The short version: AI phone agents now answer calls, book jobs, and text customers back for £29 to £300 a month, and for most small businesses they pay for themselves inside the first fortnight. But they don’t fix a broken sales process, they just answer it faster, and that’s the part nobody selling you the software wants to say out loud.
Why I even started looking at this
A client of mine, Mike, runs a landscaping business near Tunbridge Wells. Three vans, a busy diary, and a habit of leaving his phone in the van while he’s up to his knees in someone’s flowerbed. He was missing calls constantly. Not a handful. In one week I tracked with him, he missed 31 calls and only called 8 of them back. That’s 23 potential jobs going to whoever answered their phone first, probably a competitor with a bigger crew and a worse quote.
We set up an AI phone agent (Goodcall, though there are half a dozen decent options now) to pick up anything that went unanswered after four rings. It asked for name, postcode, the job, and a rough budget, then texted Mike a summary and offered the caller a slot from his actual calendar. Six weeks in: it caught 23 of those missed calls, booked 9 estimates, and those 9 estimates turned into roughly £14,000 of quoted work. Not all of that landed as booked jobs, but even at a 40 percent close rate that’s £5,600 that used to just disappear.
That’s the case for it. Now the bit most people skip.
What these tools are, in plain terms
There are three tiers, and it matters which one you’re buying because the price gap is huge and the wrong choice wastes money either way.
- Screening apps like Simple Phone (around £29 a month) sit on your existing mobile number, decide if a call is worth taking, and either put it through or take a message. Good for a one-person business drowning in cold calls and spam.
- Booking receptionists like Goodcall or Rosie (Ruby’s AI product) run £59 to £199 a month, answer in a proper business voice, take details, and book directly into your calendar. This is what most trades and service businesses need.
- Custom voice agents built on platforms like Retell AI or Bland AI charge per minute, roughly 8p to 15p a minute, and are what agencies build for clients who want something specific, like a booking flow tied to a scheduling system or a script that handles objections. Costs here can run £200 to £1,000+ a month depending on call volume, and you’ll usually need someone technical or an AI implementation coach to set the logic up rather than paying an agency a fortune to do it for you.
Most small businesses only need tier two. Tier three is for volume you probably don’t have yet, and paying per-minute rates when you’re only getting 40 calls a month is money down the drain.
The setup that works
Skip the software for a minute. Before you touch any tool, do this:
- Count your missed calls for one week. Not a guess, an actual count. Most owners think it’s “a few” and it’s closer to a third of all inbound calls.
- Write your real script first, on paper. What do you ask a new caller? What do you need to know before you can quote? Write it exactly as you’d say it, then feed that into the AI setup rather than accepting the tool’s generic template.
- Set hard escalation rules. Decide, in writing, when the AI should transfer to a human immediately: existing customers, anything over a certain job value, anyone who sounds upset. This is the step almost everyone skips and it’s the one that saves your reputation.
- Test it with people who don’t know the script. Get a friend to call and ask something odd. If the agent falls apart on a weird question, your customers will hit that same wall.
- Listen to the first 20 real calls yourself. Not the transcript summary, the actual recording. This is where you learn what customers are really asking, which is usually not what you assumed when you wrote the script.
That last point is the one I’d bet money most owners never do. They set it up, it works well enough not to notice, and they never listen to a single call again. Which means they never find out that half their callers are asking about something the AI handles badly, or that people keep asking a pricing question the AI dodges because nobody told it the answer.
The uncomfortable bit
Here’s what doesn’t get said in the “AI receptionist saves you £X a month” posts: some customers hate this, and for certain businesses that matters more than the money you save.
If you sell something high-trust, a wedding, a funeral, a big renovation, a first legal consultation, a chunk of your callers want to hear a nervous, real, slightly imperfect human on the other end because that’s part of what they’re buying reassurance from. Hand that call to a voice agent, however well built, and some of those people will hang up, or book with someone else, and you will never know why because you weren’t the one who lost the sale, the AI was, and it doesn’t tell you it felt off.
