The interview is in French. Below is an English summary and a translated transcript.
In short
- Philippe Wellens, co-founder and CEO of Kleio, was interviewed by Radio Immo at RENT 2026, the real estate and new tech show held in Paris on November 4–5, 2026.
- Kleio is a brand AI agent: it understands a buyer's or renter's project in their own words, recommends the right properties and hands a qualified lead to the agent in the CRM.
- At Orpi, the agent runs on all 1,250 agency websites. In an A/B test run over several months, the site with the agent generated 15% more leads, and those leads converted 20% better.
- Visitors who go through the conversation convert three times more often than other site visitors. Beyond 10 messages, one request in three goes straight to a human agent.
- Close to one million conversations in six months give the brand a unique source of market intelligence.
- Next step: personal AI assistants such as ChatGPT, Claude, Meta Muse and Gemini, where buyers will be able to book a viewing without visiting the website.
- Philippe's advice to agencies: clean up their data and write down their local know-how, so AI assistants cite them more often.
Chapters
- 00:43 – What is Kleio?
- 01:53 – How it differs from a basic assistant
- 03:20 – Example: a family looking for a house near a train station
- 04:29 – The journey at Orpi, up to the human agent
- 06:36 – Results of the rollout
- 08:24 – Property listings in ChatGPT, Claude and Meta Muse
- 10:49 – Accurate information and control of customer data
- 12:26 – Partnerships and priorities for the next 12 months
- 14:20 – How agencies can prepare now
Transcript (translated from French)
Interview recorded on the Radio Immo set at RENT 2026, November 5, 2026. Translated and lightly edited for readability.
What is Kleio?
Radio Immo: We're back on the Radio Immo set for the second day of RENT 2026. Let's talk about agentic AI: describing your real estate project in your own words, chatting with an AI and getting tailored suggestions. How can these new journeys benefit clients and agencies? We're discussing it with Philippe Wellens, co-founder and CEO of Kleio, which works with Orpi on this transformation. Tell us what Kleio is.
Philippe Wellens: Kleio is a brand agent that helps every consumer and every prospect find the property that suits them. For Orpi, it's an AI agent on their website, and soon in ChatGPT, Claude and other agentic channels. It understands the client's needs and qualifies them: what do you need, what is your buying, investment or rental project? From there, it recommends the right property, whether an apartment, a house or a plot of land, and takes the client all the way to the contact request, which lands in the CRM. The real estate agents then take over and call the client with a very rich starting point: their project. That makes it easier and faster to get to the viewing and the next steps.
How is it different from an assistant that just answers questions?
Philippe Wellens: The agent really represents the brand's interests, here Orpi, and also Cogedim, which we work with. At Orpi, we deployed the AI agent on all 1,250 agency websites to welcome every prospect.
The first results are very interesting. We ran an A/B test over several months with Orpi: we're at +15% leads. Beyond that, those leads convert 20% better, because they're more qualified. The reason is that the project was described in natural language. An agent reads the conversation, sees the project in a structured way and thinks: "I understand this project. There are two children, they need something bigger, one more bedroom. I want to help, I want to advise them well, I'll call them." That's what's happening today.
Example: a family looking for a house near a train station
Radio Immo: Let's take an example. A family is looking for a house near a train station on a tight budget. What can your solution do for them?
Philippe Wellens: The agent helps them understand their own project. Why near a station? How many children? What are they really looking for? Which features matter, such as light or space, and how much? Once that's clear, the agent starts suggesting properties. The idea is to do it iteratively, to co-build the project together and converge on the right property over the conversation. We've also seen that people often don't know exactly what they want: they have a few constraints in mind that they can't express with keywords on traditional websites. They tell their story, we help them, and the project gets sharper as they go. It's a real conversation.
How does the journey work at Orpi, up to the human agent?
Radio Immo: At Orpi, how does the journey work, from the visitor's first request to the handover to an agent?
Philippe Wellens: Today it starts on the website. The visitor sees the AI agent, clicks and starts the conversation, and the back-and-forth helps them find the right property. Once they find it, we personalize the whole property page to make clear why it was recommended. For the family looking near a station, the agent checks nearby points of interest by geolocation, makes sure there is a station, and puts that argument at the very top of the listing. Instead of reading five pages of description and switching to Google Maps, everything is there upfront, with personalized arguments.
From there, we can offer to put the person in touch with an agent. That's important: a human always takes over. Interestingly, the more people chat, the more they want to talk to a human. Beyond 10 messages, one request in three goes straight to a real estate agent. It frees up agents' time, because all the early-stage hesitation is handled: when the lead arrives, it's qualified and ready. The others are kept for later.
