Key takeaways
- Most UK adults now know what AI is and many use chatbots regularly, so an AI voice is no longer a shock.
- Acceptance varies: older callers use voice assistants far less, and many people worry about decisions made without humans.
- Callers mind being misled or stuck far more than they mind the technology itself.
- Say it is automated at the start, keep it short, read details back and offer a person.
- Measure acceptance with hang-ups, requests for a person and complaints, not with guesses.
What the surveys say
Familiarity has grown fast. In a 2024 government tracker survey, six in ten UK adults reported using AI chatbots in the past three months, 44% used them at least once a month for personal or work reasons and 23% at least once a week[1]. Personal monthly use rose from 34% in the previous wave to 40%[1], and 71% of adults said they could at least partially explain what AI is[1].
Familiarity is not the same as enthusiasm. Around four in ten expect AI to have a positive impact on society and on them personally, while three in ten expect a negative one[1]. Two in five adults foresee a negative impact of AI on job opportunities for themselves and their families[1]. Some callers will be curious, some indifferent and some uneasy, and a heating firm's phone line hears from all of them.
Who is most likely to mind
Age matters. Even before the recent wave of chatbots, use of voice assistants varied widely: in 2020, 49% of adults aged 25 to 34 used a virtual assistant speaker or app, against 17% of those aged 65 and over[2]. Older people are no strangers to digital services, with 65% of over-65s shopping online in the same survey[2], but they have less practice talking to machines.
- Older homeowners and tenants, who make up a large share of breakdown and safety check calls.
- Callers in distress: no heating with a baby in the house, water coming through a ceiling, a smell of gas.
- People with hearing loss, speech differences or strong regional accents, if the system copes badly.
- Regular customers who are used to speaking to the same person in your office.
- Callers who have had a bad experience with a trader before and are already wary.
Trust in home repair firms is fragile anyway. In a government consumer study, 25% of adults had bought home maintenance and repair services in the past year, and 11% of those had a consumer problem[3]. A phone experience that feels evasive adds to that wariness; one that feels efficient and honest reduces it.
What callers actually object to
Complaints about automated phone systems are rarely about the technology itself. They are about being misled, being stuck, or being made to repeat everything later. People want to know who they are talking to, to be understood the first time, and to be confident that something will happen.
- Not being told it is automated, then realising halfway through.
- Long menus and repeated questions before they can explain the problem.
- Being misheard on the address or phone number, with no read-back.
- No way to reach a person when the situation is unusual or upsetting.
- Hearing nothing afterwards, and having to ring again.
Design choices that keep trust
Say it is automated, first
One short sentence at the start: the firm's name, that this is an automated assistant, and that calls are recorded.
Let callers talk
An open question such as “What's happening with your heating?” works better than a menu.
Ask only what the engineer needs
Address, postcode, the problem, the appliance, access and a callback number. Nothing more.
Read details back
Postcode, house number and phone number, so mistakes are caught on the call.
Say what happens next
Booked for Thursday morning, passed to the engineer on call, or a call back by a stated time.
Offer a person
A clear way to ask for a human, and an honest answer about when that will happen.
Stay calm and plain on emergencies
For a gas smell, your approved safety message and the national gas emergency number, at once, with no technical advice.
Scripts that work and scripts that do not
Small wording changes make a large difference to how callers react. Compare these pairs and use them as a starting point for your own.
- Weak: “Welcome to our virtual assistant service. Please listen carefully as our options have changed.” Better: “Northside Heating, you're through to our automated assistant. What's the problem today?” (Example firm name.)
- Weak: “I didn't understand that.” Better: “Sorry, I missed the house number. Could you say it again?”
- Weak: “An engineer will be in touch.” Better: “I've passed this to the engineer on call now. If you don't hear back within the time we've set, ring this number again.”
- Weak: silence after the caller asks for a person. Better: “I can't transfer you tonight, but I've marked that you'd like a call from the office first thing tomorrow.”
