We sell an AI receptionist, and I still wouldn't put one on every coach's phone. Several things can go wrong, and one of them is far more serious here than in most small businesses.
A mistake on most business lines means a wrong address or a missed booking. On a coach's line, the caller at 11pm might have reached the end of their rope. So start with that one, and decide if you want a machine on your phone at all.
Mind the evidence. I found no independent test of AI receptionists on UK coaching phone lines, and nothing on what they do with a distressed caller. Software sellers fund much of the survey work on AI customer service, as the table further down shows. So this is the wider evidence, the official guidance, and my judgement.
The short version
- A caller in distress is the safety case. An AI can't assess risk. It needs a short script that points to 999 and Samaritans, approved by a qualified person and tested.
- It must never promise results or suggest coaching can treat anything.
- It must never confirm that someone's a client.
- Give current clients a direct number that a person answers.
- Tell callers it's an AI, and tell them if you record. Be careful what ends up in the summary.
Problem 1: A caller in distress
My guess is that people who ring a life coach aren't always looking for coaching. Some may be lonely, or frightened, or in crisis, and yours was the number they found. A business coach could get the same call after a redundancy. Your existing clients could ring as well.
An AI receptionist can't tell how serious this is. On a number the AI answers, ours only knows a call is urgent if the caller says so, and I'd expect that to be true of most systems. A caller may say something oblique. "I can't do this any more" could be about a job or about their life. No script catches everything.
So don't design this to assess anyone. Design it to do one plain thing every time: if a caller mentions suicide, self-harm, wanting to die or being in danger, give the official numbers and tell you.
The plain step, not a finished script
"I'm an automated assistant and I can't help with this. If you're in danger, or you might hurt yourself, ring 999 now. Samaritans are free, day or night, on 116 123. I'm trying to reach [Name] now."
Your business needs its own script, approved by a qualified person. These lines are only the official numbers put into a sentence.
Here's what the official guidance says:
- The NHS says to call 999 or go to A&E if you've seriously harmed yourself or may be about to. It lists 111, Samaritans on 116 123, and Shout by texting SHOUT to 85258, day and night.
- Samaritans say a call to 999 is the quickest way to get help for someone in immediate danger, and that they can call an ambulance for you if you'd prefer.
- In Scotland, NHS 24 runs a 111 mental health hub, 24 hours a day, every day. Breathing Space is on 0800 83 85 87, Monday to Thursday 6pm to 2am, and 6pm Friday to 6am Monday.
- In Northern Ireland, Lifeline is on 0808 808 8000, 24 hours a day.
- Under-19s can ring Childline on 0800 1111.
In my judgement, what the AI leaves out matters just as much. It shouldn't counsel anyone, or keep them talking to fill in a form, or promise you'll ring back soon unless that's true.
With ours, you decide what goes into that wording, and we make the changes. It isn't something you can assume is switched on. You must write it, have it approved by a qualified person, and ring the number yourself before it goes live, using the words a frightened person would use. Test again every few months.
Then decide what happens next. Ours tries to transfer an urgent call to you. If you can't pick up, it takes details and puts them in your summary. So what if you're in a session, or asleep? You need to read your summaries often, and have a plan for what you'll do with one that worries you. If you can't do that, don't put an AI on your number. Use a recorded greeting that gives the numbers instead.
Who shouldn't use one
If your work is close to mental health, trauma, bereavement or recovery, I wouldn't put an AI receptionist on your main line. That's my opinion. A human answering service, or a clear recorded message with the support numbers, is safer.
Problem 2: It breaks a confidence
Coaches work on trust, and an AI doesn't know what's private.
- A manager rings and says, "Is Sarah one of yours?" A person would know not to answer. The AI needs telling. It should say it can't confirm anything about who you work with, and take a message.
- A sponsor asks how the coachee's getting on. Same answer.
- A caller leaves deeply personal details, and those go into a summary and sit in a contact record.
- A partner picks up your phone and sees a summary.
How to reduce it:
- Tell it never to confirm or deny that anyone's a client.
- Don't load client names into it.
- Have it ask only for a name, a number and a one-line reason.
- Don't record calls unless you need to.
- Lock your phone, and keep notifications off the lock screen.
Under UK GDPR, details of someone's mental or physical health are special category data. If callers tell your AI these things, you're holding that data. Say so in your privacy notice, and tell clients in your contract how you keep it. Don't assume your provider's systems suit your duty of confidence. Ask. The Global Code of Ethics also asks coaches to take dilemmas from AI and technology to supervision, so do that.
Problem 3: It promises something you can't deliver
An AI can state something false with total confidence, and you can't necessarily blame the machine. Air Canada tried that in Moffatt v Air Canada, and a Canadian tribunal made the airline answer for wrong information from its website chatbot. No UK court is bound by that. Even so, I wouldn't want to answer a coaching complaint with "the AI said it".
The risky areas for a coach are:
- Results. "Our clients double their income." The ASA ruled against claims like this in October 2025.
- Health. "Yes, coaching can help your anxiety." CAP advice says treatment of conditions such as depression or OCD needs suitably qualified medical supervision. Don't let an AI say it down the phone.
- Fees. "Coaching's £100 a session" is a promise you didn't make.
- Credentials. "Yes, she's a qualified psychologist."
- Availability. "You can start tomorrow."
How to reduce it:
- Give it a written list of what it may say about fees, usually a range.
- Tell it what you don't do, in plain terms.
- Ban all advice. If someone asks if coaching will help, it says you'd be glad to talk it through on a discovery call.
- Have it check your credentials against the exact wording on your website.
- Read the summaries each week, and correct anything that's off.