What results has the rollout delivered?
Radio Immo: Looking back, what feedback do you have on lead quality, appointments and adoption by the teams?
Philippe Wellens: In the A/B test, comparing the site without the agent and the site with it, it's +15% leads, and leads that convert 20% better. Another interesting point: people who go through the conversation convert three times more often than other site visitors, because they are very well guided.
The market is a bit tight and people have to be selective, but it hasn't stopped. There are lots of questions like: what can I get on this budget, how much will it cost, is it the right time to buy? A first layer of advice is starting to happen in the conversations.
We now have close to one million conversations in six months. That's market intelligence: what happened last month, what's happening this month, how trends are shifting. It helps the brand understand its customers as a whole and adapt its channels and campaigns, in SEO and SEA. These conversations are a gold mine, including for knowing what to improve in the catalog or where to look for new seller listings.
Property listings in ChatGPT, Claude and Meta Muse
Radio Immo: You're also working on making listings available in assistants such as ChatGPT or Claude, which are now part of everyday life. What is actually possible today, and how does a buyer reach the property?
Philippe Wellens: That's the 2027 question, and it's happening now. The next generations will go straight to ChatGPT or Claude to look for a property. Demand is really shifting from portals to personal AI agents. Europe and France are a bit behind, but what's happening in the US is impressive. Meta Muse, for example, has had the strongest adoption since its launch in September this year. For the first time, it lets people go all the way: to the transaction in simpler cases, and in real estate, to booking a viewing slot directly from the AI, without going through the website. That removes even more friction.
These personal agents can do part of the job, but they aren't real estate experts. They need a brand agent on the other side to guide them: what is the user really asking for, what is the wider project, what are the real constraints? That happens very quickly, agent to agent. Then the brand agent recommends the right properties and offers the agency's services: insurance, the application file, financing. That's what we call agentic commerce, agent to agent, to reach all the people who start their search well before they reach the website. It started on ChatGPT and Claude, it will be on Meta Muse and Gemini: everyone in the US is already doing it. It's a new channel, and one that goes further than lead capture, because it can lead to a strong action such as booking an appointment, a phone call or even the transaction.
How do you guarantee accurate information, and who controls the data?
Radio Immo: How do you guarantee that the properties offered through this channel are still available and that the information is accurate? And who keeps control of the customer relationship data?
Philippe Wellens: Accurate information is the core of what we do: processing the data so AI agents can use it, and making sure everything they say and reason on is factual, with no hallucinations. We have verification mechanisms in place. My two co-founders and I have worked in AI for ten years, including at a US company in California that went public in New York, before we started Kleio. We know how to handle data and how to organize the different models so that everything an agent says is useful and correct.
As for customer relationship data, our client controls it, since it is their data that we take as input. And they keep control of the customer relationship, which is essential. We are a layer that makes things smoother and removes friction, but the brand stays in control.
Partnerships and priorities for the next 12 months
Radio Immo: You're preparing new partnerships in France. What can you tell us, and what is your priority for the next 6 to 12 months?
Philippe Wellens: We're preparing new partnerships in France and internationally. Kleio is already active in the US, Australia, the UK and elsewhere in Europe. The goal is to keep expanding in real estate and beyond, since we specialize in complex sales: real estate, travel, automotive, complex retail, insurance. Anywhere you need to match a vague, loosely defined customer need with a very complex and extensive product catalog. Once we've done it for one industry, we can roll it out to others.
The big focus is personal AI agents such as Meta Muse, ChatGPT, Claude and Gemini, to bring our clients into these channels and answer end customers directly there. But also all the other channels: websites, WhatsApp, where we follow up on lost leads, and so on. The platform supports many use cases and is very modular. So we can also open it to advanced partners who have their own engineers and want to use some building blocks to build their own AI agent. Kleio is then their technology partner.
How can agencies prepare now?
Radio Immo: For a professional listening to us: what should they do now to prepare their listings and their agency for these new habits?
Philippe Wellens: Work on their data. Clean it up and make it as relevant as possible, with the unique things the agency has that others don't: knowledge of the neighborhood, for example, and all of the agency's own know-how, as long as it can be written down. That will attract a new type of traffic, make listings more relevant, better ranked and more often cited, and ultimately help them meet demand better.
Radio Immo: Thank you very much, Philippe Wellens, co-founder and CEO of Kleio, exhibiting at RENT on stand SE04.
Philippe Wellens: Thank you for having me.
About
Interview broadcast by Radio Immo at RENT (Real Estate & New Tech) 2026. Watch more Kleio videos on the Kleio YouTube channel.







