The property market has one characteristic that sets it apart from most other industries: enquiries arrive when the agent isn't working. People browse listings in the evening after work, at the weekend, over lunch. They message three properties at once and book a viewing with whoever replies first. Not with whoever has the best property.
That is exactly why first contact in real estate is one of the most sensible places to deploy an AI agent. Not because it replaces the agent, but because it covers precisely the hours when the agent isn't available — and the stretch of conversation that repeats almost word for word on every single enquiry.
What the agent actually handles
First contact with an enquirer almost always has the same structure: confirmation that the property is still available, a few follow-up questions that weren't in the listing, and arranging a viewing. An agent handles all three steps, provided it has access to current data.
- Answering follow-up questions about the property — aspect, floor, condition, energy certificate, monthly costs, parking, what's included in the price. These are facts that sit in the system; they just never made it into the listing.
- Checking availability — whether the property is free, reserved or sold. This has to come live from the internal system, not from what the agent remembers.
- Booking a viewing — offering free slots from the agent's calendar, writing the appointment in, and confirming it to both sides.
- Capturing and classifying the enquiry — contact details, which property they're interested in, whether it's a purchase or a rental, whether financing is sorted, and what their decision horizon is.
- A first filter for sellers — for an enquiry along the lines of "I want to sell my flat", the agent collects the address, type, floor area and condition, and passes it on as a prepared brief.
Where the line is
Real estate is a regulated industry, and part of the communication can't be delegated to software even when it would be technically possible.
| Activity | Agent | Human agent |
|---|---|---|
| Confirming availability, property details | yes | — |
| Booking a viewing | yes | — |
| Capturing the client's contact and intent | yes | — |
| Negotiating on price | no | yes |
| Valuing the property | no | yes / a surveyor |
| Reservation and brokerage contracts | no | yes |
| Mortgage and financing advice | no | yes / a financial adviser |
The agent also shouldn't comment on a property's legal status — charges, easements, disputes. Even when it has that data in the system, misinterpreting it has direct consequences. The correct behaviour is to hand the question to a person, not to answer approximately.
A similar line applies to price. Even though the asking price is public in the listing, "would you come down on that?" isn't a question about a fact — it's the opening of a negotiation, and that is a human's job. We looked at the same principle for sales agents in our article on lead qualification and automated follow-up.
What it needs to be connected to
An agent with no link to current data is worse in real estate than no agent at all. If it confirms a flat is available when a colleague reserved it yesterday, the client turns up to a viewing for nothing and the agency loses more than one enquiry.
- The property listing system — not just the list, but above all the status (available / reserved / sold) in real time.
- Agents' calendars — without them the agent can't offer slots, only promise that somebody will be in touch.
- A CRM or enquiry register — so every enquiry is logged against the right property and the right agent, rather than landing in an inbox. We described the technical side of that kind of connection in our piece on integrating an agent into a CRM.
- Assignment rules — who receives enquiries for which properties, what happens when an agent is on holiday, and how long an enquiry may wait before it is escalated.
If the agency keeps this data in spreadsheets or only in its agents' heads, the first step isn't an agent — it's making the data accessible. Without that you can only build a chatbot that answers in generalities, and customers spot that immediately.
Where the effect is largest
In practice it pays to focus on three moments:
Evenings and weekends. This is the main reason to deploy at all. An enquiry that arrives at ten on a Friday night and gets an answer on Monday morning has, in most cases, already gone to a competitor.
Repeat questions on large developments. New-build and developer projects attract dozens of near-identical questions about layouts, completion dates and specification standards. This is the most measurable saving.
Follow-up after a viewing. A short "how did you find the flat, any questions?" the next day is the thing that gets skipped when everyone is busy. An agent handles it without feeling automated, as long as the message is specific to that property.
How to approach it
A sensible first step isn't to put an agent on all communication, but on one clearly bounded task — answering questions and booking viewings for a single development or a single branch, for instance. That shows whether the availability data is reliable enough and lets you tune the handover rules. Only then does widening the scope make sense.
What proves hardest at a given agency depends mainly on the state of the existing systems — on whether current availability and calendars can be read programmatically or are maintained by hand. If you're considering something similar, we'll go through it specifically in a no-obligation consultation, or take a look at our AI and automation solutions.
Summary
An AI agent in an estate agency makes most sense at first contact: it confirms availability, answers follow-up questions and books a viewing at the times when the agent isn't by the phone. Price, negotiation, contracts and valuation must stay with a person. And the whole thing hinges on one thing — whether the agent can see the current state of the listings. Without that its answer is fast but wrong, which is worse than not answering at all.