The short version: AI voice agents, the software that answers your business phone and books the job before a human says a word, typically cost between 150 and 600 pounds a month once you add setup, per-minute charges and the retraining nobody warns you about. They work well for short, predictable calls like bookings and opening hours, and badly for anything emotional, unusual or complicated. Run one on a single phone line for 30 days before you touch your existing setup.
What an AI voice agent really does
Strip away the marketing and an AI voice agent is a phone line connected to a language model, a calendar, and a script. Someone calls your business, the AI answers in a voice that sounds close to human, asks a few questions, checks your calendar or CRM, and either books the appointment, takes a message, or passes the call to a person. Tools in this space include Synthflow, Retell AI, Bland, Goodcall and Vapi, and most of them were built on top of the same handful of speech models, so the differences between them come down to setup, integrations and price, not magic.
None of this is new technology dressed up. Call centres have used interactive voice response for decades. What’s changed is that the AI can now hold something close to a real conversation instead of making you press 1 for sales, and it can write the follow-up email itself. That’s the genuine leap forward, and it’s worth taking seriously if your business lives or dies on the phone.
The real cost, not the headline price
Every vendor quotes you the cheapest number they can. Here’s what I’ve seen small businesses pay once the full picture is in:
- Setup and script build: 0 to 800 pounds, depending on whether you do it yourself with a template or pay someone to write the conversation flow
- Monthly platform fee: 99 to 349 pounds for most small business plans
- Per-minute usage: 4 to 15 pence per minute of call time, which adds up fast if your calls run long
- Integration with your booking system or CRM: often included, sometimes an extra 30 to 100 pounds a month if you need a custom connection to something like a trade-specific job management tool
- Retraining time: nobody quotes this, but expect to spend two or three hours a month tweaking the script once you hear the calls that went wrong
Add it up and a small business handling 200 calls a month, average call length three minutes, is looking at roughly 250 to 400 pounds a month all in. That’s cheaper than a part-time receptionist on minimum wage, which in the UK for 2026 sits at 12.21 pounds an hour for over-21s, meaning even ten hours a week costs you around 530 pounds a month before tax and National Insurance. On pure cost, the AI usually wins.
A client, a boiler firm, and 40 missed calls
I worked with a small heating and plumbing firm in Kent last year, three vans, one office phone, and the owner answering calls from a job site whenever he could. He was missing around 40 calls a week. Industry research from Invoca has put the average small business missed-call rate as high as 60 to 70 percent during busy hours, and his numbers weren’t far off that.
We set him up with a voice agent that answered, asked the postcode and the problem, checked the job type against a simple decision tree, and booked straightforward jobs like boiler services directly into his calendar. Anything that sounded like an emergency, a leak, no heat in winter, got flagged and pushed straight to his mobile as a text with the caller’s number.
The result after eight weeks: he booked an extra 11 jobs a month directly from calls that would previously have gone to voicemail and never been returned. At an average job value of 180 pounds, that’s close to 2,000 pounds a month in bookings he wasn’t getting before, against a cost of about 280 pounds a month for the AI. That’s the kind of return that makes the decision easy, and it’s the same principle Alex Hormozi hammers on in his business lessons: the money is in the leads you’re already generating and failing to follow up, not in finding new ones.
What the AI didn’t do well was the awkward calls. An elderly customer worried about a gas smell called in a panic, and the AI’s calm, scripted tone made her feel unheard rather than reassured. She hung up and rang a competitor. That’s the uncomfortable bit vendors don’t put in the case study: the AI is excellent at the calls you’d happily hand to a junior member of staff, and bad at the calls where a person needs to hear a human voice slow down and say “I understand, let’s sort this.”
Where these tools quietly fall apart
The failure points I keep seeing across different businesses:
- Regional accents and background noise, particularly on mobile calls from vans or building sites, still trip up transcription more than any demo video shows
- Complex pricing that depends on a judgement call (how big is the job, how far is the drive) confuses the script and the AI either quotes wrong or stalls
- Customers who want to negotiate, complain, or ask something outside the script tend to get looped, repeating the same question back at them
- Trust: some customers hang up the second they realise they’re talking to a bot, especially older callers and anyone calling about a bereavement, a break-in or anything with emotional weight
This is the part most write-ups on this topic skip over entirely because it doesn’t fit a tidy success story. The honest picture is that AI voice agents are brilliant at capturing the bookings you’re currently losing to a full voicemail box, and they are not a replacement for a person who can read a room. If your business does a lot of emotionally loaded or high-value calls, hairdressers taking a first booking after a big life change, funeral directors, anything medical adjacent, keep a human answering and use the AI as backup only, not the front door.
