The short version: AI phone answering for a small business runs anywhere from about £25 a month for a basic missed call text back tool to £250 plus a month for a full AI receptionist that books jobs and answers questions. Most small businesses don’t need the expensive version. They need someone (or something) to stop losing the caller in the first thirty seconds, and that’s a much cheaper problem to solve than the software companies want you to believe.
What I mean by AI phone answering
There are three different products hiding under this one label, and nobody selling them makes the difference clear.
- Missed call text back. You don’t answer, the caller gets an instant text asking what they need and when’s good for a callback. No voice AI involved. Cheap, dumb, works.
- AI voice agent. A synthetic voice answers, holds a conversation, books an appointment or takes a message, and hands off to a human when it gets stuck. This is what most people mean when they say “AI receptionist.”
- Hybrid human plus AI. Real people answer the call but AI is doing the scheduling, note taking and follow up in the background. Smith.ai works this way, and so does a chunk of the market that markets itself as “AI” but is really trained humans with software behind them.
Knowing which one you’re buying matters, because the pricing and the failure points are completely different for each.
What it costs in 2026
Real numbers, not “starting from” fog:
- A basic missed call text back tool through a CRM add on typically sits between £20 and £40 a month.
- Dialpad’s AI voice features come bundled inside plans starting around £15 to £20 per user a month, which is cheap if you already use them for business calls anyway.
- Dedicated AI voice receptionist products (Rosie, Air AI style tools, and similar) tend to land between £40 and £150 a month depending on call volume, with per minute overage charges once you go past your bundled minutes.
- Smith.ai’s hybrid human plus AI service starts around £230 to £280 a month for roughly 30 calls, and climbs fast after that because you’re paying for real receptionist time, not just software.
- A custom built voice agent set up by a developer or consultant, wired into your booking system and CRM, usually costs £500 to £2,000 to build plus £50 to £200 a month to run.
Notice the gap between the cheapest and most expensive option is nearly ten times the price for the same basic outcome, which is: someone or something answers when you can’t.
The locksmith who taught me this the hard way
A client of mine runs a locksmith and key cutting business covering three towns in Kent. Call it what it is: emergency work, lock outs, van keys, the sort of job where the caller is standing outside their own front door in the rain at 9pm. He was missing roughly 40% of calls that came in outside his 8 to 6 working hours, and every single one of those was a caller who then rang the next locksmith on the Google list.
We worked out, from three months of call logs, that he was losing around £2,800 a month in jobs he never even knew he’d lost. Not because he was bad at his job. Because nobody was picking up the phone at 9pm.
We set up an AI voice agent first, the full version, because it looked like the proper solution for an emergency trade. It took the caller’s postcode, the type of lock, gave a rough quote, and offered a callback slot. Sounds great on paper.
In practice, over six weeks, 22% of calls to the AI agent ended in a hang up within the first 15 seconds. We pulled the transcripts. The pattern was obvious once we saw it: callers in genuine distress, standing in the cold, wanted a human voice that sounded like it understood urgency, and the AI’s calm, evenly paced responses read as indifferent even when the words were fine. One transcript had the caller say “just get me an actual person” three times before hanging up.
Where these systems quietly fall down
This is the bit that gets left out of most write ups on this topic, usually because the people writing them are selling the software.
First, accents. We tested the same AI agent with callers using broad Glaswegian, Scouse, and a strong Kent accent, and the misrecognition rate on postcodes and street names was noticeably worse for the Glaswegian caller, roughly one in four attempts needed a repeat. The models are trained on a huge amount of American and standard Southern English speech, and regional UK accents still trip them up more than the marketing pages admit.
Second, and this is the uncomfortable part: a chunk of your callers, especially older customers and anyone in a genuine hurry, will hang up the second they clock they’re talking to a machine, regardless of how good the voice sounds. This isn’t a technology problem you can fix with a better model. It’s a trust problem. Richard Thaler’s work on how people behave rather than how we assume they’ll behave is worth reading here, and there’s a good breakdown of his business lessons from Richard Thaler that applies directly to this: people don’t respond to a phone call the way a spreadsheet assumes they will. They respond emotionally, and a synthetic voice reading a script, however smooth, doesn’t pass the emotional test for a real slice of your customer base.
Third, the ROI on the expensive full voice agent, once we measured it against the cheap missed call text back tool, was barely better. The text back tool costing £30 a month recovered about 60% of the missed calls we’d have otherwise lost, because most people, given a text within ten seconds, will happily reply with their name and problem rather than wait on hold or repeat themselves to a machine.
