AI in Air Pricing: Why It Matters for Buyers

A little more than a month ago, Delta Air Lines president Glen Hauenstein on an earnings call commented about the carrier's use of the product Fetcherr and AI in revenue management, stating that it was used on about 3 percent of its domestic network by the end of the second quarter, and that the goal was to have it used for 20 percent of the domestic network by the end of the year.

Those comments set off a firestorm of coverage and concern. Several media reports indicated that Delta was using the AI tool for individualized pricing by way of personal information. 

The U.S. Department of Transportation Secretary Sean Duffy said the agency would investigate anyone who does that. Three U.S. senators—Mark Warner (D-Va.), Richard Blumenthal (D-Conn.), and Ruben Gallego (D-Ariz.)—on July 21 wrote a letter to Delta expressing concern about the carrier's use of AI to set "individualized fares," and noted that the practice "present[s] data privacy concerns, but will also likely mean fare price increases up to each individual consumer's personal 'pain point' at a time when American families are already struggling with rising costs."

But the carrier issued a statement, an item on its News Hub and a July 31 response letter to the senators stating that "there is no fare product Delta has ever used, is testing, or plans to use that targets customers with individualized offers based on personal data." The airline further noted that "a variety of market forces drive the dynamic pricing model that's been used in the global industry for decades, with new tech simply streamlining this process. Delta always complies with regulations around pricing and disclosures."

The chatter did not die down, however, and on Aug. 11, Delta SVP of global sales Bob Somers sent an email to Delta Business customers reiterating the carrier's denial of using fare products that target customers with individualized prices based on personal data, along with a copy of Delta's letter confirming as much to the senators. 

Somers again addressed the topic during the carrier's Aug. 20 Showcase event for corporate customers, stating misinformation had been spread and repeating Delta's response.

While this issue might not seem that big of a deal—AI has been used in revenue management for years—it does harken back to early concerns when New Distribution Capability was introduced by the International Air Transport Association more than 10 years ago. Some critics raised privacy worries and that the use of NDC would increase fares because offers would be based on customer profiles. 

IATA countered that NDC was intended to accommodate either an anonymous shopping request or customized offers if a customer provided more information about themselves. The goal was to have airlines offer products that more closely fit consumer needs and would be more attractive—a win-win for the supplier and the customer.

We know NDC did not exactly take off as anticipated, though the train is picking up speed, and much has been noted that NDC and its partner continuous pricing can actually lower fares and not just raise them. 

So why the ongoing noise?

Most corporate travel sources BTN spoke with on the topic of AI in air pricing gave what amounted to a collective shrug. 

"It feels like just a new tool to do the same thing that's already been done, and maybe make it more efficient or more precise," former travel manager and current consultant Colleen Kearney said, adding that airlines have long taken historical behaviors, looked at patterns, supply and demand, and priced accordingly. 

ZS travel and meetings manager Suzanne Boyan added that "nothing about this bothers me because this is where all industries are headed, … and I don't want our partners to go bankrupt," she said. "They should make a healthy profit. That's part of what a partnership is, and if this helps them to do that, to maximize their revenue while still being fair to corporate clients, then I don't see anything wrong with this. It's going to come down to transparency, to communication."

AmTrav CEO Jeff Klee said he takes Delta at its word, and since the carrier doesn't have an NDC program yet and it's not experimenting with personalized offers, that means they are filing fares with ATPCO, he said. That process used to be manual, but Klee took Delta's use of Fetcherr to mean that the carrier "is now using AI to more smartly pick what fares they want to file," he said. "If everyone's going to get hysterical every time a job that used to be done by a human is done by an AI bot, we're going to be in for a pretty rough few years."

Increased Sophistication

Others, however, while agreeing that the use of AI is nothing new in airline revenue management, said that the increased use of AI in pricing is becoming more sophisticated and will affect travel programs. 

Advito managing director and BCD Travel SVP April Bridgeman said during a session at July's Global Business Travel Association convention in Denver that monitoring and adjusting air contracts was key, particularly when "you can't predict how your travel patterns will change, what's happening in the marketplace and which new tools and technology airlines are investing in, whether it's NDC or now something like Fetcherr, which is probably going to make every buyer's life miserable with their new AI pricing scheme."

Bridgeman further explained to BTN that the use of such tools will make the role of a travel manager or a sourcing professional that's involved in air more challenging. "And it's not just air, it's anywhere that pricing becomes more sophisticated, and that will be in the future primarily driven by AI," she said. "It does make things less transparent and predictable and less measurable for travel managers."

One way to measure the value of an air program is to benchmark using filed fares in the global distribution systems. But that is beginning to change with NDC and dynamic and continuous pricing. 

"Supplier leverage increases because the airline can present a discount that sounds substantial, but that could be anchored to in some cases an inflated personalized price," Bridgeman said. "What AI is going to be able to do is make it for an airline to serve different fares to different buyers or booking channels based on things like search history, booking behavior, route demand, even device or location data. And those things are not necessarily personal data. They're things that we already share just based on our behavior on our phones."

Colluding AI Bots?

Traveler identity network Travlr ID founder Gee Mann penned an Aug. 5 post on his Substack about recent research that found AI agents had begun to collude on pricing. Researchers "simulated financial markets and watched AI bots, designed to compete, begin cooperating," without talking to each other, he wrote.

"I don't think it's t4e privacy part or the personal profile part that is the issue," Mann told BTN, adding that the AI model doesn't need to know the person, and the type of AI model used doesn't matter. "The ultimate goal is to get maximum revenue. The incentives are always going to be the same, it doesn't matter the technology or the tech stack."

Using the research example, Mann said that if one supplier drops too low, somebody else is going to go up. If someone is too high, someone else will come in the middle. Eventually, the models will start to price at the same level to not be volatile and to get maximum revenue across the board. "And that's where the collusion started happening," he said. 

What's needed, Mann said, is backup on how those models got to their prices, the outcomes. "If we can get an audit trail to how the outcomes are being derived, then that's a different scenario," he said. "But that's not what's happening right now. I'm concerned that there's not enough regulations around this either."