What does an AI receptionist really cost in the UK - and what changes the price?

Ask "how much does an AI receptionist cost in the UK?" and you will get answers so far apart they seem to describe different products - because they do. The phrase covers a chat widget bolted to one website, a six-channel system booking appointments around the clock, and a voice agent answering the phone, and the honest price of each depends on choices nobody has asked you about yet. So this piece does something slightly unusual for an article written by someone who sells the thing: it names no prices at all. Figures printed in blog posts go stale within months, and a figure without a spec misleads more than it informs. What stays true are the six factors that move every quote you will ever get, from us or from anyone. Our current numbers live on the pricing page, dated and kept up to date; this article is the part that helps you read them - and everyone else's.

The six factors that actually move the price

1. Which channels it covers

A receptionist that only watches your website chat is a different machine from one answering WhatsApp, web chat, Instagram and Facebook DMs, SMS and email in one place. Every added channel means more connection work, more testing and more surface to keep accurate - and more of your actual enquiries caught, because UK customers scatter across all of them without warning. The 9pm WhatsApp, the Instagram comment, the form fill and the Friday voicemail are the same buyer using different doors. Pull up your last 50 enquiries and list where they arrived: that list is your channel spec, and the first thing that sets your price.

2. Voice or text

Text is where most UK enquiries now start, and it is the cheaper half: a message tolerates a second of thought before the reply. A voice agent answering live calls is a harder problem - telephony, natural interruptions, zero patience for delay - and it prices accordingly, which is why voice is usually offered as an add-on rather than the starting point. The sequence that fits most businesses: cover the text channels first, then add voice if ringing phones are genuinely a leak. There is a halfway option too: missed-call rescue, where a call you could not take gets a text back within seconds and the AI books the caller in the thread - much of the value of voice at a fraction of the complexity.

3. Conversation volume

Every AI conversation consumes computing resource that somebody pays for, so volume shows up in every pricing model on the market: as tiers, as allowances, as usage rates. A sole trader fielding sixty enquiries a month and a clinic fielding two thousand should not pay the same, and honest pricing reflects that. When you compare quotes, ask what volume is included, what happens to the price in a busy month, and - more important - what happens to the answering. A system that slows down or stops exactly when demand spikes fails at the moment it was bought for. The good news: volume is the most transparent factor on this list, and the easiest to check against your own inbox before you buy.

4. What it plugs into

An AI that answers questions is useful; one that ends conversations with an appointment in your actual diary is the point. That means connecting your booking system or calendar, sometimes a CRM, sometimes reminder flows - and integration effort ranges from an afternoon to a project depending on what you run. The test to apply before signing anything: watch a booking land in the real calendar, with contact details captured before it is confirmed. An answered question without a booking is just a politer missed call.

5. Who trains it

The same engine gives a brilliant answer or a confident wrong one depending entirely on what it was taught, which makes training the most price-sensitive line on any quote. At the self-serve end, a guided wizard drafts the AI's knowledge from your website and you correct and approve it yourself: cheap in cash, real in hours. At the done-for-you end, a person sits with you for hours, extracts the answers that live only in your head, writes and tests the lot, and has you sign it off before launch. I unpacked the difference in what "trained on your business" actually means; it is the gap customers can actually feel when they message you.

6. Who keeps it sharp

The quiet factor, and over a year often the biggest. Prices move, staff change, services appear and vanish - and an unreviewed receptionist drifts from useful to merely present without ever visibly breaking. Someone has to read the conversations every month, fix the weak answers and add what changed. If that someone is you, the cost is your hours; if it is your provider, it is in the fee. A quote that says nothing about ongoing reviews is not cheaper. It has just moved the maintenance onto your desk without telling you.

Build, buy or DIY: the honest triangle

Those six factors get assembled in three different ways, and each corner of the triangle genuinely suits somebody. The corners do not really compete on price; they compete on where the work lands. Every route pays for training, testing and maintenance somewhere - in cash, in your own hours, or in mistakes - and the honest comparison is about which currency your business can best afford.

• Build it in-house. Full control, and occasionally right for a business with a developer on staff and unusual needs. For everyone else it is the most expensive corner in practice, because the build is the cheap part: models change, channels update their rules, the person who built it leaves, and the maintenance never ends. Most small businesses should not be running a software project in order to answer WhatsApp.
• Set it up yourself on a platform. The least cash and the most of your attention: you approve the knowledge, connect the channels, test the awkward questions and keep the monthly review habit. Right for simple service lists, tool-comfortable owners and low-stakes answers. Our self-serve platform opens this autumn for exactly this corner - and what you build there carries over if you later upgrade to done-for-you.
• Buy it done for you. The most cash and the least of your time: discovery, training, testing, sign-off and monthly reviews are the provider's job, with accountability to match. Right where every enquiry carries real money, where the field is regulated - clinics, dental, legal, finance - or where you already know the maintenance habit will not survive contact with your diary.

I keep a fuller decision guide in set it up yourself or have it done for you. The pattern that repeats: one strong reason - regulation, high-value enquiries - usually justifies done-for-you on its own, while self-serve earns its place when the whole picture is simple.

What "trained personally" buys, in practice

Since training splits the market's prices widest, it deserves concreteness. Personal training means a discovery session of two to four hours in which someone extracts your services, prices, tone of voice and booking rules, plus the awkward questions customers really ask. It means the system is tested against those questions before launch - the price nobody likes saying out loud, the 11pm rambling voice note, the complaint - and it means sign-off: you read the answers before any customer meets them, and a named person stays responsible afterwards. A setup wizard is honest work too, and surprisingly good at drafting from a website. What it cannot do is know what is missing from that website, and what is missing is usually the twenty answers that decide bookings. When you pay the higher price, you are not paying for a nicer wizard. You are paying for judgement, testing and someone to call when something needs to change by Friday.

The costs that never appear on an invoice

Three of them, and they decide more buying decisions than the visible fee. Your time is the first: a cheap subscription that quietly needs your evenings is not cheap, and a dearer service that needs two hours of discovery and a monthly check-in can cost the business less in real terms. Wrong answers are the second: a template bot quoting last year's price confidently at 9pm creates rework and mistrust no invoice ever itemises. And the third is the baseline everyone forgets - the cost of doing nothing. Enquiries answered within the first minute convert up to 391% better (Velocify, from an analysis of around 3.5 million leads); a reply inside the hour makes qualifying the lead about seven times more likely (Harvard Business Review); and 80% of patients say they want to book any time, from their phone (Experian Health, 2025). Whatever a receptionist costs, the fair comparison is against the enquiries currently going unanswered. A free strategy call puts your own numbers on that.

How to compare quotes like for like

Whoever you shortlist - us included - make every quote answer the same spec sheet: which channels, text or voice or both, what conversation volume is included, which calendar and booking connections, who does the training and to what standard, who reviews conversations after launch and how often, and what happens to your data if you leave. Cheap quotes usually go quiet on training and reviews, which is precisely where cheap goes wrong; I keep a nine-point version of this test in the honest comparison of UK AI receptionists, and it works on any provider. Then, and only then, put real numbers beside the factors. Ours are on the pricing page: the receptionist itself, voice as an add-on, the self-serve platform, and the Growth Partner plan for businesses that want the whole engine - capture, convert, rebook, grow - run by one accountable team. No figure appears in this article, deliberately. Every figure is one click away, kept current, which is the only place prices belong.