AI in real estate: client enquiries, leases, valuation and tenant service

The real estate sector is really several different businesses. A developer sells flats in its own developments, an agency acts as an intermediary in the sale and letting of other people’s properties, and a property manager looks after buildings and tenants. Each has a different bottleneck, so the sensible first project differs too.

The table below shows where it is usually worth starting in each of these types of company. It is a starting point for discussion, not a fixed rule – what counts is where your team loses the most time.

Enquiries come in from your own website, property portals, adverts and referrals. Some concern a specific flat, some are general questions, and some come from people whose budget or plans do not match the offer. An adviser who answers them all one by one has less time for clients who are ready to buy.

An AI assistant, in the form of a website chat or a voicebot that calls back after an enquiry, replies immediately. It asks a few questions: about location, budget, timing and how the purchase will be financed. The adviser receives the contact with a short note and knows straight away whether to propose a meeting or a different offer.

Work it out on your own data. Suppose a company receives 300 enquiries a month and the first contact with each takes an adviser 10 minutes. That is 50 hours a month on initial conversations alone, some of which end with the conclusion that the offer is not a fit.

The assistant must meet the conditions listed below. Without them, a fast reply can do more harm than a late one.

A property manager handling dozens or hundreds of leases must keep track of rent indexation dates, lease end dates, contractual penalties and security deposits. This information is often recorded only in the text of the leases. AI can read it, collect it in a table and send reminders about upcoming deadlines.

The second use is comparing a lease with the company’s template. AI points out clauses that depart from the standard, such as a different indexation method or a right to early termination. The lawyer then does not read the whole contract from scratch, but checks the flagged passages.

It works similarly when buying a property or an entire portfolio, when hundreds of documents need to be reviewed: contracts, decisions, reports. This review of the legal status and documents before a purchase is known in practice as due diligence. AI speeds up its first, mechanical part, but the conclusion about the risk is drawn by a lawyer.

This kind of work can often be handled by Microsoft Copilot with a Microsoft 365 Copilot licence, if the contracts are kept in SharePoint. Another option is a separate AI assistant connected to the company’s document archive. Leases with private individuals contain personal data, so the tool must operate under a GDPR-compliant data processing agreement.

An AI model can suggest a price range based on similar listings and transactions, location, floor area and standard. This is useful to an agency when talking to a seller with inflated expectations and when setting the asking price.

Such a pointer is not a valuation in the legal sense. A formal valuation report, needed for example for a mortgage, in court or when dividing assets, may only be prepared by a licensed property valuer. At most, AI can help the valuer gather data on comparable transactions.

The quality of the pointer depends on the number of similar transactions. In a large city, for a typical flat in an apartment block, the model has something to go on. For an unusual house, a plot of land or premises in a small town, its result is not very reliable, and the client must be told this clearly.

Tenants report faults by phone, by email, through messaging apps and in person to the building administrator. Without a shared place, requests easily get lost, and the building owner does not know how quickly the property manager responds.

AI recognises what the request is about, assesses its urgency and assigns it to the right person or maintenance company. The system tracks the agreed response time and sends reminders about requests that have been waiting too long. Urgent matters, such as a flood or a heating failure, go straight to the person on call, not into the queue.

The same data is used to produce a report for the building owner. It shows the number of requests, response times and recurring faults, for example the same lift breaking down every month.

At many developers, the sales report is put together manually in Excel from several sources: the sales system, a reservations spreadsheet and bank statements. Power BI, Microsoft’s tool for reports and charts, combines this data and refreshes it automatically.

Since 2025, developers have also been required to publish up-to-date prices for all units on offer on their website. When the price list lives in several places, discrepancies easily arise between the website, a salesperson’s offer and the chatbot’s answer. Automatically transferring prices from a single source to the website and the AI assistant removes this problem.

An AI assistant can also answer the management board’s questions about the data, for example about the pace of sales in a given development. However, the answer is best treated as a starting point and checked against the source report.

AI in real estate touches on several groups of regulations. The table below lists those that most often matter for the projects described in this article.

In addition, every company using AI must take steps to build its employees’ knowledge of the technology. This obligation stems from Article 4 of the AI Act and has applied since 2 February 2025.

The most common mistake is rolling out a chatbot before the company has put in order the data it is meant to use. An assistant that does not know the current price list or availability answers quickly, but wrongly. The steps listed below make sense in this order.

Each step ends with checking the effect against figures measured before the start: time to first response, number of meetings booked, response time to requests. How long each stage takes depends on the number of channels and systems that need to be connected.

A small agency with a dozen or so listings and a few enquiries a day usually does not need an AI assistant. An adviser can reply quickly on their own, and a good CRM will be a better investment.

The same goes for premium properties, where the client expects to talk to a specific person from the first contact. There, AI works better in the background, for example in preparing documents, than in conversation with the client.