How to test one before you commit
- Pick one phone number, not your main line, and forward it to the AI for 30 days. Keep your existing number answering normally.
- Write the script yourself first, in plain language, before you let a vendor’s “AI-generated” template do it for you. You know your customers better than a generic prompt does.
- Set a hard rule for what gets escalated to a human immediately: emergencies, complaints, anything a customer says twice because the AI misunderstood.
- Listen to every single call transcript for the first two weeks. This is the step people skip and it’s where you catch the tone problems before a customer does.
- Compare booked jobs and average call value against the same 30 days last year, not against a guess.
If you’d rather have someone set this up, do the integration work, and steer you away from the vendors overselling what their tool can do, that’s exactly the kind of practical project an AI consultant for small business should be handling for you, not a five-figure “digital transformation” strategy deck.
What this has to do with your wider marketing
A phone call is a conversion event, the same as someone using an interactive calculator on your website to work out a quote before they ever pick up the phone. Both exist to catch someone at the moment they’re ready to buy and stop them drifting off to a competitor’s tab. If you’re investing in an AI voice agent to catch phone leads but your website still makes people fill in a five-field contact form and wait three days for a reply, you’ve fixed the smaller leak and left the bigger one running.
Brands that get this right treat every touchpoint, phone, chat, website, as one continuous experience rather than a collection of separate tools. Look at how Airbnb built trust through consistent, human-sounding communication at every stage, as I’ve written about in the Airbnb marketing strategy breakdown, or how Away turned a simple returns policy into a whole brand promise, covered in the Away marketing strategy piece. The tool answering your phone is one small part of that bigger picture, not the whole strategy.
The bit that took me longest to accept
I resisted AI voice agents for longer than I should have, partly because I’ve spent years telling clients that people buy from people, and a synthetic voice felt like the opposite of everything I preach. What changed my mind wasn’t a slick demo, it was watching how many calls small businesses lose to nothing at all: no voicemail set up, no callback, no record the call ever happened. An imperfect AI that books 60 percent of straightforward calls correctly beats a voicemail box nobody checks, every time, on pure arithmetic.
That said, don’t let a vendor talk you into believing this replaces your judgement or your team’s relationship with regular customers. Napoleon Hill’s oldest lesson, covered in business lessons from Napoleon Hill, still holds here: definiteness of purpose matters more than any tool. Decide exactly what job you want the AI to do, booking simple appointments, answering opening hours, taking messages after 6pm, and refuse to let it creep beyond that without you checking the transcripts first.
When it isn’t worth it
Skip an AI voice agent entirely if:
- You take fewer than 20 calls a week; the monthly fee will cost more than the missed bookings ever did
- Most of your calls are already handled well by an existing receptionist or answering service you trust
- Your customer base skews older or calls you specifically because they want a person, common in funeral care, care homes, and some legal services
- You don’t have the time to listen to transcripts and adjust the script in the first month, because an unmonitored AI agent left running badly can lose you customers faster than a full voicemail box ever did
Content-led businesses tend to overlook the phone entirely and pour every pound into blogging and SEO instead, which works brilliantly for a company like Ahrefs, whose growth through useful free content is covered in the Ahrefs marketing strategy piece, but a plumber, hairdresser or dentist lives and dies by the phone in a way a software company never does. Match the tool to how your customers reach you, not to what’s trending in your industry group chat.
Free resource: The Voice Agent Call Flow Template.
Frequently asked questions
How much does an AI voice agent cost for a small business?
Expect 150 to 600 pounds a month all in, including the platform fee, per-minute charges and any integration cost with your booking system. A typical small business handling 150 to 250 calls a month usually lands around 250 to 400 pounds a month once everything is included.
Can an AI voice agent completely replace a receptionist?
For most small businesses, no. It handles simple, repeatable calls, bookings, opening hours, basic questions, well, but struggles with emotional, unusual or high-value calls where customers want to be heard by a person. Most businesses that succeed with these tools use them as backup and overflow, not a full replacement.
Will customers know they’re talking to an AI?
Often, yes, especially on longer calls or when the AI is asked something outside its script and starts repeating itself. Some customers won’t mind for a simple booking. Others, particularly older callers or anyone calling about something distressing, will notice and disengage.
How long should I trial an AI voice agent before deciding?
Run it for at least 30 days on a separate phone line while your existing setup keeps working normally. Listen to every transcript in the first two weeks and compare bookings against the same period last year before making it your main line.
Related reading: AI Receptionists for Small Business: What They Really Cost (And When They’re a Waste of Money) and AI Agents for Small Business: What They Are and Where to Start in 2026.
Want this done for you? See write for us about AI.
Free resource: grab The Brand Voice Prompt Template from the resource library.
