Putting generative AI to work

Generative artificial intelligence (genAI) isn’t a new form of technology. But when its latest and greatest iterations — Open
AI’s ChatGPT, Microsoft’s Copilot and Google’s Gemini — barreled onto the technology scene more than a year ago, it was with great fanfare and much hype.

That hype extends to the world of corporate travel, where TMCs are mostly deploying relatively uncomplicated genAI use cases, but are testing and preparing for more complex applications of the technology down the road. 

“There are tremendous opportunities for generative AI, and I think there are plenty of low-risk opportunities to start testing it,” said Heather Wright, vice president of global product marketing at Atlanta’s BCD Travel (No. 4 on the Travel Weekly Power List). “I’m sure everybody’s in a race for the best idea.”

That race is expected to yield genAI uses that Kevin Fliess, chief marketing officer of the Global Business Travel Association (GBTA), believes will be “a game-changer” for the corporate travel world. 

“I have to remind myself this is as bad as AI will ever be, and it’s already mind-blowing in terms of what it’s able to do,” Fliess said. “Our industry has a history of embracing innovation: web, mobile, the advent of the internet and online booking tools, the mobile revolution, search. AI is another chapter, but AI is as big as what we saw with the advent of everything moving onto the internet. It’s as profound and transformative. I see this as a sea change.”

Today’s genAI use cases in business travel are generally limited to basic applications like answering queries about corporate travel policies and email parsing. But excitement is high about what the future holds, such as supercharged customer relationship management systems or AI assistants.

“We’re very optimistic about what genAI can do for the travel industry, what it can do for us,” said Steve Glenn, chairman and chief visionary officer at Executive Travel (No. 67 on the Power List) in Lincoln, Neb. “And I’m talking just at the surface level, just simple areas that can make life better for our agents, for our customers and for us. I’m very excited.”

‘Period of experimentation’

The business travel world is in what Fliess called “a period of experimentation” when it comes to genAI.

Many GBTA-member TMCs are brainstorming potential use cases of genAI, he said. Teams are picking several to develop and test.

Executive Travel, for example, is working on integrating genAI to help travelers query and better understand their travel policies. Glenn called it “low-hanging fruit,” the kind of genAI application that is relatively quick to implement. The integration is expected to launch in the fourth quarter.

The idea started with a simple premise, according to Glenn: “Nobody reads their corporate travel policy.” As a result, compliance is “iffy, at best,” he said.

GenAI can step in, acting as a simple interface where travelers can ask questions and engage with their travel policy without having to read through what can be a lengthy document.

How, exactly, the technology will be implemented depends on the client, Glenn said. Some might want it to exist on the company’s internal network, while others might want it to exist fully outside of its systems. Different security teams will have their own preferences.

BCD is also working on applying genAI to simplify searching through a travel policy, Wright said. 

Additionally, BCD is testing a chatbot for booking, rebooking and canceling travel. While BCD does have a messenger product that uses both AI and human agents, the chatbot that applies genAI would fully automate those processes.

However, Wright said, BCD is taking a cautious approach with extensive testing. The product needs to be nearly 100% accurate before it moves beyond the testing phase. Right now it’s close, at about 99% accuracy, “but that 1% matters,” she said.

GenAI use cases

American Express Global Business Travel (Amex GBT, No. 3 on the Power List) earlier this year assembled a team to increase the New York-based TMC’s focus on AI. Its first task is adopting genAI in various areas.

Marilyn Markham, vice president of engineering and AI strategy, said the team has a number of genAI proofs of concept that are maturing quickly.

The first is an email AI assistant. Markham said Amex GBT owns its online booking tools, unlike many TMCs, and has an engine in place that can find an offer for a traveler already compliant with their travel policy before presenting it. But that requires the traveler to submit a structured request via a web interface.

“The gap that AI is helping us bridge is to actually go from a completely unstructured block of text, which is your email request, and actually create that structured request that allows us to then use these pre-existing platforms,” Markham said.

For agents, it saves time — that email would have otherwise gone to an advisor, who would need to extract the information. Markham said Amex GBT is working on calculating how much time it would save at scale. Right now, the AI assistant is more than 94% accurate.

That kind of application would not have been possible before recent advancements in genAI, she said.

Another area Amex GBT is targeting is airline fare rules. GenAI can succinctly summarize terms and conditions, and the user can interact with it, she said. That’s helping accelerate the training of new advisors.

