The Phone Still Rings. Who Answers It at UK Businesses Is Changing

Picture a joiner in Leeds halfway up a ladder when his phone buzzes in his pocket. He lets it ring out, and the caller, a homeowner with a kitchen job, rings the next firm on the list. An AI receptionist UK firms can now switch on would have picked up that call, taken the details and booked a site visit before the joiner climbed down.

So what does an AI receptionist actually do? In short, it answers your business line with a natural-sounding voice, handles routine requests such as bookings, opening hours and messages, and passes anything complex to a person. The technology is now cheap enough for small firms, but whether it helps depends on how well you design the handover, the scripts and the data behind it.

Why the phone still matters to UK businesses

Calling is in long-term decline, yet it remains huge. According to Ofcom’s telecoms market data for Q2 2026, UK mobiles still originated 35.2 billion minutes of calls in that quarter alone. Fixed lines carried 3.02 billion minutes, down 16.3% on a year earlier.

Behind those numbers sit millions of small operators. The Department for Business and Trade’s business population estimates for 2025 count 5.7 million private sector businesses, and 75% of them have no employees at all. In other words, for most firms the person answering the phone is also the person doing the work.

That tension shows up in missed calls. Research by the answering service Moneypenny, reported by Small Business, found that a third of the micro businesses it surveyed failed to answer incoming calls. Moreover, 69% of callers who reached voicemail hung up without leaving a message. Since Moneypenny sells call answering, treat those figures as useful but not neutral.

How an AI receptionist works, step by step

Under the bonnet, an AI receptionist chains together several services. First, a telephony provider routes the call to the software. Next, speech recognition turns the caller’s words into text, and a large language model decides what to say or do. Finally, a voice engine speaks the reply, usually in well under a second.

The useful part, however, is what happens around the conversation. A good set-up connects to your calendar, your CRM and your job system. As a result, it can check a free slot, create the booking and log the call against the right customer record. Without those connections, it is little more than a talking voicemail.

Most platforms follow a similar pattern on each call:

1. Greet and disclose. The agent answers in your business name and says it is an automated assistant.
2. Identify the caller. It matches the phone number against your CRM, or asks for a name and postcode.
3. Understand the request. It works out whether the caller wants a booking, a quote, an update or a complaint.
4. Act or collect. It books, answers from approved information, or records the details you need.
5. Hand over when needed. It transfers the call, or promises a callback and creates a task for a named person.
6. Write it down. It saves a summary and the outcome so nobody has to listen back to the recording.

If you want a deeper explanation of inbound and outbound use, our guide to how an AI call assistant handles calls covers the mechanics in more detail.

What an AI receptionist UK firms can buy today actually costs

Pricing for an AI receptionist UK businesses can deploy has fallen sharply, and much of it is now published. Take ElevenLabs, one of the better-known voice platforms. Its agents pricing page, checked on 15 October 2026, lists a Pro plan at $99 a month with 1,238 included minutes, and additional call minutes at $0.08. However, the page also notes separate usage charges for the language model. Telephony comes on top at cost.

Telephony is the cheaper layer. According to Twilio’s UK voice pricing, also checked in October 2026, a UK local number costs $3.50 a month. Receiving calls on that number costs $0.01 a minute, while calls out to UK mobiles cost $0.0305 a minute.

A worked example for a small firm

Picture an illustrative dental practice that takes 600 calls a month, averaging three minutes each. That adds up to 1,800 minutes. Here is how the published prices stack up, in US dollars as the vendors list them.

That figure excludes build time, testing and ongoing tuning, which often cost more than the minutes. Even so, it shows why the economics have shifted. A part-time human receptionist costs far more each month, but a person also brings judgement the software lacks.

The caller experience is the real test

Customers are wary, and the data says so plainly. A Gartner survey published in July 2024 found that 64% of customers would prefer companies did not use AI for customer service. Their top worry was that reaching a person would become harder.

That fear points straight at the design brief. Callers will tolerate an automated voice if it books them in quickly. On the other hand, they will not forgive one that traps them in a loop when their boiler has burst.

Service standards in the UK are also rising, which raises the bar. The Institute of Customer Service reported that its UK Customer Satisfaction Index for July 2026 stood at 78.3 out of 100, up one point on a year earlier. It also noted that ease of dealing with an organisation has grown more important to many customers.