- Weak: reading a long privacy statement. Better: one sentence on recording, with the full notice on the website.
Listen to real calls before changing scripts. The words that confuse callers are rarely the ones you expect, and a small change tested for a week tells you more than a long debate in the office.
When to keep a person on the line
Some calls are better handled by a person whenever one is available: complaints, disputes about bills, bereavements and callers who are clearly confused or distressed. Set your system to recognise these where it can and pass them on, or promise a call back with a time. Automation that knows its limits earns more trust than automation that tries to handle everything.
Concerns to address openly
Public worries about AI centre on data and decisions. In the same government survey, around nine in ten adults saw important decisions being made without human input as a risk[1]. On a heating firm's phone line, the honest answer is that the AI takes down the job and follows your rules, while people decide on diagnosis, pricing and anything unusual. Say so in your privacy notice and on your website, and make sure it is true.
Businesses that use AI mostly keep people checking the output: 84% reported at least some human checking in 2025[4]. Telling customers that someone reads every job summary is reassuring, as long as you actually do it.
Measuring acceptance in your own firm
| Signal | What it tells you | What to do if it is high |
|---|---|---|
| Hang-ups in the first seconds | Callers put off by the greeting | Shorten it; make sure the firm's name comes first |
| Requests for a person | Calls the system should hand over | Review recordings; add rules or a faster hand-off |
| Repeat calls about the same job | Callers unsure anything will happen | Improve the what-happens-next step and follow-up |
| Address or number errors | Read-back is not working | Test with different accents and background noise |
| Complaints mentioning the phone | Trust problems | Listen to those calls first and fix the cause |
Compare these with the period before AI where you can, such as missed calls, abandoned calls and complaints. Many firms find the honest comparison is not between AI and a perfect receptionist but between AI and voicemail at nine at night.
Worked example: adjusting after a month
Example: a fictional heating firm switches on AI answering for evenings and weekends. After a month, the owner listens to a sample of calls. Most callers give their details without fuss, but early hang-ups are higher than expected, and several older callers ask whether they are talking to a real person. He shortens the greeting so the firm's name and the word automated come in the first five seconds, adds a line offering a call back from a person the next morning, and asks the system to slow down when reading back numbers. The next month's early hang-ups fall, and requests for a person drop to a handful, mostly complex complaints that should go to a person anyway.
Telling your customers in advance
- Add a line to your website: calls out of hours are answered by an automated assistant that takes down your job and contacts the engineer on call for urgent problems.
- Mention it on service reminders and invoices for a few months after you start.
- Tell letting agents and landlords directly, since their tenants will be calling.
- Brief your office staff, so they can answer questions with confidence.
For the legal side, read AI call handling and UK GDPR. For choosing a tool, use the AI phone tool checklist, and see appointment windows for setting expectations about arrival times.
Frequently asked questions
Do customers mind talking to an AI on the phone?
Many do not, especially if it is quick and honest. Callers object most to not being told, being misunderstood and having no route to a person.
Should an AI phone assistant say it is AI?
Yes, at the start of the call. It is honest, it supports your data protection duties and it avoids callers feeling misled later.
Are older customers less comfortable with AI calls?
Surveys show older adults use voice assistants much less. Keep the conversation short, read details back clearly and offer a person.
How do I know if customers accept it?
Track early hang-ups, requests for a person, repeat calls and complaints, and listen to a sample of recordings regularly.
References
Department for Science, Innovation and Technology (GOV.UK) (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑
Office for National Statistics (2020). Internet access - households and individuals, Great Britain: 2020 (opens in a new tab) ↑
Department for Business, Energy and Industrial Strategy (GOV.UK) (2021). Consumer protection study 2022: understanding the impacts and resolution of consumer problems (opens in a new tab) ↑
Department for Science, Innovation and Technology (GOV.UK) (2025). AI adoption survey report 2025 (opens in a new tab) ↑