Which AI should a real estate agent use?

There is no single tool that covers everything. What works best is a set of targeted agents connected to what you already use, your CRM and inbox, rather than one more general-purpose piece of software to open every morning. The first project worth tackling is almost always contact follow-up, ahead of listing copy or valuations, which are already well covered by industry software.

Can AI write a property listing?

Yes, a first draft, built from the property's characteristics. It is a real but modest time saving, and a topic already covered by most real estate software. A review still matters: a generative AI can get a figure or a legal detail wrong, and an inaccurate listing is your liability, not the tool's.

How do you follow up with contacts without losing your day to it?

By letting an agent flag which contacts need a follow-up based on rules you set, such as the last exchange, the stage of a project, or the season, then draft a message using the real context of the file, and wait for your approval before sending. You keep the decision, the agent handles the watching and the drafting, which is the part that eats the most time day to day.

Does AI replace a real estate agent?

No, and no serious source claims otherwise. It handles preparation and repetitive follow-up. The viewing, the negotiation, and the trust a client places in a person stay out of reach for an automated system, however advanced it gets.

AI for real estate agents: what actually helps

What AI for real estate agents can actually do goes beyond listings and property valuations, whatever most content on the topic suggests. For a professional working alone, the real gain sits in following a network of several hundred contacts, too many to hold in one head, and in everything that happens between two appointments.

Where AI genuinely saves time in real estate

Most content on this topic stops at two uses: generating listing copy and automated property valuation. Both are real, but already covered by most industry software, so they offer little edge to anyone still searching for them.

According to the France Num 2025 barometer, across all business activities and with no figure specific to real estate, the share of small and mid-sized businesses using AI has doubled in a year to reach 26 percent. Usage is dominated by generative AI (22 percent, up 12 points) and chatbots and assistants (14 percent, up 9 points), well ahead of document analysis (6 percent) or task automation (5 percent).

Task	Real gain	Limit
Listing copy	First draft from the property's characteristics	Review required, already covered by industry software
Contact follow-up	Reminders and monitoring without depending on one person's memory	Human approval before anything reaches a client
Document analysis	Fast lookups in a mandate or a survey report	Never carries legal judgment
First inquiries	Sorting and drafting a reply	Fine qualification stays human

The real subject: following a portfolio of contacts

A professional working alone who tracks several hundred active contacts cannot remember, for each one, what stage they are at, when they were last in touch, or which newly listed property might suit them. Today, that tracking depends on a spreadsheet, a poorly maintained CRM, or a memory that eventually gives out.

Vu sur le terrain

A real estate investment advisor working alone, with a network of several hundred active contacts. Already comfortable with AI tools, he needed no training: what he wanted was systems that run in the background and plug into what he already used.

We set up a group of agents connected to his existing tools, with one requirement set from the start: human approval on anything that reaches a client. A market watch feeds his outreach instead of producing generic content.

Follow-up no longer depends on his memory or on when he happens to open his tools. What needs doing surfaces on its own.

What did not go as planned: his demand for reliability forced the project's scope to narrow. Three reliable automations beat ten approximate ones, and a demanding client is a good client, as long as you give them proof rather than promises.

It is the same principle we cover in our article on automated client follow-ups: the machine detects and drafts, the person keeps the decision to send.

Should AI be allowed to write to clients?

No, not without human approval. A message going out to an investor or a seller carries your reputation, and a generative AI can get an amount, a date, or a property detail wrong without it being obvious at first glance.

The right practice is the one we describe in how to stay in control of an AI agent: a defined scope, human approval on anything leaving the business, and a record of what was sent. This matters even more in real estate, since a contact file holds sensitive personal data (income, purchase plans, family situation), a point the CNIL's practical AI factsheets flag for any system that processes this kind of data.

What AI will not do in your place

Three things do not get delegated, whatever tool you choose:

The viewing. Feeling out a place and answering a hesitation on the spot cannot be automated.
The negotiation. It requires reading a human situation in real time and adjusting a position, something no system does reliably.
The trust. What a seller or an investor places in a person rather than a tool is built over time and delegates to nothing.

Content that promised otherwise would be selling a solution that does not exist. Saying so plainly is what keeps the rest of the article credible.

How much does a project cost, and how long does it take?

A first project focused on contact follow-up starts at 1,500 euros as a flat fee, never a daily rate, depending on how many existing tools need connecting and how wide the agreed scope is. Running costs then stay limited to a few dozen euros a month in API calls.

The time saved scales directly with the size of the portfolio tracked, and it is better announced below what is actually observed than above it: a professional disappointed by an inflated promise does not come back. For a portfolio of several hundred contacts, the time recovered is measured in hours per week, not minutes, but that figure varies too much from one file to another to state as a general rule.

Where to start this week

The simplest starting point is not a tool, it is a list: your active contacts and the last piece of information you hold on each one. That list, however imperfect, is what the first follow-up agent gets built on, not a generic system installed without knowing what it should watch for.

Before any commitment, a free 30-minute assessment checks what is genuinely automatable in your current setup, starting with whatever costs you the most time today. Take a look at our AI sales agent or our AI agents to see where this kind of project fits.

The takeaway

AI for a real estate agent is not primarily about listings or valuations, already well covered by industry software. The real gain lies in following a portfolio of contacts too large for one memory, with systematic human approval on anything reaching a client. The viewing, the negotiation, and the trust remain, and will remain, the person's work.

To find out what in your contact portfolio deserves automating first, our free 30-minute assessment starts from how you actually work today. See how our method works or check out our case studies.