The other honest truth is that AI phone agents don’t fix bad businesses, they just answer the phone faster. If your quoting process is slow, if you don’t call people back within the hour, if your pricing is unclear, an AI receptionist will capture the lead beautifully and then the lead will sit there while the rest of your process falls apart exactly as it did before. I’ve seen this with more than one client: the call gets answered, the booking gets made, and then nobody follows up on the quote because that part of the business was always the weak link, not the phone.
Think of it the way you’d think about tracking a website. Watching where a customer drops off, the way Hotjar’s marketing strategy is built around showing you exactly where people get stuck, is only useful if you fix what you find. An AI phone agent is the same. It shows you where your business is losing people. Whether you do anything about it is a separate decision entirely.
Getting the voice and follow-up right
Most of these tools let you choose a voice and a tone, and this matters more than the pricing page suggests. A calm, unhurried voice suits a spa or a therapist’s booking line. A brisk, friendly one suits a plumber. Get this wrong and it sounds like a corporate call centre bolted onto a two-van business, which undoes the trust an AI agent is supposed to build. Calm’s entire marketing strategy is built on one consistent, reassuring voice across everything they make, and that’s worth copying at a much smaller scale: pick a tone and keep it identical across your voicemail, your texts, and your AI agent, rather than letting each one sound like a different business.
The follow-up sequence matters just as much as the call itself. Once the AI books an estimate, what happens next? Most tools will fire off an automatic text confirmation, but you can go further and set up simple “if this call happens, then this follow-up sequence starts” rules, the same logic that made IFTTT’s whole product useful for years before AI made it easier. A missed-call catch, a booked estimate, a no-show, each should trigger a different next step, not one generic “thanks for calling” text that gets ignored.
And if you want people to turn up to the appointment the AI booked, a short reminder sequence with a bit of repetition works better than one text and hoping. Duolingo built an entire brand on the idea that gentle, repeated nudges beat one perfect message, and the same logic applies to a booked estimate: one reminder the day before and one an hour before will cut no-shows more than any clever script ever will.
What it costs, all in
For a one-person or small trade business, budget £60 to £150 a month for a decent booking-style AI receptionist, plus your time to write the script and check the calls weekly for the first month. For a business with genuine call volume, 100+ inbound calls a month, budget £200 to £500 and expect to want a proper setup done by someone who’s built these flows before rather than relying on the default template. If you’re not sure which tier you need, that’s exactly the kind of decision worth a short conversation with an AI consultant before you commit to a monthly fee you might not need.
One more number worth knowing: most of these tools offer a free trial or a free tier capped at a small number of minutes. Use it. Run it alongside your normal phone for two weeks, compare the missed-call count before and after, and only then decide if the paid tier earns its keep. Mike’s trial paid for itself in the first ten days. Not every business will see that, and you won’t know until you’ve counted.
If you’re capturing leads on your website as well as your phone line, it’s worth thinking about both together rather than treating them as separate problems. A caller who hangs up without booking might still convert if your site has something interactive to catch them first, which is the whole idea behind using a calculator or quote tool to keep a visitor engaged instead of losing them to a closed tab.
Frequently asked questions
How much does an AI receptionist cost for a small business?
Expect £29 to £59 a month for a basic call-screening app, £59 to £199 a month for a full booking-style AI receptionist that answers, qualifies and books calls into your calendar, and £200 to £1,000+ a month for a custom-built voice agent charged per minute, which only makes sense once you have real call volume to justify it.
Will customers know they’re talking to AI?
Most current tools sound convincingly human on a first call, but many businesses now disclose it upfront anyway, both for trust and because some regions are moving toward requiring it. Being upfront rarely loses the booking; a robotic-sounding script that fumbles a simple question loses far more.
Is an AI phone agent worth it for a one-person business?
Yes, if you’re currently missing calls because you’re on the tools or with a client, and no, if your problem isn’t the phone but slow follow-up afterward. Count your missed calls for a week first; the answer usually becomes obvious once you have the real number in front of you.
What’s the biggest mistake people make setting one up?
Using the tool’s default script instead of writing their own first, and never listening to the actual call recordings once it’s live. Both mean you never learn what customers are really asking, and the agent keeps making the same small mistakes for months without anyone noticing.
Related reading: Why Your Analytics Are Hiding How Many Customers AI Is Sending You and Dropbox Marketing Strategy: How They Built a Brand That Wins.
