Key takeaways
- AI phone answering now handles routine calls well: property facts from approved details, viewing and valuation requests, repair reports and routing.
- It should not give valuations, negotiate offers or judge referencing; those need a person.
- Most agencies are micro businesses, and small firms adopt AI less: 28% of businesses with up to 9 staff used AI against 49% of large firms[1].
- Callers want a human route: 91% of UK adults worry about important decisions made by computers without human input[2].
What "AI phone answering" means in an agency
An AI phone answering service picks up your branch calls, speaks naturally with the caller, works out what they need and either deals with it or passes it on. Unlike a recorded menu, it understands a sentence such as "I'm calling about the flat on Mill Road, is it still available and can I see it on Saturday?" and responds to all of it. Unlike a simple message-taking service, the better systems can complete defined tasks, such as booking a valuation into a calendar, and hand a full summary to your team.
Agencies are looking at it for practical reasons. Branches are small: about 90% of the UK's 31,430 real estate agencies are micro businesses with 0 to 9 employees[3]. When two negotiators are out on viewings and the third is on another call, the phone rings out. Evenings and weekends, when many buyers and tenants make their calls, are even harder to cover.
Adoption is rising, but small firms lag
Business use of AI has grown quickly. Self-reported AI use among UK businesses with 10 or more employees rose from around 12% to around 35% between late 2023 and 2026[4]. Small firms are behind: 28% of businesses with 0 to 9 employees reported using at least one AI technology, compared with 49% of those with 250 or more employees[1].
Source: Office for National Statistics, 2026[1]
Show the data
| Item | Value |
|---|---|
| 0 to 9 employees | 28% |
| 250 or more employees | 49% |
Adoption has not, so far, meant fewer jobs for most firms. Of UK businesses using AI to improve operations, 63% reported no change in headcount, 6% a decrease and 1% an increase[5]. In agencies the realistic aim is to stop losing calls and to take routine admin off negotiators, not to replace the people who win instructions.
What it does well today
| Call type | What good AI answering handles | Where it stops |
|---|---|---|
| "Is it still available?" | Confirms status and key facts from details you approve: tenure, council tax band, parking | Anything not in the approved details |
| Viewing request | Takes the property, preferred times and the caller's position | Confirming a slot without a negotiator, unless diaries are connected |
| Valuation request | Takes the address, reason for moving and timescale; books into the valuer's calendar | Any opinion on price |
| Rental enquiry | Gives the advertised rent and the same next step to every applicant | Discussing offers, affordability judgements |
| Repair report | Logs the problem and access details; flags urgent issues | Diagnosing the fault or promising a contractor time |
| "Can I speak to Sarah?" | Routes by person, branch or department; warm transfer when free | Nothing: this is routine |
Afterwards, the call should arrive with your team as a recording, a transcript and a short summary, attached to the contact. That record is often as valuable as the call itself, because it removes the "who took that message?" problem.
Where a person must take over
Some calls should always reach a human, either live or by fast callback. Write the list down and test it.
- Offers, counter-offers and anything about negotiating price.
- Valuations and price opinions of any kind.
- Referencing outcomes, holding deposit disputes and decisions about who gets a tenancy.
- Complaints, and callers who are angry or distressed.
- Signs of vulnerability, such as confusion, bereavement or someone who sounds under pressure.
- Emergencies: gas smells, flooding, no heating for a vulnerable tenant, security risks. These go to your out-of-hours contact at once.
- Any caller who asks for a person.
Public attitudes back this up. In government tracking research, 91% of UK adults said they were concerned about important decisions being made by computers without human input[2]. People are generally comfortable with automation for logistics. They want a person when money, homes or disputes are at stake.
Current limits to plan around
- It only knows what you give it. Out-of-date listing details produce wrong answers, so decide who updates them and when.
- Accents, poor lines and background noise still cause errors. Check transcripts for misheard names, numbers and addresses.
- Diary booking depends on your systems. Where calendars are not connected, it can take a request but not confirm a slot.
- It cannot see the property. Questions such as "is the garden north-facing?" need to be in the approved details or passed on.
