Using AI to answer enquiries and take bookings

An AI receptionist for a small business answers calls, texts and web chats when nobody is free, then either resolves the enquiry or books it into your calendar.

In practice most of the value sits in three small jobs: the question you get asked twenty times a week, the call that rings out while you are with a customer, and the booking request that arrives at 10pm. Get those right and the rest can wait.

What an AI receptionist does for an Irish small business

It handles the first reply, so every enquiry gets an answer within a minute or two, whatever time it lands. The simple version answers the questions you hear every day: opening hours, whether you take walk-ins, how far you travel, whether you are booked out this week. A fuller version checks your calendar and offers a real appointment, takes a deposit, or passes on a message with enough detail that you are not ringing a stranger back cold.

It is also not one product. There are four common shapes, and which one you need depends on where you actually lose enquiries:

  • Web chat. A box on your website that answers questions and takes booking requests.
  • Missed-call text-back. Your phone system texts anyone whose call you didn’t pick up, with a link to book or an offer to ring back.
  • A voice assistant. Software that answers the phone, talks to the caller, and books them in or takes a message.
  • Inbox and form replies. Emails and contact-form messages get a quick, sensible acknowledgement, and a short summary lands with you.

Why do small businesses lose enquiries in the first place?

Because the people who answer the phone are the same people doing the work. Say you run a small salon with two staff on the floor most of the day. A phone that rings during a cut goes to voicemail, and plenty of callers ring off rather than leave a message. A web enquiry sent at 10pm sits unread until the next afternoon, by which point the person may have booked somewhere else. Both are simply what happens when a small team is busy doing the job it is paid for.

The businesses that feel it hardest get their enquiries outside working hours: trades who get called once a customer is home in the evening, cafés and restaurants fielding bookings on a Saturday morning, clinics where people ring on their lunch break. The trades version, where the evening call and the unchased quote are one problem, is in AI for tradespeople.

How does an AI booking assistant handle a real enquiry?

An AI booking assistant asks the two or three questions you would ask yourself, then either books the slot or hands over to a person. Here is how that might run for a hypothetical physio practice in Douglas:

  1. A call comes in at 7.40pm. The assistant answers, says it is the practice’s automated assistant, and asks what the caller needs.
  2. The caller wants a first appointment for a sore back. The assistant asks the set questions: new or returning, and any preferred day.
  3. It offers two real slots from the diary. The caller picks Thursday at 10, and a confirmation text goes out straight away.
  4. A summary lands in the practice inbox, so whoever opens up in the morning knows who is coming and why.

For anything outside its script, a good setup says so and hands off: “I can’t answer that one, but I’ll have someone ring you in the morning.” That handoff matters more than any other part of the setup. An assistant that guesses at a price it doesn’t know, or promises a slot that doesn’t exist, does more damage than the missed call would have.

AI for missed calls: the cheapest place to start

For most small businesses the best first step is a missed-call text-back, which is simpler and cheaper than a voice assistant. Many business phone systems and VoIP services can already send an automatic text to anyone whose call goes unanswered: “Sorry we missed you, here’s a link to book, or reply and we’ll ring you back.” That catches a caller who would otherwise try the next name on the list. Check what your current provider offers before you buy anything new.

After a month, count the texts that went out. That number is how many calls you were missing, and it is a fair test of whether a fuller assistant is worth the bother.

Where does it go wrong, and what should stay human?

Anything sensitive, complicated or emotional should reach a person. A regular customer asking about a discount, a complaint, a first-time patient with a medical question before treatment: these need judgement an automated assistant does not have, and forcing them through one annoys exactly the customers you most want to keep.

Tone is the other trap. A script that sounds robotic, or repeats the same line whatever the caller said, tells people nobody set it up properly. Listen back to real calls or read real chat transcripts after the first fortnight, and fix the three or four things that keep tripping it up.

Two rules are worth settling before it goes live:

  • Say it is automated. The EU AI Act’s transparency rules expect people to be told when they are dealing with an AI system, unless that is already obvious. A one-line greeting covers it. I wrote more on this in should you tell customers you use AI?
  • Know what it keeps. Calls and chats carry names, numbers and sometimes health or account details. Find out where the tool stores recordings and transcripts, for how long, and whether callers are told. This is general information, not legal advice, and is it safe to put client data into AI tools? goes further.

Is an AI receptionist worth it for a small business?

It is worth it when you can point to a specific, recurring leak: calls after 6pm that go unanswered, weekend enquiries nobody reads until Monday, a booking form that sits in an inbox for three days. It is not worth it as a general “modernise the front desk” project with no target, because that is how a business ends up paying monthly for a system nobody configured and nobody trusts.

These tools are usually billed by the month, and voice assistants often by the minute of call time as well, so a quiet line can cost more per booking than it brings in. The fuller cost picture is in how much AI automation costs a small business in Ireland.

If you are not sure the gap is big enough, the AI Assessment answers that before you sign up to anything: it looks at how enquiries actually reach you across a normal week and tells you which channel to fix first, or whether to leave it alone. Start with the one channel that loses you the most, get it working properly, then add a second. A missed-call text-back that works beats a voice assistant that half works, every time.

Book the AI Assessment

The assessment is credited in full if you go on to a build, and if I can’t find you five hours a week you pay nothing. See how it works.