Op Ed: Jeff Klee On Artificial Intelligence In Corporate Travel

Jeff Klee, CEO of TravelPerk’s AmTrav, looks back at his early days in the industry for guidance on navigating the hype and realities of artificial intelligence in travel management.

Back in the Stone Age when I started in travel, there were situations where, for one reason or another, we’d actually have to hand write airline tickets. This was a huge pain in the ass, requiring someone to transcribe a bunch of crazy codes and fare breakdowns. To save our agents time, we built an application that could print correctly positioned text onto a ticket booklet. Over time, we iterated to make that application useful for more scenarios. But at some point it became obvious that paper tickets were going away, so we stopped enhancing a lame duck tool.

I think about that automated ticket writer a lot these days. Industry leaders who I respect – like Corporate Travel Management managing director Jamie Pherous in a recent interview – are raising questions about whether online booking tools might also be headed for obsolescence. In forums and at conferences, you hear that AI will kill not just OBTs, but also other mainstays of our industry, like human customer support, GDSs and maybe TMCs altogether. As someone who spends a lot of hours dreaming up ways to enhance booking tools, I can’t help but wonder: Am I wasting my time? Are all of us who are working to deliver traditional travel management services – even those employed by the most “modern” TMCs – really just skilled carriage makers about to enter the age of the automobile?

The argument against TMCs in an AI world is that, with an army of automated agents, travelers and companies won’t have the pain points that necessitate travel management in the first place. With agentic AI, every traveler will have their own bot that, with a full understanding of corporate policies and traveler preferences, can book trips directly with suppliers, organize receipts, facilitate reimbursements and subsequently crunch the data every which way imaginable. Ironically, the strongest argument against this outcome may be the very reasons that travel management is so ripe for disruption: the closed systems, arcane technology and fragmented unstructured data which TMCs have learned to navigate but which might be challenging for large language models to get at, at least for a while.

If we focus more narrowly on OBTs, the idea of an AI agent knowing each traveler so well that it can find and book the perfect trip automatically is compelling at first blush, and technically possible. But will travelers prefer that? Or will most still want to see their options and make their own choices?

If AI agents do an end run around the intermediaries and connect directly to the suppliers, it’s conceivable that GDSs might become unnecessary. On the other hand, GDSs have lots of data, and data is exactly what large language models are likely to compete on as computing power becomes a commodity.

And what about human customer support? Conventional wisdom says that’s definitely going away. Even in a pre-AI world, travelers want the power to self-serve for most things. But when disruptions occur and solutions aren’t simple, a human expert capably using AI-powered tools to deliver service with a side of empathy and even humor strikes me as something that will still be welcome and valued.

At some point, the hype for AI in business travel will die down, we will cut through the crap and the story will be what it always is – value. Who is doing what to actually make life better for travelers and companies?

I roll my eyes when I hear companies announce AI strategies, as if AI is not the means to an end but the end itself. Travelers don’t care if the technology they’re using is AI-powered, blockchain-enabled, cloud-native or mobile-first. When they need to change a flight, they just want it to work.

We are at this extraordinary moment in history where things that never used to be possible now suddenly are. It would be a tragedy if we as an industry conclude that the best use of AI is to simply deliver the same value we always delivered, in the same way, just more cheaply. There is also a danger that resources will be wasted on gimmicky offerings that create headlines and showcase technology but don’t create new benefits that businesses actually care about.

It seems likely that AI will mark another fork in the road where a host of startups emerge with blank slates and fresh approaches that challenge the conventional thinking about the mechanics of how travel should be managed. Successful or not, that will be disruptive and healthy.

A full two years after we decided to stop enhancing our ticket writer, the adoption of e-tickets was dragging and we were still handwriting way too many tickets. In retrospect, our decision to deprecate the application so early was a loser. Over those two years, it probably cost us hundreds of avoidable agent hours. But handwritten tickets did eventually go away. Is there a lesson there as we contemplate the value of “old work” on the brink of an AI revolution?

Threats of replacement notwithstanding, it makes sense to me to keep enhancing booking tools, making traditional support better and thinking of new ways to improve the lives of companies and travelers with current tools, workflows and technology. There are obviously ways to take advantage of AI already, and many companies are doing it, but we should do so where it genuinely adds value, not just for the sake of it.

AI is neither hoax nor panacea but there’s no question it will raise the bar. The future for business travel will be awesome in ways that we can predict, and in ways that we cannot yet imagine. Whether the service providers that deliver the most transformative change are incumbents or startups is still TBD. But, however it plays out, it will be a fun story to watch with twists, turns, unicorns and cautionary tales. Casting isn’t finished yet. Plenty of big roles are still open.

This Op Ed was created in collaboration with The Company Dime‘s Editorial Board of travel managers.