Since its emergence, travel companies have been experimenting with artificial intelligence (AI).
But any new technology comes with challenges, and that remains true with AI.
Shilpa Ranganathan, chief product officer at Expedia Group, which moved quickly to implement AI, opened up about Expedia Group’s AI innovation process.
She discussed lessons from early-stage chatbot efforts and AI hurdles including development speed, employee buy-in and more.
Chatbot reality check
Two or three years ago, as many travel companies did, Expedia Group began testing chatbots, aiming to enable conversational booking. Publicly, Expedia Group launched the now-defunct AI assistant Romie in 2024 and has made continued development efforts.
“It was an interesting learning experience, because travelers' trust is very important,” Ranganathan said at Skift’s Data + AI Summit in New York last week.
“When that trust breaks, it's going to be very difficult,” she said. “And so when we shipped this chatbot, we tried to see what kind of questions people were asking, and slowly we realized that if we weren't grounding our generic responses in our data, in our inventory, in our rates of reservations, customers were just like, ‘this is not working for me, I'm going to leave.’”
That served as a reckoning for Expedia Group, she said.
Following the pullback, she said her team has spent the last year focused on the AI framework and the AI platform that needed to be built.
Expedia Group has keyed in on evals and is using taste and judgment to assess a model’s output. She also said the travel company is focused on building architecture that scales so that they can multiply the number of agents they ship.
“From the outside it looks like, ‘Oh my god, it's so slow, why aren't they shipping things faster?’” she said. “But creating that foundation has really helped us now to not have to build a penthouse on a foundation of sand.”
Low barrier to entry with AI testing
While taking the time to build a strong foundation is necessary, the barrier to begin new projects related to artificial intelligence (AI) is very low, Ranganathan said.
That’s a major upside of the technology.
“We want people across all our functions to be able to experiment and learn, but to me the true test is in customer adoption outcomes,” she said.
But that doesn’t mean all experimentation sticks.
“I love that employees want to try new things,” she continued. “But we have a huge responsibility to make sure that customers actually appreciate the fact that this is going to solve a problem for them.”
Travelers don’t care whether Expedia Group is using “AI, BI, CIA,” she said—they just want to get their travel booked and squared away.
Change management
Appealing to the traveler isn’t the only hurdle amid AI innovation: Change management has to be a priority.
Expedia Group, like other travel companies, needs to secure companywide investment from employees. Ranganathan said she fundamentally believes that AI is not a technological challenge but an organizational one.
That’s manifested in thinking intentionally about culture management, Ranganathan said, noting she spends a good amount of her own time handling that side of things.
"How do you get everybody in the organization to feel like they're empowered to take risks and experiment? How do you give them the right tools?”
That’s complicated by the fact that access to AI tools is expensive. But if a company doesn’t democratize, employees are less likely to change their daily behavior. So it’s been a priority, Ranganathan said, to set up an environment with the proper “scaffolding” to strike the right balance there.
But ultimately, Ranganathan reiterated, it all circles back to customer response.
“You can experiment all you want in AI,” she said. “But if it's not translated to a real business problem, it's very unlikely that the employees are going to make progress in that area.”