The TMC also maintains knowledge bases that are both agent- and customer-facing. A genAI proof of concept is helping extract answers to questions faster than when a person searches through the information themselves.

“That’s one of the fortes of generative AI, is that it does a great job at search-and-find in large, core pieces of data,” Markham said.

Finally, Amex GBT is working with a partner to offer live translated captions during its virtual events, said Markham, who noted the importance of working with third-party partners for some use cases.

‘Felonies’ and ‘misdemeanors’

Concerns about AI accuracy have made the business travel side of the industry a bit slower to embrace genAI than others, with fears that even a small error can have big consequences.

TMCs have been using different forms of AI technology for years, but BCD’s Wright said genAI brings with it its own set of challenges.

“New technology is exciting, but it’s also scary, because you don’t know what you don’t know,” she said. “We’re approaching it cautiously.”

Her team describes genAI errors, known as hallucinations, as either “misdemeanors” — minor errors that don’t result in much harm — or “felonies.”

“It’s a felony if you cancel the flight of a CEO who’s going to conduct some very important business,” she said. “We’re looking at low-risk but high-value ways to integrate AI. Not just generative, but all AI — generative is no different than that.”

Important considerations also include security and privacy, and vetting third-party vendors when they’re used, she said.

Late last year, the GBTA surveyed members about AI and its impact. While about one-third of respondents were excited about the technology’s promise, Fliess said, another third said it was too early to predict its impact. The final third had a negative outlook.

“They saw it as a necessary evil, or they were concerned about AI due to ethics and privacy or other considerations,” he said. 

However, Fliess said he suspects if a similar survey were fielded today, months later, it would indicate more adoption of AI and genAI among TMCs.

What does the future hold?

BCD is currently finalizing its internal AI policy, something many companies are in the midst of doing, Wright said. And for BCD in particular, it’s helping shape the TMC’s procedure on partnering with third-party technology companies.

Wright outlined some of the questions the policy will define: “What are we willing to use and not use? What kind of protocols do you have to go through to procure an AI company? Who’s allowed to use it — who has access to it?” 

Wright believes a key factor in genAI’s use cases in the future will be its ability to help humans do their jobs better. She’s specifically watching spaces like policy development, automating data and even how to better use data.

“We’re not looking to get rid of people,” she said. “We need humans in our industry. That’s never going to change.”

Executive Travel’s Glenn said he believes the next step for genAI is a vocal interface (like Amazon’s Alexa) with a database like a GDS enabling users to make reservations. GenAI specifically will allow for more “ebb and flow” between such an interface and the user.

Glenn is also looking to genAI to help make a “CRM on steroids.” 

“This is an accelerant,” he said. “This is like gasoline on fire, and that’s what it’s going to do to CRM and data and the ability to provide comprehensive information at the point of sale for agents to serve their customers better.”

GenAI can also quickly offer support and service for travelers who face trip disruption, Glenn added.

While there are obvious uses for genAI in corporate travel, Markham sees a lot of potential in leisure travel. She pointed, in particular, to Expedia Group’s Romie, a new AI “assistant” the company (No. 2 on the Power List) is testing.

Part of the promise of Romie is its ability to listen in on group chats, offer suggestions when appropriate and store talked-about topics like hotels on the Expedia platform for ease of booking.

“I thought that was a really interesting way to bring AI into the community of conversation around a trip and have a very dynamic assistant there,” Markham said.

She also sees great possibility in using genAI to create video previews of experiences specific to what people have booked.

In business travel, having an assistant that joins a chat could be very helpful, Markham said. It could also offer a level of personalization travelers would appreciate.

In essence, she said, an AI assistant could better complete traveler profiles. 

Historically, business travelers haven’t been the most forthcoming about personal likes and dislikes at work, but AI could parse natural language for more information and a more accurate profile.

“Data still rules any AI application you have, whether it’s machine learning, [natural language processing], genAI, whatever,” Markham said. “You still need just really, really good data to feed it, because ‘garbage in, garbage out’ still applies. I think that people think sometimes that AI is magic. It really isn’t. It’ll get you the wrong answer if you give it that data.”

Update: This report was updated July 25 to clarify the accuracy American Express Global Business Travel's AI assistant; Amex GBT said is more than 94% accurate.