Key takeaways
- A human answering service brings warmth and judgement but usually only takes messages unless you pay for more.
- A basic AI receptionist covers every hour and every peak, but may also only take messages.
- The real comparison is what happens after the call: message, booking or a call to your on-call engineer.
- Compare cost models over a full winter, including call-backs your office still has to make.
- Many firms do best with a hybrid: one route for routine calls, a clear path to a person for the rest.
Four ways to cover the phone
Most plumbing and heating firms are tiny. About 95% of registered businesses in the trade are micro firms with fewer than ten employees[1], and around 48% of plumbers and heating installers are self-employed[2]. For many of them, the person who answers the phone is also the person under the sink. When that does not work, there are four options: voicemail, a human telephone answering service, an AI receptionist that takes messages, or an AI system that books and routes jobs, sometimes called an AI employee.
Voicemail is free and loses jobs. Callers with no heating ring the next firm rather than leave a message. The other three options are worth comparing carefully, because they differ less in how they sound than in what they do once the caller has explained the problem.
Side by side
| Factor | Human answering service | Basic AI receptionist | AI that books and routes |
|---|---|---|---|
| Hours | Depends on the package; out-of-hours often extra | Every hour | Every hour |
| Winter peaks | Queues possible when many clients are busy | Answers every call at once | Answers every call at once |
| What the caller gets | A friendly person taking a message | A message taken by an AI voice | The job taken down, urgent calls routed, routine visits booked |
| Booking | Sometimes, if given diary access and scripts | Usually not | Into your diary, by your rules |
| Urgent calls | Can phone your engineer if instructed | Message only, unless configured | Follows your on-call rota and no-answer step |
| Consistency | Varies by operator | Same script every time | Same rules every time |
| Judgement with upset callers | Strong | Limited | Limited; needs a clear path to a person |
| What you receive | Message by email or text | Message or short summary | Job summary, often with recording and transcript |
Products differ, and some answering services offer booking while some AI products do little more than voicemail with a voice. Use the table as a list of questions, not a verdict.
Cost models: compare like with like
Human services usually charge by the call or by the minute, often in bundles, with extras for out-of-hours cover, call transfers or booking. AI products tend to charge by usage or by a subscription with an allowance. Neither model is cheaper in general. What matters is what the whole arrangement costs your firm across a year, including the work it leaves behind.
- Count your calls by month for a year. Heating calls are seasonal, and household gas demand in the first quarter of 2026 was more than five times the level of the third quarter of 2025[3]; usage-based bills follow that curve.
- Add the call-backs. A message that needs a call back costs office time, and some callers will have gone elsewhere by then.
- Add set-up time: scripts, rules, diary access and testing.
- Add what happens on overflow: who answers when the service is busy or the system cannot help.
- Compare against the jobs each option is likely to keep, not just the cost of answering.
Urgent calls and escalation
The difference between options is sharpest on urgent calls. A human operator can ring your engineer if you have given them the rota and the authority. A basic AI receptionist usually cannot. A system built to route calls follows the rota: engineer on call first, backup second, then a clear message to the caller if nobody answers.
Gas escapes are different again. When told of a gas escape, the gas conveyor must attend as soon as reasonably practicable and stop the escape within 12 hours[4]. Whoever answers your phone, human or AI, must give a caller who smells gas your approved safety message and the national gas emergency number straight away, and must not attempt technical advice. Test this script with every option before going live.
Script for any answering option: “If you can smell gas, please leave the property if you can, don't use switches or flames, and call the national gas emergency number now. I'm flagging this to our engineer as urgent as well.” Adapt it to your approved wording.
Will callers accept an AI voice?
Many already talk to machines. Six in ten UK adults reported using AI chatbots in the past three months in a 2024 government survey[5]. Even in 2020, 35% of adults in Great Britain used a virtual assistant speaker or app, though use varied sharply by age, from 49% of 25 to 34 year olds to 17% of those aged 65 and over[6].