- Long, rambling calls with several issues are harder. The best systems summarise each issue separately for the right team.
Setting it up: the information it needs
Most of the work in a good deployment is not technical. It is writing down what your best negotiator already knows. Before going live, gather the material the service will answer from and the rules it will follow.
- A property sheet for every live listing: status, price or rent, tenure, council tax band, service charge, parking, EPC rating and any approved notes from the seller or landlord.
- Branch details: opening hours, addresses, who covers which area and department.
- Routing rules: who takes sales, lettings, property management, new instructions and complaints, with backups.
- Urgent issue definitions and the out-of-hours contact for each type.
- Words to use and avoid: your greeting, how to describe your services, and phrases that must never be said, such as any price opinion.
- A short list of frequently asked questions with your approved answers.
Assign an owner for each item. The single biggest cause of poor AI answering is a property sheet that was accurate in June and wrong by September.
Greeting and tone
Callers decide in the first few seconds whether they trust the line. Keep the greeting short, name the branch, say plainly that it is the branch's AI assistant and that calls are recorded, and ask an open question. Avoid a long list of options. Then let the service speak in the same plain, friendly tone your staff would use, without slang and without pretending to be a named human colleague. People respond well to honesty and speed; they respond badly to discovering halfway through that they were not told what they were speaking to.
How to tell whether it is working
Measure the baseline
Before switching on, count missed calls by hour and day for two weeks.
Start with overflow
Let it answer only when staff do not pick up within a few rings, plus evenings and weekends.
Review every summary for the first fortnight
Look for wrong facts, missed handovers and callers who asked for a person.
Track outcomes
Valuations booked, viewing requests captured and repair reports logged out of hours.
Listen to callers
Watch reviews and complaints for any mention of the phone experience.
Keep the first two weeks deliberately narrow. If you switch it on for every call on day one and something goes wrong, you will not know whether the problem is the greeting, the property data or the routing. Widen in steps: first overflow, then lunchtimes, then the early evening, reviewing at each stage before moving on.
A clear sign of success is boring: fewer "I rang three times and nobody answered" complaints, and Monday mornings that start with a tidy list of requests rather than a voicemail backlog.
Example: a Saturday afternoon
Example (fictional): at 3.40pm on a Saturday, with both negotiators on viewings, a seller calls to ask for a valuation. The AI answering service takes the address and the reason for moving (a job relocation in the spring), books a slot on Tuesday in the valuer's calendar and sends a confirmation. Ten minutes later a tenant reports water coming through a ceiling; the call is flagged as urgent and passed to the out-of-hours contact with the access details. A buyer asks about a flat that went under offer on Thursday; the service explains the position and takes backup buyer details. On Monday morning the team has three summaries, one booked appraisal and no missed calls.
That is the realistic picture today. Ava, for example, takes viewing requests for a negotiator to confirm rather than booking them straight into diaries, and never says what a home is worth. For how this kind of tool compares with a basic message-taking service, read AI receptionist vs AI employee, and before you choose any supplier, work through the questions to ask before AI answers your calls. Calls about properties already under offer are a good test case.
Frequently asked questions
Can AI answer calls in my branch's name?
Yes. It answers with your greeting and follows your rules on what to say, what to book and when to transfer.
Will callers know it is AI?
They should. Be open about it in the greeting or when asked, and always offer a route to a person.
Can it give a valuation figure?
It should not. It can take the details and book a valuation with your valuer.
Does using AI mean fewer staff?
Not for most firms so far: 63% of UK businesses using AI to improve operations reported no change in headcount[5].
References
Office for National Statistics (2026). Artificial intelligence in UK businesses: 2023 to 2026 (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 (Nomis) (2026). UK Business Counts - enterprises by industry and employment size band (SIC 68310, 68320), 2026 (opens in a new tab) ↑
Office for National Statistics (2026). Artificial intelligence in UK businesses: 2023 to 2026 (opens in a new tab) ↑
Office for National Statistics (2026). Artificial intelligence in UK businesses: 2023 to 2026 (opens in a new tab) ↑