For many small businesses the phone is still where the money arrives. A plumber, a clinic, a letting agent or a law firm can lose a job simply because nobody picked up while the team was on site, with a patient or in a meeting. Callers rarely leave voicemails any more; they ring the next name on the list.
The scale of the problem is now being measured. In October 2026 BT Business said UK small businesses lose an estimated £3.7 billion a year to missed calls, and that two-thirds of small firms miss at least one call a day that could have turned into work. BT also sells an answering product, so treat the headline figure with some caution, but most owners do not need a survey to know the pattern.
That is why AI receptionists, voice agents that answer the phone, talk naturally and take action, have moved from novelty to a common question. This guide covers what they can realistically do, how off-the-shelf tools compare with a custom build, and what to check before you switch one on.
What an AI receptionist can do well
A modern voice agent listens, understands what the caller wants, replies in a natural voice and can use your systems during the call. Done properly, it handles the repetitive majority of calls:
- Answering at any hour. Evenings, weekends, lunchtimes and the five minutes when everyone is busy.
- Booking and rescheduling. Checking real availability in your calendar or booking system and confirming the slot.
- Answering common questions. Opening hours, prices you are happy to quote, service areas, parking, what to bring.
- Qualifying enquiries. Asking the questions a receptionist would ask, such as postcode, type of job or urgency, and logging the answers.
- Routing and handover. Transferring urgent calls to a person, or taking a message and alerting the right team member straight away.
- Writing everything down. A summary of each call in your CRM or inbox, so nothing depends on someone remembering.
What it should not be trusted with
The technology is good, not magic. Keep a person in the loop for:
- complaints, distressed callers and anything sensitive
- negotiation, discounts or commitments you would not put in writing
- medical, legal or financial advice
- taking card details over the phone, unless the payment flow is built and certified for it
A good AI receptionist knows its limits. The most important feature is a fast, reliable route to a human when the caller asks or the conversation goes off script.
Off-the-shelf tools: quick to start, limited to their templates
Plenty of products now offer an AI receptionist on a subscription. You sign up, connect a phone number, describe your business, and it starts answering. Pricing is usually per call, per minute or a flat monthly plan, and entry plans published by vendors tend to sit in the tens of US dollars a month, with extra charges once you pass the included calls or minutes. Check current pricing directly, as it changes often.
Off-the-shelf is often the right first step when:
- your calls are simple, mainly bookings and FAQs
- you already use a booking tool or phone system the product integrates with
- you want to test whether callers accept an AI voice before investing more
The limits show up as you grow. You are tied to the integrations the vendor supports, the call logic is whatever their settings allow, and your data, recordings and transcripts live on their platform. Per-minute pricing can also climb quickly for busy lines.
A custom voice agent: built around how you work
A custom agent is built on a voice AI platform (for example Vapi, Retell or similar) and connected to your own systems by a development team. It costs more up front, but it fits your process rather than the other way round. It usually makes sense when:
- Calls need your data. Looking up an order, a job status, a policy or a patient's next appointment in your own database or CRM.
- The logic is specific. Different questions for different services, postcode-based pricing, triage rules, multiple locations or languages.
- Calls should trigger work. Creating a job, sending a quote, updating a deal or starting a follow-up sequence once the call ends, often through an automation tool such as n8n. Our guide to n8n automation for small businesses covers that side.
- Volume is high. At scale, paying the voice platform and model costs directly can work out cheaper than a per-call subscription.
- You want control. Your own prompts, your own data retention rules, and the option to change providers later.
One example from our own work: for Brixven, an agency serving small businesses across the UK and Ireland, we built a Vapi voice agent that answers inbound calls and books appointments, connected to n8n workflows that handle email follow-up and CRM updates.
The legal and trust checks to do first
Rules differ by country and change over time, so take advice for your own situation, but these are the points every business should check:
- Tell callers they are speaking to an AI. It is good practice everywhere and increasingly expected by regulators. A short line in the greeting is enough.
- Call recording and transcripts. In the UK, recordings and transcripts are personal data under UK GDPR, so you need a lawful basis, a clear notice to callers and a sensible retention period. In the US, several states, including California, require every party's consent to record a call, so say clearly at the start that calls are recorded.
- Outbound calls are different. In the US, the FCC ruled in 2024 that AI-generated voices count as "artificial" voices under the Telephone Consumer Protection Act, so automated outbound calls need the same prior consent as robocalls. An inbound receptionist avoids most of this, but keep it in mind before you let the agent call people back.
- Where data is processed. Ask which providers handle the audio and transcripts, and where. Your privacy notice should reflect it.
How to roll one out without upsetting customers
- Pick one call type. Start with bookings or after-hours calls, not everything at once. The playbook in voice AI for sales follows the same principle.
- Write the script and the exits. List the questions, the answers you are happy to give and every situation that must go to a person.
- Connect the systems. Calendar, CRM and notifications. Without integrations, the agent is just a polite voicemail.
- Test with real calls. Have staff and friendly customers call in with awkward accents, background noise and unusual requests.
- Launch on overflow first. Let the agent take calls your team cannot answer, then widen its role once the results look right.
- Review every week. Listen to a sample of calls, read the transcripts and fix the gaps. Track answered calls, bookings made and handovers, not just call minutes.
Which route is right for you?
If your calls are simple and you want to start this month, try a reputable off-the-shelf tool and measure the results. If your calls depend on your own data, follow detailed rules or need to trigger work in other systems, a custom voice agent will serve you better and usually pays back faster as volume grows.
We design and build voice agents and the automations around them for businesses in the UK, US and Gulf. You can see what that involves on our AI solutions page.
If you would like a second opinion on which approach suits your calls, tell us how your phone line works today and we will give you a straight answer.