Source: Office for National Statistics, 2020[6]
Show the data
| Item | Value |
|---|---|
| All adults | 35% |
| Aged 25 to 34 | 49% |
| Aged 65 and over | 17% |
Heating firms serve many older customers and landlords' tenants with every kind of need. Whatever you choose, give callers a way to reach a person, say clearly when they are speaking to an automated system, and keep the conversation short. Read more in do customers mind talking to AI on the phone.
Which fits which firm
- Sole trader, no emergency work: a human service or basic AI receptionist for messages may be enough, if you return calls the same day.
- Sole trader or small firm with out-of-hours work: you need something that can reach you for urgent calls and book routine work while you are on a job.
- Multi-van firm with an office: an AI that books and routes can cover overflow, lunchtimes, evenings and weekends, with the office handling exceptions.
- Firms with many elderly or vulnerable customers: a human service, or an AI system with a quick, reliable route to a person, during the hours those customers call.
- Firms with social housing contracts: whatever you use must record report times and route emergencies without delay.
Hybrid set-ups that work
You do not have to choose one for every hour. Common patterns include the office answering in working hours with AI covering overflow and out of hours; AI handling routine bookings while a human service takes complaints and complex calls; or a human service during the day for a firm whose customers prefer a person, with AI covering nights and weekends. Keep one diary and one set of rules underneath, so every route books the same way.
Whatever the mix, tell your regular customers and letting agents how calls are handled, especially if anything changes. A short note on invoices or your website saves surprise and complaints in the first weeks.
Questions to ask any provider
- Exactly what do we receive after each call, and how fast?
- Can you book into our diary, and how do you avoid double bookings?
- How do you reach our on-call engineer, and what happens if nobody answers?
- What do you say to a caller who smells gas? Show us the script.
- How are calls recorded, where is the data held, and for how long?
- What does the bill look like in our busiest month, not an average one?
- How quickly can we change the script, the rota or the areas we cover?
- How do we leave, and what happens to our recordings and records?
Write the answers down and compare them side by side. A provider that cannot answer the gas question clearly, or the busiest-month question honestly, has answered the more important one.
Running a fair trial
Write five test calls
A winter breakdown, a gas smell, a landlord safety check booking, an out-of-area enquiry and a complaint.
Run them on each option
Ring as a customer, at different times, including a busy weekday morning and late evening.
Score what you receive
Could an engineer act on it without calling back? Was urgency right? Was the gas script correct?
Trial on live calls
For at least two weeks that include a cold spell if possible.
Count the call-backs
The number of calls your office still had to make is the clearest measure of value.
Worked example: one firm, two trials
Example: a fictional three-engineer firm compares a human answering service and an AI system that books routine visits, each for two weeks in November. The human service is warm and handles an upset tenant well, but every call becomes a message and the office makes most of them into call-backs the next morning. The AI system books most service and safety check calls straight into the diary and routes two Saturday breakdowns to the engineer on call, but one caller asks for a person and has to wait for a call back. The firm chooses the AI system for evenings, weekends and overflow, with the office answering during the day and a clear option for callers to request a person.
Frequently asked questions
Is an AI receptionist cheaper than an answering service?
Not necessarily. Compare the whole year, including winter peaks, out-of-hours extras and the call-backs your office still has to make after each message.
Can a telephone answering service book jobs?
Some can, if you give them diary access and clear scripts. Many only take messages as standard, so ask exactly what is included.
What happens if a caller wants a real person?
Any option should give a clear route: transfer to the office in hours, or a promised call back with a time. Test this on a trial call.
How should gas smell calls be handled?
With your approved safety message and the national gas emergency number straight away, no technical advice, and the call flagged to your engineer as urgent.
References
Office for National Statistics (Nomis) (2026). UK Business Counts - enterprises by industry and employment size band (SIC 2007 class 43.22), via Nomis (opens in a new tab) ↑
Office for National Statistics (Nomis) (2026). Annual Population Survey - regional - occupation (SOC2020) by sex by employment type, via Nomis (opens in a new tab) ↑
Department for Energy Security and Net Zero (2026). Energy Trends: UK gas, ET 4.1 natural gas supply and consumption (quarterly), September 2026 (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (1996). The Gas Safety (Management) Regulations 1996, regulation 7 (opens in a new tab) ↑
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) ↑