Artificial Intelligence will drive growth in travel but human expert intervention is required to make sure it is reliable, according to leading Chinese online travel agent Trip.com Group.

Delivering the opening keynote at this year’s ITB Asia in Singapore, James Liang, co-founder and executive chairman of Trip.com Group, said AI won’t make travel cheaper.

He said it will make everything else cheaper giving people more disposable income which they will spend on luxury items, like travel.

“AI and automation is going to make the travel industry much bigger,” he said, “not because it’s going to make travel cheaper.

“It’s going to make everything else cheaper and travel, relatively, expensive. When people have more money and discretionary resource they will spend more on travel.”

Liang said travel satisfies people’s higher intellectual and emotional needs so demand rises when people have more time and money.

He said AI will make travel easier throughout the booking and experience journey from conceptualisation to consumption.

“People spend hours trying to plan a trip,” he said. “We find with AI we can make this not very pleasant or efficient aspect of travel easier, and maybe even fun.”

Liang said large language models like Microsoft’s ChatGPT and Google’s Bard, that have emerged recently, have solved the problem of understanding the question.

The technology is also capable of “reading” all the content that has every existed, including travel content, in the search for answers.

But Liang said there are still two significant pieces of the puzzle that large language models miss: real time data and insights, and independent expertise.

“In order to do travel planning you want real time data and the latest trends including pricing and inventory, what’s cheap, what are the deal, the hot trends, events coming up,” Liang said.

“The other big piece it’s missing is input from independent experts. Are the sources of information reliable, are they [recommendations] independent, are they generated by AI robots with a conflict of interest? You need experts.

“Our strategy is to provide those two missing pieces by integrating the capabilities of large language models with real-time data and also with human inputs from experts. With that, we think we can improve the planning capabilities of our OTA AI agent.

“We want to provide customers with the latest trends and deals. We also have experts to make sure that recommendations are reliable and up to date. I think human input is still very important.

“Having large language models and thousands of human consultants together will make sure that the answers are reliable and correct.”

Trip.com Group has developed its Trip.Best rankings for the most highly rates hotels for its users to use to based their decisions on under a number of specific themes, like family travel or beach or city trip.

Although generated automatically, the lists are checked by human experts to ensure they are both independent and up to date.

Liang said the trip building stage, following recommendation, is more complicated, particularly when multiple cities and flights are involved.

Trip.com Group has also developed technology for its Trip Genie itinerary planner that is able to take account of time and space and can adapt the recommended trip based on specific customer needs.

However, Liang said large language AI models are not reliable enough yet, so human intervention is required at this stage to verify the output of the technology.

“I think we believe we have the right strategy. But that’s still very much work in progress. It’s still something no one has done in this industry.

“Our engineers love to play with it, but there are only a few months into getting the best out of it. Over time it will continue to improve.

“We are experimenting with different strategies. I’m not saying we have the right formula but it’s a very exciting direction and I hope to see many other companies trying different things. We want to help the consumer generated the perfect trip, just like a Genie.”