The cheaper fix most small businesses need
For the locksmith, we ended up running both, but with the AI voice agent switched off outside emergency hours and the text back tool covering everything else. His missed call recovery went from 40% lost to about 12% lost, and the monthly cost dropped by more than half compared to running the full voice agent around the clock.
If you’re a small business owner reading this wondering where to start, don’t start with the flashiest option. Start with:
- Turning on missed call text back through whatever CRM or phone system you already have.
- Writing three lines of script for that text: who you are, what you need to know, and when you’ll call back.
- Reviewing every missed call for one month and noting what time of day they came in, because that tells you whether you need after hours cover at all.
- Only then deciding if a full AI voice agent is worth the extra £100 to £200 a month, based on actual missed call volume, not on how impressive the demo looked.
A well built website with clear pricing and an interactive calculator that lets a visitor work out roughly what a job will cost before they even pick up the phone will cut your call volume in the first place, which is often a better use of £500 than any voice AI subscription. Phone cover is just one AI workflow for small business that saves time; the same build-once, measure-first approach applies to email, scheduling and follow-ups.
How to set this up well in a week without wasting money
This is the order I run it in with clients:
- Day 1: Pull three months of call logs from your phone provider. Count missed calls by hour and day. Most owners guess wrong about when their busy periods are.
- Day 2: Set up missed call text back on your existing system if you have one, or a £20 to £40 a month add on if you don’t.
- Day 3: Write the greeting and the three follow up questions your text needs to ask. Keep the tone the way your brand talks. Coco Chanel understood better than almost anyone that consistency of tone builds trust faster than novelty does, and there’s a lot to take from her approach in these business lessons from Coco Chanel when you’re deciding how a script for your business should sound.
- Days 4 and 5: Run it live. Don’t touch anything else yet.
- Day 6: Review the transcripts and replies. Look specifically for hang ups, repeats, and confused customers.
- Day 7: Decide, based on actual missed call volume from day one, whether a full AI voice agent is worth adding, and if so, which provider fits your call type, since a trades business and a professional services firm need very different scripts and escalation rules.
When to bring someone in rather than DIY it
If your business takes more than 50 calls a week and even a small percentage of missed calls is costing you real jobs, the maths usually justifies paying someone to set this up rather than fumbling through it yourself over a month of trial and error. This is exactly the kind of narrow, practical project worth getting outside help on rather than a broad “do everything with AI” engagement, and if you’re weighing that up it’s worth reading through what an AI consultant costs before you commit to a monthly software subscription that might not fit how your customers behave on the phone.
Bigger brands get this right because they test relentlessly before rolling anything out nationally. GoDaddy’s approach to building a brand small businesses trust is worth a look for exactly this reason, they never assume a tool works just because it’s new, they test it against how their actual customers behave first.
The uncomfortable bit nobody selling AI phone systems wants to say out loud
The AI voice agent industry sells the story that within a year or two, callers won’t be able to tell the difference between a human and an AI receptionist, and that trust concerns will disappear as the voices get smoother. I don’t think that’s true, and I’ve watched enough call transcripts to say so with some confidence. A synthetic voice can now sound remarkably close to human, but a distressed or hurried caller isn’t listening for vocal quality, they’re listening for whether the thing on the other end understands what they need right now. Smoothness doesn’t fix that. What fixes it is knowing which calls need a human and routing them there fast, and being honest with yourself about the fact that a chunk of your customers will always prefer a person, no matter how good the technology gets.
Free resource: The AI Agent ROI Calculator Cheat Sheet.
Frequently asked questions
How much does an AI receptionist cost for a small business in the UK?
Expect £40 to £150 a month for a straightforward AI voice agent handling a moderate call volume, £230 upward for a hybrid human plus AI service like Smith.ai, and as little as £20 to £40 a month if you only need missed call text back rather than a full voice agent.
Is AI phone answering worth it for a small trades business?
Usually yes for after hours cover, since a missed emergency call almost always goes straight to a competitor. But test missed call text back first, since it often recovers most of the lost business at a fraction of the cost of a full voice agent.
Do AI phone systems understand strong regional UK accents?
Not consistently. In testing, broad regional accents such as Glaswegian produced noticeably more misrecognitions on postcodes and place names than standard Southern English accents, so it’s worth testing with your actual customer base’s accents before committing to a system.
What’s the biggest mistake small businesses make with AI phone answering?
Buying the most expensive full voice agent before measuring how many calls they’re missing and when. Most businesses find a cheaper missed call text back tool recovers the bulk of lost business, and the expensive AI voice agent only earns its cost on top of that for a specific slice of calls, usually after hours emergency work.
Related reading: Your Business Now Has an AI Reputation, and You Probably Haven’t Checked It and Why ChatGPT Might Already Be Recommending Your Competitor.
