Here's how AI helps business travelers. Source: Shutterstock

Here’s how AI helps business travelers. Source: Shutterstock

AI can drive savings for business travelers says AMEX GBT

SAVING money and time is one thing artificial intelligence (AI) can do well, and American Express (AMEX) Global Business Travel (GBT) is using the technology to deliver savings every time clients book a hotel room.

In an exclusive interview with Tech Wire Asia during a trip to Singapore, Chief Information and Technology Officer David Thompson spills the beans on the company’s AI-based booking solution.

As AMEX GBT books hotel rooms for clients, it follows travel policies handed down to them (preferred hotels, agreements, etc). However, as most hotel’s inventory changes frequently, opportunities to move the traveler into a lower cost hotel room with the same amenities often arises.

“This is where we found the opportunity to leverage AI. We trained AI to understand where the hotel room is and what the amenities are.”

The descriptions for the rooms and amenities can vary dramatically from region to region. What’s called high-speed internet in Europe, for example, is called WiFi in Asia.

“We trained the AI engine to understand all the content globally, and now have the ability to constantly re-shop the hotel booking on behalf of the traveler. If we find a better opportunity to move the traveler into a better economic situation for the corporation, we present that opportunity to the traveler.”

AI helps AMEX GBT scan massive amounts of content, looking for like-for-like or better opportunities for travelers, all to ensure that the company is able to optimize its client’s spend in the travel management space.

The solution took two months to build before the initial deployment, and from the initial baseline, helped the company create an improvement ranging from three and seven percent over the first few iterations.

One of the things that Thompson’s team learned early on is that in order to make sure that the AI model was robust and efficient, the data set needed was much larger than the company has originally estimated.

“We went through historical data for a one year period and used AI to look for opportunities (to save costs) as against opportunities found by human agents, and we saw significant improvement during that time frame.”

Obviously, although in production, the AI model is constantly learning. This means that results are constantly improving — however, sometimes, the actual cost-savings depend on the client that is being served rather than the AI model.

Sometimes, strict corporate policies prevent AMEX GBT’s AI solution from actually making changes to traveler bookings, rendering it impossible to deliver cost savings.

However, Thompson’s team is often able to show how much economic value they would able to bring if the company is willing to alter its travel policies — and clients usually accept, proving AI can really make a difference.

In the future, more companies could be using AI, but the two things they need to keep in mind while developing their model is that the data set that is used to train the solution must be large enough to ensure it is robust, and that the model is tested for results against a strong baseline before it is deployed.