Problem 4: Some callers don't want to talk to AI
Coaching's a relationship, and the first call is part of it. A person ringing a coach may be more put off by a machine than a person ringing a garage. That's my guess, but the wider evidence points the same way.
| Survey | Finding | Who paid |
|---|---|---|
| Gartner, 5,728 customers, December 2023 | 64% would prefer companies didn't use AI in customer service | Independent research firm, countries not stated |
| YouGov for Pega, 2,222 UK adults, November 2025 | 50% rarely or never get a successful result from AI-only contact | Pega, a software firm |
| Five9, 1,000 UK adults, 2026 | 76% say being told it's AI is very or extremely important | Five9, a software firm |
The fair comparison isn't AI against you. It's AI against a voicemail, and in the YouGov survey for Answer4u, only 6% of UK adults said they'd leave one. Even so, for someone choosing a coach, a warm text from you might work better than either.
What helps:
- Open by saying it's an AI.
- Keep the greeting short.
- Offer a clear route to you: "I'll take your details and Kate will ring you back by 4."
- Read the first 20 summaries, and call back quickly. The call-back is where the relationship starts.
Problem 5: It mishears names and details
I found no independent figure for AI receptionists mishearing callers. There's a general warning, though. In a 2025 preprint, researchers testing speech recognition on Newcastle English recorded a word error rate of about 32% for one system's UK setting, and cited work finding higher error rates in Northern England and Northern Ireland than in Southern British English. That study used recorded speech, not phone calls, so treat it as a reason to test.
A caller who's upset, or crying, will be harder still. Have it read back the name and number, and always capture the caller's number.
Problem 6: Privacy and call recording
The ICO's guidance says you must tell callers if you're recording a call and why. It's filed under worker monitoring and it's under review. I'd follow it regardless. Other details can go in a link or on your website.
Providers differ. Fasthosts says its AI receptionist doesn't record calls and keeps only a summary and basic details. For a coach, that's a real advantage. Ours can record calls, and the greeting says so. Ask any provider what's recorded, how long it's kept, and if it trains their models.
As far as I could find, no UK rule makes a business announce an AI on an inbound call. I couldn't check the very latest position, so announce it regardless. I wouldn't use an AI to cold call either. Outbound marketing calls have their own rules, so take advice before you try it.
Problem 7: Sales calls run up the bill
A published coaching number attracts sales calls, I'd expect, and on a usage-priced service those can count. Ask how a call's measured, if spam is billed, and what you pay beyond the allowance. Ours comes with a monthly usage credit, and anything above it is added to your bill.
Problem 8: It takes details, and then nobody acts
This is the most common failure I'd expect. A receptionist that captures five enquiries is worthless if you don't ring them back. Set a daily time for call-backs, and fix the rest of the chain, as set out in why coaches lose clients after the enquiry.
What ours does
To compare against anyone else's:
- It says it's an AI at the start of the call.
- It answers from what you've told it about your services, areas, fees and hours.
- It takes the caller's details, and you get a summary.
- On its own, it doesn't book. You call back and book. The full AI Business Engine adds booking.
- It can record calls, and the greeting says so.
- You can have a separate direct number for current clients that comes straight to you, and an enquiry number the AI answers.
- On a number the AI answers, it only knows a call is urgent if the caller says so. Then it tries to transfer the call to you. If you can't pick up, it takes details and puts them in your summary.
- For someone in distress, it only says what you've set. You must set that wording, get it approved and test it. Don't assume it.
No receptionist, ours included, takes these risks away. Whether one's worth having at all is a separate question, and I've answered it here.
Questions coaches ask
Is it safe for a life coach to use an AI receptionist?
It depends on your clients. For a career or business coach with few distressed callers and a tested script, I think it can be. If your work is close to mental health, I wouldn't use one on your main line.
What should an AI receptionist say to a suicidal caller?
It should say it can't help, then give 999 and Samaritans on 116 123, and try to reach you. It shouldn't try to counsel anyone. Your own wording must be approved by a qualified person, and tested. This is general information, not clinical advice.
Is there a legal duty to say it's an AI?
Not that I could find for inbound calls, though I couldn't check the very latest position and this isn't legal advice. I'd say so regardless.
Can I use it for existing clients?
I'd give them a direct number instead. They've trusted you, not a system. Use the AI for new enquiries only.
What happens if it tells someone coaching will cure their anxiety?
Treat it as your responsibility. Call them, correct it and fix the script. Then read your summaries more often. To avoid it, ban all health claims and have it offer a discovery call instead.
Sources
- NHS, help for suicidal thoughts (reviewed 26 April 2024)
- Samaritans, if someone you know is suicidal
- Samaritans, when you should let someone else help
- NHS 24, support for your mental health
- nidirect, Lifeline
- Global Code of Ethics, version 4 (December 2025)
- UK GDPR, Article 9
- McCarthy Tétrault, Moffatt v Air Canada (19 February 2024)
- Osborne Clarke, ASA rulings on earnings claims (19 December 2025)
- CAP, mental health treatment claims (15 July 2021)
- Gartner, customers and AI in service (9 July 2024)
- YouGov and Pega, reported by Compare the Cloud (25 February 2026, vendor)
- Five9 survey, reported by IT Brief (26 October 2026, vendor)
- Speech recognition on Newcastle English (preprint, June 2025)
- ICO, monitoring workers (under review)
- Fasthosts, AI receptionist cost (9 July 2026, vendor)
- YouGov for Answer4u, June 2026, 2,046 UK adults
Prices and rules change. Check the source before you rely on a figure. This is general information, not legal, tax or financial advice.
