AI Receptionist UK: What Should a Small Business Check First?

Hannah runs a three-person lettings agency in Leeds. Tenants ring about a leaking tap at seven in the evening, landlords ring about viewings on Sunday, and applicants ring at lunchtime when everyone is out doing viewings. The calls she misses are the ones that cost her business. When she asked around, three different people told her to try an AI receptionist.

If you run a small UK business and you are weighing the same thing, the product matters less than a handful of rules that apply here and are glossed over in most American-written guides. This is what to check.

Here is what matters most:

• UK GDPR applies to every call your receptionist handles, because a caller's name, number and reason for calling is personal data. The provider is acting on your behalf, and you remain responsible.
• Answering calls people make to you and ringing people to sell to them are governed by different rules. Automated marketing calls need the person's specific consent under PECR.
• Tell callers when a call is recorded, and tell them early.
• Insist on a written contract covering how the provider uses and stores your call data.
• Keep your own number, so you can change provider without changing your letterhead and website.

What an AI receptionist does for a UK small business

An AI receptionist, also sold as an AI answering service or AI phone answering, answers your phone in a natural voice, deals with the routine questions, books appointments into your diary and sends you a summary of each call. It is not the same as a voicemail box, which waits for the caller to talk to nobody, and it is not the same as a traditional virtual receptionist service, where real people answer in your company's name.

In the UK it tends to be bought by the trades and by small professional practices: electricians, plumbers and heating engineers, letting and estate agents, solicitors and accountants in busy periods, private dentists and physiotherapists, salons, vets, driving instructors. The reason is the same everywhere. These are businesses where the owner is often busy doing the job, the caller wants something quickly, and a missed call goes to the next name in the search results.

The question most owners ask is whether it works. It works for the predictable calls: opening hours, whether you cover an area, what a visit costs, booking a slot, taking details for a callback. It works less well for the unpredictable ones: an upset customer, a complicated dispute, a caller who cannot say what they want. A sensible setup lets it handle the first group and passes the second to you straight away.

Take a heating engineer with a van and no office. A customer rings at half past six on a winter evening and says the boiler has stopped. A good AI receptionist asks the few things the engineer needs to know: the address, the make of boiler if the caller knows it, whether there is any smell of gas, whether anyone vulnerable is in the house. It then either sends the engineer an urgent text and a call, or, for a smell of gas, tells the caller to ring the National Gas Emergency Service. That is a very different outcome from a voicemail that is checked at eight the next morning.

Two kinds of call, two sets of rules

The most useful distinction is between calls coming in and calls going out.

Inbound calls are the ones your customers make to you. An AI receptionist answering them is handling personal information, so UK GDPR is the main framework: you need a lawful reason to collect the details, you should tell people what you are doing with them, and you should keep them only as long as you need to. A short spoken notice at the start of the call, backed by a privacy notice on your website, covers most of this for an ordinary small business. If your business handles personal data, you will also usually need to be registered with the Information Commissioner's Office and pay its data protection fee, unless an exemption applies.

Outbound calls are the ones you make to them. If you want an AI to ring leads, chase quotes or sell, the Privacy and Electronic Communications Regulations (PECR) come into play. The Information Commissioner's Office, the UK's data regulator, is clear that automated marketing calls need specific consent, and that general marketing consent, or even consent to live calls, is not enough. Automated calls must also give your name and a contact address or freephone number. Live marketing calls have their own rules, including screening against the Telephone Preference Service (TPS) and the Corporate TPS for business numbers.

Keep the two uses apart in your planning. A receptionist is a low-risk starting point. An AI that rings people unprompted is a separate project with a separate legal footing, and it is worth speaking to a specialist before you start.

Recording and telling people it is an AI

If calls are recorded, say so at the start of the call. Do not bury it in a privacy policy nobody reads. The usual approach is a short opening line such as "this call may be recorded, and you are speaking with an automated assistant for the agency". It tells callers what is happening and gives them a chance to ask for a person.

On whether you must say it is an AI, the sources I read are divided, and the position is still developing. Some commentary links disclosure to the UK GDPR requirement to be fair and transparent with people. I would not rely on silence being allowed. Say it. It costs nothing, and callers who discover afterwards that they were talking to a machine tend to feel tricked, which is the outcome you are trying to avoid.

You can hear what a voice agent sounds like in the live demo below. It is one of our own demo agents, and the salon one speaks British English. Talk to it as a customer would, and notice how it handles interruptions and changes of mind.

Questions to ask the provider

Treat this like any supplier who will hear your customers' details, because that is what it is.

