AI receptionists in the UK: what they cost, which to buy, and the one case where you should build your own.

An AI receptionist is software that answers your business phone, talks to the caller in natural language, answers the questions it has been given answers to, takes a message or books an appointment, and sends you a summary. It is not a virtual receptionist, which is a remote human answering on your behalf at £25 to £40 an hour, and it is not an automated attendant, which is the "press one for sales" menu that has been irritating callers since the nineties.

In 2026 the category has matured quickly. The two largest UK hosting companies now sell one for £39 a month, setup takes fifteen minutes, and the voices are good enough that, in our own production experience with voice agents, most callers do not ask whether they are speaking to software unless the agent tells them, which it should.

This guide does three things. It surveys what thirteen providers charge and include, from their own pricing pages, dated. It says honestly which product suits which kind of business, including several we have no connection with. And it draws the line, with the running costs worked out, between the cases where a subscription is the right answer and the narrow case where building a bespoke agent earns its price. We build bespoke agents. Most readers of this page should not buy one.

Every vendor in this market opens with a missed-call statistic. Most of those statistics are the vendors’ own. The most-cited UK study, 142 small and medium businesses measured in 2025, found that 47 per cent of first-time calls went unanswered during business hours and 27 per cent of all calls across the week, with the misses peaking at lunchtime, after 5pm and on Saturdays. Miss rates were highest in heating and home services (62 per cent) and plumbing (38 per cent), lowest in dental practices (22 per cent) and estate agency (20 per cent). That study was run by Paperclip, a telephone answering company, and to its credit its own write-up says the research in this field is sponsored by businesses with "vested interests" and relies on self-reporting.

The line you will see everywhere, that 85 per cent of callers never try again, has no primary source we could trace. The conservative figure comes from PA Media’s May 2025 research: one in five UK customers will not attempt a second contact after a single missed call. Answer4u, another vendor, estimates missed calls cost UK businesses £30 billion a year, about £5,500 each; BT Business puts the average value of a call to a UK service business at £1,200. Treat those as the industry’s view of itself. The number that matters is yours: how many calls you miss a week, and what one of them is worth. The worked examples below use that instead.

Prices as published on the providers’ own pages in September 2026, excluding VAT unless stated. Where a provider does not publish, the figure comes from the comparison guides cited in the method section, and the table says so.

Yell’s AI Receptionist is bundled with its Amplify and Accelerate marketing plans and has no standalone price. Prices change; the date at the top of this page is the last check.

Reconciled, the market has three tiers of software and one of people. The gaps between them are the whole decision.

Employed cost from the ONS Annual Survey of Hours and Earnings 2025: full-time median for receptionists £24,678, before employer National Insurance and pension.

We have no commercial relationship with any of these. These are the fits the published features and the comparison guides support.

Dentistry is the one sector where the generic products above are usually the wrong answer, because a dental call has to end inside the practice management system, with the right diary, the right patient type and the right escalation. A receptionist that books into a Google calendar is no use to a practice running Dentally or Software of Excellence. So the first filter is not price or features but integration: which products write into the system you already have.

The pressure is real and it is measurable from official sources rather than vendor claims. NHS dentistry in England saw 18.8 million adults in the two years to March 2026, 40 per cent of the adult population, and urgent courses of treatment rose 8.4 per cent to 3.97 million. Of people who tried to get an NHS appointment, 78 per cent succeeded. On the other side of the desk, the British Dental Association’s vice-chair told the BBC that one in seven NHS patients failed to attend at his practice over the past year, at a cost of about £56,000, and that the rate had been under 5 per cent before the 2006 contract removed the right to charge for missed appointments. Since an NHS practice cannot charge for a DNA, confirmations, reminders and fast rebooking are the only lever, and they are exactly what an integrated receptionist does well.

Vendors quote missed-call rates for dental practices of 22 to 40 per cent. We could not find an independent UK figure and we do not repeat theirs as fact; the honest position is that a practice can measure its own in a fortnight from the phone system’s logs, and should, before buying anything.

Integration and feature claims are the vendors’ own, read on their pages in September 2026 and not independently tested. Prices exclude VAT where stated.

They are inbound products. They answer, they book into a calendar, they take a message, and the better ones transfer. Read the feature lists and the same five gaps appear.

This is the part vendors of bespoke agents, including us, tend to leave vague. The platforms that bespoke agents are built on publish their prices, so the running cost can be worked out to the penny.

Dollar prices converted at roughly $1.30 to the pound. Add-ons such as knowledge bases and PII removal are fractions of a cent per minute.

Five hundred minutes of calls a month on a bespoke agent costs £40 to £75 to run. Two thousand minutes costs £160 to £300. That is the same money as the subscriptions charge. The running cost is not where bespoke wins or loses. The build is: £4,000 to £15,000 for an agent that reads messy inputs, applies your criteria and writes into your systems with a person approving the risky steps, against £39 a month and fifteen minutes for a subscription.

So the rule is short. Do not build a receptionist. If the call only has to be answered, booked and summarised, buy one of the products above and spend the money you saved on something else. Build only when the conversation has to do something a receptionist cannot: screen against your criteria, book against your real rota or system, call out as well as answer, or hand off a decision rather than a transcript. When that is true, the £4,000 to £15,000 buys an outcome the subscription cannot produce at any price, and the running cost is a rounding error against it.

One honest caveat from our own production data on the outbound side. When an agent calls out rather than answers, about half of calls reach voicemail, a quarter never connect, and roughly one in eight reaches a person. Inbound agents do not have that problem, which is one reason the subscriptions work so well; outbound agents have to be designed around it, with a text in the same minute as the fallback. That design is part of what the build price pays for.

Illustrations priced from the tables above, not client accounts. Put your own call count and call value in and the arithmetic holds.