearly cost me a client because it sounded too polished and too corporate for the kind of trust-based work I do. If you run a small business and you're drowning in calls you can't get to, this is worth trying, but only if you write the script yourself first and check the transcripts every week.
Why I even bothered with this
For years my mobile rang and I answered it, or I didn't, and either way it felt like a small emergency each time. I run a consultancy on my own, no assistant, no office manager, just me between client calls, writing, and school pickup. Some weeks I'd get 15 to 20 calls that weren't in my diary: a prospect who found me through a podcast, a journalist, a supplier chasing an invoice, and the occasional wrong number from a courier trying to deliver to the house three doors down.
Every time I missed one, it went to a generic voicemail greeting I recorded in about 2019 and never updated. Call tracking companies have been saying for years that somewhere around one in three calls to small businesses never gets answered, and that most people who hit voicemail simply don't call back, they call the next name on the list. I believed that stat before I started this experiment. I believe it more now, because I watched it happen with my own missed calls in real time.
What I tested
I picked an AI phone answering tool built for small businesses, one of the newer ones that uses a voice model rather than a simple menu tree, and connected it to my existing business number using call forwarding. Setup took about 40 minutes. Pricing on tools like this typically sits somewhere between £30 and £150 a month depending on call volume, with the cheaper tiers capping you at a set number of minutes before you pay per minute on top. I went with a mid tier plan to be safe.
The pitch is simple: instead of a caller hitting voicemail, they get a voice that answers using your business name, asks what they need, answers straightforward questions from information you've fed it (opening hours, services, pricing ranges, location), books appointments straight into your calendar, and forwards anything urgent to your mobile with a text summary.
I ran it for exactly 30 days, from the start of a working month to the end of it, with no exceptions and no manually picking up calls "just this once."
The numbers after 30 days
The system logged 118 calls in total. Here's how they broke down:
- 34 were spam, PPI-style calls, or automated sales bots, filtered out before they even reached my day
- 61 were genuine questions the AI answered completely on its own, mostly "do you do X" or "what's your availability like" or "how does pricing work roughly"
- 23 needed a human, and the system correctly flagged and forwarded 19 of those to my mobile within seconds
- 4 it got wrong, either misunderstanding the request or routing it to the wrong place
That's a 90% hit rate on the calls that mattered. Before this experiment, my honest answer rate on incoming calls, based on my own phone log for the previous three months, was closer to 55%. That's the concrete win. Second, less obvious win: transcripts. Every call gets transcribed and dropped into a dashboard, which means I suddenly had a written record of the exact language prospects use to describe their problem, in their own words, not my marketing words. That turned out to be more useful than the calls themselves.
What went wrong
Here's the uncomfortable part nobody selling these tools puts in their case studies. A prospect who'd been referred to me by a former client called on day 12. The AI answered, sounded polished, professional, slightly American in its cadence even though I set the accent to British, and asked her the standard qualifying questions. She hung up before booking anything. When she eventually emailed me directly two days later, she said, and I'm paraphrasing closely because I kept the email, that she "wasn't sure if she'd reached a real business or a call centre" and nearly didn't bother following up.
That's the trade-off nobody talks about enough. An AI receptionist can make a one-person business sound bigger and more established, which is often exactly what you want. But if your entire pitch is "you're working directly with me, not a faceless agency," a slick automated voice undercuts the very thing you're selling. I sell trust and personal attention. A voice that sounds like it belongs to a 40-person call centre works against that positioning even while it's doing its job well.
I fixed it, but only after nearly losing that client, by rewriting the script from scratch in my own voice rather than using the tool's default template. I did this the same way I'd approach any piece of writing meant to sound like me: I sat down with a set of prompts designed to pull out how I naturally talk, similar to the process in my piece on using ChatGPT prompts for self-discovery and reflection, and fed the answers into the script instead of the generic corporate copy the tool suggested by default. The rewrite mentioned my name in the first line, used the phrase "I'll get back to you personally" instead of "a team member will follow up," and dropped every piece of jargon the default script had added. Complaints stopped after that.
