Your phone rings when everyone is busy. A client is mid-treatment, the bar is three deep, or the office shut at half five. The call goes to voicemail, and plenty of people don’t leave one. They ring the next business on the list.
An AI receptionist answers those calls for you. Here’s what one does, where it should stop, and how to pick one that sounds like your business rather than a call centre.
What does an AI receptionist actually do?
An AI receptionist answers your phone or website chat, talks to people in plain speech, and handles the routine parts of a call without a person. A good one does far more than read out your opening hours. It can:
- answer common questions about your services, location, parking and hours
- take a booking or an enquiry and send it where it needs to go
- take a proper message with the caller’s name, number and reason for calling
- pass urgent or awkward calls to a person
- remember someone who has been in touch before, so they don’t have to repeat themselves
It works late at night, on a Sunday and while your team is with someone else. It doesn’t replace your front desk. It covers the gaps your front desk can’t.
What shouldn’t an AI receptionist do?
It shouldn’t pretend to be a person, and it shouldn’t make decisions that need a person’s judgement. That line matters more than any feature.
If you run a clinic, an aesthetics business or a dental practice, the assistant should never give medical advice. Any clinical question goes straight to a member of your team. The same goes for anything legal, financial, a complaint, or a situation where a wrong answer could cause harm.
It also shouldn’t make things up. When it doesn’t know, the right answer is “I’m not sure, let me take your details and someone will call you back.” An assistant that guesses confidently is worse than voicemail.
Should it tell callers it’s an AI?
Yes. In my view it should say so early and plainly. People are far more forgiving of an assistant that is honest about what it is. For health businesses I’d treat this as non-negotiable.
Being upfront also sets expectations. The caller knows they can ask for a person, and they know a person will follow up on anything the assistant can’t handle.
How do I choose an AI receptionist?
Choose the one that knows your business, sounds like it, and hands off cleanly when it should. Before you commit to anyone, ask these questions:
- Can I hear it answer as my business before I sign anything?
- What does it do when it doesn’t know the answer?
- Where do bookings and messages go: my inbox, my booking system, my phone?
- Does it remember returning callers, and where is that stored?
- Who updates what it says when my hours or services change?
- Can I read back what it said on every call?
Here’s how I’d read the answers:
| Good sign | Warning sign |
|---|---|
| You hear a demo in your own business’s name | You only hear a generic sales demo |
| It takes a message when unsure | It always has an answer |
| Messages land where your team already works | You need to log in somewhere new to find them |
| You can read every conversation | You can’t see what it said |
| It says it’s an AI | It’s designed to sound human and hide it |
When I built Zeena, I gave her a voice people talk to, a web page, a workflow engine with 43 automatic steps (one run fires 39), a database that remembers every person, and her own server. On my site she plays a made-up clinic’s receptionist, and you can hear her answer as your own business. That’s the test I’d want any owner to run before deciding.
What should I check before switching one on?
Check how it handles people’s data, then test it yourself as a difficult caller. An AI receptionist hears names, phone numbers and sometimes sensitive details, so ask any supplier where that information is stored, who can see it and how long it’s kept. Check your obligations under UK GDPR with the ICO. If you’re regulated, by the CQC, GDC or GPhC for example, check with them too.
Then ring it yourself before your customers do:
- Call as a nervous first-time customer and see if it’s patient.
- Call as someone annoyed and see if it offers a person.
- Ask something it shouldn’t answer and check it refuses politely.
- Ask something odd and check it takes a message rather than guessing.
For the first week, read every conversation. You’ll spot gaps quickly, and fixing them early is what turns a decent assistant into one your callers trust.
If you’d like to see how I approach this, my AI receptionist page explains what I build and how it fits around your team.
If you want it done for you
I’ll design a free preview of your new homepage, with a short video of me walking you through it, within two working days. Ask for your preview here, or just tell me which calls you’re missing and what you need answering, and I’ll tell you honestly whether an AI receptionist is the right fix.