Search with AI for Homes & Villas by Marriott Bonvoy
Travelers describe the trip they want and get matching vacation homes. I led the Publicis Sapient work on Homes & Villas by Marriott Bonvoy's generative AI search.

Our starting point was an observation from the team: every travel site opens with the same two boxes, where do you want to go and when. Homes & Villas by Marriott Bonvoy had more than 140,000 vacation rentals behind those boxes. In March 2024 it started testing a search that skips them.
Travelers described the trip they wanted in plain language, with as much or as little detail as they liked. They got back matching destinations and homes, or close alternatives, with the weather and things to do. They could still pick the traditional search.
I led the work at Publicis Sapient. We conceived, designed and built it with the Homes & Villas team. As of March 2024 it ran on GPT models in Microsoft's Azure OpenAI Service.
From "where to" to "what do you want to do"
That's how I put it in the video: "We moved away from 'Where to' to 'What do you want to do?'" One of the launch examples was "Best A-frame home in the mountains with a fireplace and stunning view."
A guiding principle for the generative AI layer was that people should feel heard. One traveler in the video said they typed exactly what they'd have said to a person, and ended up on a lake they hadn't known about.
Brand-new technology, handled carefully
I pulled the team aside and asked if they were in: "We're going to try to do something that nobody's ever done before." We did it somewhere supportive and careful, because, as I said then, "We're using brand-new technology in brand-new ways." Travelers kept the choice of the old search, and the launch was a test, on the website first with mobile to follow.
The client made that possible. "It's a hospitality ecosystem," I said in the video, and being allowed to experiment and innovate inside it was "a real positive signal to us."
What we learned
We were very early. This was the beginning of the AI boom, and we had a lot to learn about building software with AI. We started, stopped and learned many times.
We also learned that people's reaction to chatbots was negative at the time. We were designing something nobody had seen, so we went as far as building an alternative UI so people wouldn't get hung up on the design. What worked better was skipping the chat entirely. We asked people where they wanted to go, then matched them to a property in the search results.
What it did
The case study reports more than 40,000 AI-powered searches, which now inform marketing strategy for upcoming seasons. Saved properties doubled. Rollout time for other brands fell from one year to three months.
It won a Bronze Stevie Award for Technology Breakthrough of the Year in Artificial Intelligence at the 2024 American Business Awards. The case study has the rest.
It's live today on the website and mobile web, and it's performing well. Those numbers aren't mine to share.