How to set one up without the awkward bits
If you want to try this in your own business, do these in order, not the order the software vendor suggests:
- Write your own script first, on paper, in the exact words you'd use if a stranger called and asked what you do. Do not start from the tool's template.
- Record yourself saying it out loud once. If it sounds like a corporate brochure when you say it, rewrite it.
- Set a strict escalation rule for words like "urgent," "complaint," or a specific existing client's name, so those always get forwarded to you live rather than handled by the bot.
- Test it yourself by calling from a different phone and pretending to be three different types of caller: a confused prospect, an angry existing client, and someone trying to sell you something.
- Check the transcripts every single week for the first month. This is where you catch the tool misunderstanding regional accents, product names, or local slang.
- Update the script monthly based on the actual questions people ask, not the ones you assumed they'd ask.
If setting this up alongside your wider marketing stack feels like more than you want to take on solo, this is exactly the kind of small, contained project where a short engagement with an AI implementation coach earns its cost back quickly, because the biggest failures with these tools aren't technical, they're tone and trust failures that a second pair of ears usually catches faster than you will on your own script.
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.
The bit that surprised me most
The transcripts turned out to be a useful content source, which I hadn't expected going in. Every recurring question callers asked me became a line item I could answer on my site, or a short answer worth testing in the kind of structured content that AI search engines pull from. I ran a few of the most repeated questions through the same process I describe in the 15-minute AEO audit, and three of them weren't answered clearly anywhere on my site at all, which is a gap worth knowing about regardless of whether you ever touch a phone bot.
It also gave me raw material for content curation in the more traditional sense, real questions from real people rather than guesses about what my audience wants, which is the whole point behind why content curation matters for a business in the first place: you're not inventing topics, you're listening for the ones already being asked out loud.
Who this suits and who it doesn't
If you run a business where the phone rings with routine, answerable questions, a tradesperson, a clinic front desk, a local service business, a small agency fielding the same five questions on repeat, this pays for itself fast. The maths is simple: if the tool costs £70 a month and it saves even one lead a month that would otherwise have gone to voicemail and never called back, and that lead is worth more than £70 to you, you're ahead.
If your entire business is built on the promise of direct, personal access to you specifically, be honest with yourself about whether an automated voice, however good, quietly damages that promise before a prospect ever gets to speak to a human. The fix isn't skipping the tool. It's writing the script like you'd write your own bio, checking it against how you sound, and treating the first 30 days as a live test rather than a set-and-forget install. Brands that get voice and tone right, the way Red Bull built a consistent identity across every channel, didn't get there by accident, they tested the tone against real audience reactions and adjusted. Do the same with a script that's speaking on your behalf before a human even joins the call.
I keep every related walkthrough in the AI for Small Business: 25 Plain-English Guides to Automation, Marketing, Sales and Admin.
Frequently asked questions
How much does an AI receptionist cost for a small business?
Most tools built for small businesses sit between £30 and £150 a month depending on call volume, with extra per-minute charges once you go over your plan's included minutes. Budget for the mid tier if you get more than 20 to 30 calls a week, since the cheapest plans cap out fast.
Will customers know they're talking to an AI?
Often yes, and pretending otherwise usually backfires. The safer approach is writing a script that's upfront and personal rather than trying to pass the AI off as a human employee, since a caller who feels misled is more likely to hang up than one who simply knows they've reached an automated front desk that will get them to a real person quickly.
Does an AI phone answering tool replace a human receptionist?
For most solo and small businesses, no, it replaces voicemail, not a person. It handles routine questions and books straightforward appointments well, but anything emotionally loaded, a complaint, a confused elderly caller, a high-value new client, still needs a human on the line quickly, which is why escalation rules matter more than the AI's core script.
What's the biggest mistake businesses make when setting one up?
Using the tool's default script without rewriting it in their own voice first. The default templates sound like a generic call centre, and for businesses built on personal trust rather than scale, that mismatch is what loses leads, not the technology itself.
