Key takeaways
- A basic AI receptionist answers and takes a message; someone in your firm still has to do everything else.
- An AI employee takes down the whole job, follows your rules for urgent calls and books routine work into the diary.
- The difference shows on the calls that matter most: winter breakdowns, gas smells and calls nobody on the rota answers.
- If you only need messages, a basic receptionist is enough; if call-backs are your bottleneck, it is not.
- Test both with the same five difficult calls before you decide.
Two different jobs
Both answer the phone with an AI voice. The difference is what happens after hello. A basic AI receptionist greets the caller, asks for a name, number and reason, and passes a message to the office. An AI employee does the work a good office manager does on the call itself: it writes down the full job, decides from your rules whether it is urgent, gets the right person on the line or books the visit, and leaves a record the team can act on without ringing back for details.
Neither is better in the abstract. A sole trader who returns every message within the hour may only need the first. A firm losing jobs because messages sit until the morning needs the second.
Side by side
| Situation | Basic AI receptionist | AI employee |
|---|---|---|
| Boiler breakdown on a Saturday | Takes a name and number; promises a call back | Takes the postcode, boiler make, fault code and whether there is heating and hot water, marks it urgent and rings the engineer on call |
| Nobody on the rota answers | Call ends in a message | Follows the no-answer step you set and tells the caller honestly what happens next |
| Caller says they can smell gas | Takes a message | Reads your approved safety message and the gas emergency number, gives no technical advice and flags the call urgent |
| Annual service or landlord safety check | Leaves a message for the office | Books it into your diary, with landlord, agent and tenant details for access |
| Do you cover my area? | Says someone will call back | Checks the postcode against the areas you approved |
| After the call | Sends a short message | Sends a full job summary with the recording and transcript |
The table describes what each kind of system is designed to do. Individual products vary, so test the specific one you are considering rather than relying on the label.
Why the difference matters in heating
Heating calls are lumpy. Household gas demand in the first quarter of 2026 was more than five times the level in the third quarter of 2025[1], and breakdowns rise with it. On a cold Monday, a pile of messages that each need a call back is the problem, not the solution. Every call back is another conversation to collect details that could have been taken the first time.
Some calls also carry a clock. Since 27 October 2025, social landlords in England have had to investigate emergency hazards and make them safe within 24 hours under Awaab's Law[2]. If you hold that kind of work, a message that waits until the morning can put the landlord, and your contract, at risk.
Engineers' time is the other constraint. Employee plumbers and heating and ventilating installers had median total paid hours of 40 a week in 2026[3]. An on-call engineer who is rung for every message, including routine bookings and area questions, loses the evening to calls a system could have handled.
What AI employee should mean
The phrase is a marketing term, so pin it down. Before accepting that a product is more than a receptionist, check that it can do each of these on a live test call.
- Take down the job in full: address, postcode, appliance, fault, symptoms, access and a callback number, read back to the caller.
- Apply your urgency rules, not its own, and explain to the caller what happens next.
- Follow an on-call rota: engineer first, backup second, and a clear step if nobody answers.
- Handle a gas smell call with your approved script, without technical advice.
- Book routine work into a real diary, or leave it as a request when your rules say so.
- Answer the questions you approve, such as areas, hours and Gas Safe details, and say when it does not know.
- Send the team a summary that can be acted on without ringing the customer back.
When a basic receptionist is enough
- You are a sole trader who returns calls quickly and wants a polite answer rather than voicemail.
- Most calls are from existing customers who will wait for a call back.
- You do not do emergency or out-of-hours work.
- Your diary is booked by you personally and you do not want anyone else touching it.
In those cases, paying for booking and rota features you will not switch on adds cost and set-up time without changing the result. Start simple, and revisit the choice if call-backs begin to pile up or you take on emergency work.
When an AI employee earns its place
- You lose jobs because callers ring the next firm while waiting for a call back.
- You run an on-call rota and want urgent calls routed without the office.
- A large share of calls are routine bookings: services, safety checks and survey visits.
- Callers often need the same answers: areas covered, opening hours, what happens next.
