In this talk from the Pendomonium+#mtpcon roadshow – Amsterdam 2024, Booking.com’s VP of Product Joe Futty runs through the recent work at the travel marketplace to incorporate Gen AI into its products.
Joe starts by saying that he joined Booking.com about 18 months ago because he loves working in two-sided marketplaces and he's very excited about the current disruption from Gen AI. He says:
“
Booking is all about making travellers’ lives easier. We need to have user empathy in everything we do.
”
He gives the analogy of a chef to explain how Gen AI can be used successfully and the pitfalls to look out for. He says a trainee chef might join a culinary school, and they’ll create recipes, get feedback, and adjust their recipes. Then, over time, they’ll create recipes with their favourite ingredients. By contrast, a Gen AI chef can look at every recipe on the internet, learn the outcomes of the recipes, and over time create new ones. But it could also create recipes with ingredients and tools that don’t exist. Joe suggests we think of Gen AI as the Einstein in your basement, where your imagination is the limiting power.
Gen AI means Booking can become a travel advisor, but “super smart prompting” is key to its success, Joe says.
If product personalisation is the holy grail, then Gen AI can be invaluable in the quest to find it, he says. He gives three goals for powering personalisation with Gen AI:
“
You can create a magical user interface if you can pull all these three things together.
”
Booking wants to be grounded in user problems. Joe says travellers still have the same needs as 25 years ago, they want a good deal, they want to talk to experts, and they want someone they can depend on when they're away. Booking.com can leverage Gen AI to create a better user experience by revisiting the customer journey. The Booking customer journey is split into six stages and Joe explains that the company has a lot of research on travellers, it can even monitor their emotions. For example, customers find discovery and planning almost as exciting as going on a trip.
He gives the example of the many different factors that make up trip intent. He says Booking.com will build a system that will understand these factors and iterate, be carried through the user journey, be specific when it needs to be a step back and be broad when it doesn’t.
Joe also demonstrates Booking.com’s AI Trip Planner, which has been live in the US and other English-speaking countries for about a year. There were loads of learnings, he says. For example, to begin with, the hallucination was “out of control”, and Booking ended up building its own orchestrator.
He offers some advice:
Joe’s mistakes to avoid are as follows:
“Embrace AI, you will learn so much,” he concludes.
Take a look at Five things we learned at the Pendomonium+#mtpcon roadshow – Amsterdam 2024 for more insights from the roadshow and be sure to come back to the Mind the Product homepage and Pendo blog in the weeks ahead to read more recaps and full write-ups of all the roadshow talks.
