A quarter of real estate companies with ten or more employees already use AI, but in agencies, the tool is arriving faster than the change in methods. According to the published analysis, the primary obstacle is not technical: it lies in work organization.
In many agencies, the scene is similar. An agent drafts listings with an AI assistant on their phone, the manager has attended two demos and followed a webinar, and everyone agrees the topic is important. Yet, the following Monday, the agency works exactly as before.
Insee figures confirm this desire. According to the institute, 26% of real estate activity companies with ten or more employees reported using at least one AI technology in 2025, compared to 14% in 2024 and 7% in 2023. On the client side, the Xerfi study on customer journeys in real estate indicates that 28% of buyers have already used an AI tool to refine their search.
AI has therefore entered agencies, without yet entering their organization.
Rapid but individual adoption
In the agencies observed, AI most often remains a personal use. Everyone uses it in their own way, sometimes with a free account, to draft a listing, an email, or a post faster. The gain is real for the task, but it disappears as soon as you look at the whole picture: information is re-entered from one software to another, follow-ups rely on each person’s memory, and the seller is still waiting for their report.
The agency has added a tool without changing its method. This stage is common, as a technology almost always enters through individual uses, but the risk is to stop there. These scattered uses also raise a data question: a mandate, an ID document, or a buyer’s income have no place in a personal account whose terms of use no one in the agency has read.
Why the change is stalling
The first obstacle is not price. According to the same Insee survey, 54% of companies not using AI cite a lack of expertise, and 53% of those using it say they are hindered by this same lack. Among non-users, 38% cite incompatibility with their existing software, as many as concerns about data protection and lack of legal clarity.
The real estate agency compounds its own difficulties. The manager is often also the lead agent, lacking the time or IT department needed to lead a project. Part of the team works under independent sales agent status, to whom you cannot impose a method as you would on an employee. The transaction software, portals, messaging, and calendars already form a stack of tools, and an error in estimation, a mandate clause, or a message commits the agency.
AI is therefore not stumbling on technology, but on organization.
Changing the method rather than adding a tool
Agencies that are moving forward start with a process rather than a tool. They choose a task repeated every week, which costs time and causes oversights, then rewrite it from start to finish. The visit report is well-suited. Today, the agent takes notes between appointments, drafts in the evening if they have time, and the seller ends up calling to find out how the visit went.
With AI, the agent dictates their impressions upon leaving the property. The tool formats the report, gathers the buyer’s remarks, and prepares the update of the file in the software. The agent proofreads and sends it, the seller is informed the same day, and the manager has a complete history of the file. The core of the change lies in task allocation: who does what, when, and who validates. A price, a mandate, or a commitment to a client remain human decisions, and the method must be written with its limits, supplemented by a usage rule specifying the authorized tools, opened with professional accounts, and the client data that can be entered there.
The change also requires a point person, not necessarily the manager, who tests the method, gathers colleagues’ feedback, and evolves it. Training is more effective when done on the agency’s actual files: an hour spent on a real report is more convincing than a day of general presentation. Before extending the approach, the agency verifies that the report is sent the same day, that the seller calls less, and that the team uses it without being reminded. The lack of expertise noted by Insee justifies, at the start, getting support until the team appropriates the method.