• Who is responsible for the data? The provider will usually be your data processor. UK GDPR expects a written contract setting out what they may do with the data, how they secure it and when they delete it.
• Where are recordings and transcripts stored? Ask whether data leaves the UK, and what safeguards apply if it does.
• Does it train on my calls? Some providers use customer calls to improve their own models. That may or may not suit you, and you should know which.
• Can I get my data out and deleted? You should be able to export call records and have them removed when you leave.
• How is pricing counted? Per-minute plans can be expensive when spam and sales calls use up your allowance, so compare on cost per useful call.

Does it cope with how people actually speak?

Test it with regional accents as well as clear southern English. A caller from Glasgow, Newcastle or the Black Country will challenge a system that has mostly heard a news-reader voice. Ring it from your mobile on a bad line. Give it a postcode, which is an awkward thing to hear and repeat, and a UK mobile number said quickly. If it keeps asking callers to repeat themselves, they will give up.

Also decide what happens when it cannot help. For Hannah, an emergency repair report in the evening, such as no heating in winter or a water leak, should go straight to her phone or her out-of-hours contractor. Write down the triggers, and test them before you rely on them. The aim is that the receptionist takes the routine calls and the person takes the urgent ones, so the system works alongside you, not instead of you.

Your number, and what it costs

Keep your existing number. Whether it is a 01 or 02 geographic number, an 03 number, a freephone 0800 line or a mobile, you should be able to forward it to the service and switch provider later without changing what is printed on your van, website and letters. Be wary of any provider that gives you a new number it controls.

On cost, prices vary and change, so I will not quote them. Work it out yourself. Count the calls you miss in a typical week from your phone's call log. Estimate what one new customer is worth and what share of missed callers would have become customers. Multiply to get a rough weekly cost of missed calls, then compare it with the monthly fee, checking how spam and sales calls count against any allowance. If the sums are close, try a month and listen to the recordings before committing to a contract.

If you are still deciding between people and software, our comparison of an AI receptionist, an answering service and a virtual receptionist covers the trade-offs.

When an off-the-shelf tool is enough

If your calls are mostly the same handful of questions and bookings, a ready-made AI receptionist is a reasonable buy. Try a month, listen to the recordings, and keep it if it earns its cost. If you want to build your own with a do-it-yourself tool, you can, and the setup is mostly writing down your answers, opening hours and emergency rules. The hard parts are the ones you cannot see: reliability when callers interrupt, handover to a person, and keeping the data in order.

It stops being enough when the call has to do something inside your systems. A letting agent who needs a repair logged against the right property and tenant, a practice that must check a client record before saying anything, a firm that wants a call turned straight into a job in its own software. At that point the conversation is the easy bit and the connection to your systems is the real work. That is what we build in our AI voice agent development service.

FAQ

Is an AI receptionist legal in the UK?

Using AI to answer incoming calls is lawful, provided you handle callers' personal data in line with UK GDPR, tell them if calls are recorded, and have a proper contract with the provider. The stricter rules arrive when an AI rings people to market to them, where PECR requires specific consent for automated calls. If you are unsure about your own situation, a short conversation with a data protection adviser is cheap protection.

Does an AI receptionist actually work?

For routine calls, yes: it answers instantly, never misses a call because it is busy, and books appointments from your diary. It works less well for upset callers, unusual requests and poor phone lines. The way to find out for your business is a one-month trial with the recordings reviewed every week, not a demonstration.

Which AI receptionist service is best in the UK?

I would not give you a ranking, because the right one depends on your calls. Compare two or three on the same set of test calls, and judge them on how they handle regional accents, whether they book into your own diary, how they hand over an emergency, where they store recordings and whether you keep your number if you leave.

How much does an AI receptionist cost?

Plans are charged per minute, per call, as a flat monthly fee or with bundled allowances, and they change often. The number that matters is cost per useful call: take a month of your own call history and apply each plan to it. Check what happens when you go over an allowance and whether spam calls count.

Is there a free AI receptionist?

Some providers offer a free trial or a free tier with limits on minutes or features, usually without a link to your diary or records. That is useful for testing and rarely enough to run on. Check what happens when you reach the limit and whether you can still listen to recordings.

Do I have to tell callers it is an AI?

I would, even though the legal position is still being worked out. Being open fits the UK GDPR principle of fairness and transparency, and it avoids the damage of a caller feeling deceived. A single sentence at the start of the call is enough.

Is an AI receptionist better than an answering service?

It depends on your calls. An answering service uses people who take messages and can judge urgency. An AI receptionist answers instantly, can book appointments into your diary and handles several calls at once, but it is weaker with unusual or distressed callers. Many small businesses choose one, then add the other for out-of-hours emergencies.