- The office spends hours each week ringing back for details that were not taken first time.
Keeping humans in charge
Whichever you choose, your firm stays responsible for what is said in its name. Most businesses that use AI check its work: 84% reported at least some human checking of AI outputs in 2025, and 67% significant checking[4]. For phone tools that means approving the script and rules, reading summaries daily at first, and listening to a sample of recordings every week. Firms adopting AI mostly want efficiency[4], and in construction the main use is improving operations rather than changing how customers are served[5].
Source: Office for National Statistics, 2026[5]
Show the data
| Item | Value |
|---|---|
| Improve business operations | 65.1% |
| Provide or personalise customer services | 15.6% |
Five test calls to tell them apart
The Saturday breakdown
Ring as a customer with no heating and a fault code. See whether it takes the boiler make, code and heating status and what it does next.
The gas smell
Say you can smell gas. It must give your approved safety message and the emergency number, and must not offer technical advice.
The missed engineer
Ask a colleague on the rota not to answer. See what the caller is told.
The landlord booking
Book a safety check for a tenant-occupied flat and check the diary and access notes afterwards.
The awkward question
Ask something it should not answer, such as how much a new boiler would be. It should say it does not know or follow your approved answer.
Worked example: a winter Monday
Example: a fictional four-engineer firm gets most of its winter Monday calls before nine. With a basic AI receptionist, the office starts the day with a list of messages, rings each caller back for details and books urgent visits by eleven. With an AI employee set up on the same rules, the urgent callers were already passed to the on-call engineer over the weekend, routine services are in the diary with access notes, and the office starts with summaries to check rather than calls to make. The firm's own figures over a fortnight show how many call-backs disappeared, which is the number to compare against cost.
Questions to ask about the hand-off
The weakest point of any phone system is the moment the call ends. Ask any supplier to show you exactly what your team receives, using a real test call, and check it against these questions.
- Could an engineer go to the job from this summary without ringing the customer?
- Is the urgency clear at a glance, and does it match your rules?
- Is the recording or transcript attached, so a dispute about what was said can be settled?
- Does it reach the right person: the on-call engineer for urgent work, the office for routine bookings?
- What happens if the summary is wrong, and how do you correct the rules so it does not happen again?
If the answers are vague, the system is a receptionist whatever it is called.
Where Ava fits
Ava is an AI Employee designed for plumbing and heating firms. Live today, it answers calls in your firm's name, takes down the job, follows your on-call rota, handles gas-smell calls with your approved message, books routine visits into a tested diary and sends a summary with the recording and transcript; text confirmations and creating jobs directly in job-management software are planned rather than live. For the next steps in choosing any tool, see the AI phone tool checklist and AI receptionist vs answering service.
Frequently asked questions
What is an AI employee?
An AI system that does more than take messages: it records the whole job, applies your rules for urgency, follows your rota and books routine work, leaving a summary your team can act on.
Are AI receptionists and AI employees the same thing?
No. A basic AI receptionist answers and passes on a message. The difference shows on urgent calls, bookings and what the team receives afterwards.
Can an AI employee handle gas emergency calls?
It should only read your approved safety message and the national gas emergency number, give no technical advice and flag the call urgent. Test this before going live.
Will it replace my office staff?
In most heating firms it takes the repetitive calls and out-of-hours cover, so office staff spend less time on call-backs and more on scheduling, customers and problems.
References
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) ↑
Ministry of Housing, Communities and Local Government (GOV.UK) (2025). Awaab's Law: Guidance for social landlords - Timeframes for repairs in the social rented sector (opens in a new tab) ↑
Office for National Statistics (2026). Annual Survey of Hours and Earnings 2026 provisional, Table 14 (occupation by 4-digit SOC 2020) (opens in a new tab) ↑
Department for Science, Innovation and Technology (GOV.UK) (2025). AI adoption survey report 2025 (opens in a new tab) ↑
Office for National Statistics (2026). Business insights and impact on the UK economy: wave 165 dataset (AI Current Purpose) (opens in a new tab) ↑