Five design rules for an AI receptionist UK callers will accept

• Say it is AI at the start. Honesty early costs you nothing and avoids the backlash of a caller realising halfway through.
• Offer a person on request. Make “speak to someone” work at any point, not only after three failed attempts.
• Keep answers short. Phone callers cannot skim, so one fact per turn beats a paragraph read aloud.
• Escalate on emotion and risk. Complaints, safeguarding concerns and urgent faults should go to a human straight away.
• Close the loop. If the agent promises a callback, a named person must own it in your system.

Data protection and disclosure in the UK

An AI receptionist handles personal data from the first second of a call. Therefore UK GDPR applies in full: you need a lawful basis, a clear retention period for recordings and transcripts, and contracts with every processor in the chain.

The Information Commissioner’s Office is explicit on transparency. Its guidance on AI and data protection says you need to be transparent about how you process personal data in an AI system. Where you collect data directly from people, it says you should give that information at the point of collection.

In practice, that means a short spoken notice at the start of the call and a fuller privacy notice on your website. Furthermore, check where each vendor stores audio and transcripts, and whether any of that data trains their models. If you serve customers in the EU as well, the EU AI Act’s transparency rules for AI systems that interact with people, which apply from 2 August 2026, are also relevant.

Where an AI receptionist fits, and where it does not

The strongest cases share a pattern: high call volumes, repetitive requests and a clear next action. Think trades booking quotes, clinics managing appointments, letting agents logging repairs and restaurants taking reservations. In each case, the agent either completes a task or gathers exactly what a person needs to finish it.

By contrast, the weak cases involve judgement, negotiation or distress. A solicitor’s first call with a bereaved client, for example, is not a job for software. Equally, a firm with a dozen calls a week may get more value from a simple diversion to a colleague’s mobile.

Missed calls tend to be the clearest starting point. If your team already loses enquiries out of hours, our piece on why missed calls cost you customers sets out how to measure the leak before you fix it.

How to choose an AI receptionist for your UK business

When comparing any AI receptionist UK suppliers offer, start with your own calls, not a vendor demo. Listen to a week of recordings, or ask whoever answers the phone to tally the reasons people ring. Most firms find that four or five request types make up the bulk of the volume.

Then work through this checklist before you sign anything:

1. Integrations. Can it read and write to your calendar, CRM and job system, or only send you an email?
2. Handover options. Can it transfer live to a mobile, ring a group, or create a task with a full summary?
3. Voice and accents. Test it with real callers from your area, including older customers and strong regional accents.
4. Pricing clarity. Ask for the full cost per minute, including the language model, telephony and any platform fee.
5. Data handling. Confirm storage location, retention periods and whether your data trains anyone’s model.
6. Reporting. Check that you can see every call, its outcome and the reason for any transfer.
7. Ownership. Decide who in your team reviews transcripts weekly and updates the scripts.

That last point is the one most businesses skip. An AI receptionist is not a one-off install; prices change, menus change and customers ask new questions. Partners such as Riophany build the agent into the CRM and keep tuning it after launch, but you can also run that review in-house if someone owns it.

The firms that get this right

The most effective set-ups are surprisingly modest. They answer every call, solve the simple things, and hand the rest to a person who already knows who is calling and why. Nobody notices the technology much, which is rather the point.

Back on that ladder in Leeds, the joiner still wants to talk to the homeowner himself. The difference is that he now calls back knowing the address, the budget and a booked slot, rather than chasing a stranger who has already found someone else. The phone keeps ringing; the answer just arrives faster.

Frequently asked questions

What is an AI receptionist?

An AI receptionist is software that answers a business phone line using speech recognition, a language model and a synthetic voice. It can book appointments, answer common questions and take messages. Well-designed systems also transfer callers to a person when the request is complex or sensitive.

How much does an AI receptionist cost in the UK?

Costs depend on call volume and the platform. As of October 2026, ElevenLabs publishes a Pro plan at $99 a month with 1,238 minutes and $0.08 per extra minute, plus separate language model charges. Telephony through Twilio adds about $0.01 a minute for inbound UK local calls, while build and tuning work is extra.

Do I have to tell callers they are speaking to AI?

The ICO says organisations must be transparent about how they process personal data in an AI system, and give that information when they collect the data. In practice, a short spoken notice at the start of each call meets that expectation and builds trust. Firms serving EU customers should also check the EU AI Act’s transparency rules.

Will customers accept an AI receptionist?

Many customers are sceptical: a 2024 Gartner survey found 64% would prefer companies did not use AI for customer service. Their main concern was losing access to a person. Acceptance improves when the agent discloses itself, solves simple requests quickly and offers a human at any point.