While there's no doubt advancements in artificial
intelligence will transform travel management companies, expect to see it
remain in the assistant's chair, not the driver's seat, this year.
By now, you've probably heard the stats around the meteoric
rise of ChatGPT and its reaching of one million users with just five days of
being publicly available—a feat that took Instagram about two months, Facebook
about 10 months and Netflix more than three years. While users are a different
measurement than usefulness, it didn't take long for TMCs to tap its abilities,
either. Navan, the former TripActions, in February announced it had integrated
generative AI APIs across its infrastructure, both internally such as helping
to write code and externally in making a conversational experience within its
virtual assistant Ava.
"In just two years, the leading players in T&E will
be either established companies that have already embraced ChatGPT or new
startups that bring innovative experiences and capabilities to the market,"
Navan co-founder and CTO Ilan Twig predicted at the time of the company's
generative AI announcement.
At The Beat Live in December, Fox World Travel chief
information officer Sam Hilgendorf said the TMC has worked with Microsoft and a
local consulting firm to "bring in ChatGPT and put walls around it"
to develop Scout, an AI tool that can help agents out with questions, including
complex processes, they come across in their work. It's fed with data from the
many standard operating procedures that Fox has written and operates under, he
said.
For TMCs, embracing generative AI is more than just keeping
up with those "innovative experiences and capabilities" that Twig
mentioned. There's a financial motivation as well. In American Express Global
Business Travel's third-quarter earnings call last November, CEO Paul Abbott
said 40 percent of the TMC's costs come from employee costs of servicing
customers in voice channels, and generative AI is a ripe opportunity to trim
those costs.
Like Navan, Amex GBT is tapping AI for internal purposes
such as coding as well as in customer-facing processes including automating
emails and chat, Abbott said during The Beat Live. However, Abbott also showed
a degree of caution, saying that while AI is "scaling nicely" into
chat, Amex GBT is "not trying to use AI too aggressively" there, he
said.
"The last thing I want is to put someone in 'bot hell,'
where they say 'My flight is canceled,' and it says, 'Oh, that's great!'"
Abbott said. "If that means we move a little slower in some areas or
differently in some areas, the most important thing is the travelers'
satisfaction."
Already this month, CWT announced that it had added AI
capabilities to traveler messaging on its MyCWT platform in 56 countries, with
which travelers can chat with an AI virtual assistant powered by Microsoft's
Azure OpenAI platform and move on to a live agent if the virtual agent is
unable to help them. Now that there has been sufficient time to test such
capabilities—CWT tested the virtual agents with a small set of users before the
wider rollout—expect a flurry of such announcements from TMCs large and small
this year.
There are pitfalls to generative AI, such as
"hallucinations," when the technology lacks the data to answer an
inquiry and simply makes up an answer. Scout, for example, gave a name
Hilgendorf had never heard of when he asked it who Fox's CIO was, and there
have been more humiliating examples with the technology at large, such as when
lawyers used it in court only to have it discovered that it simply invented
case precedent, Hilgendorf said.
Beyond just getting things wrong, there's also the potential
of a "dark side," where the technology picks up on our "biases,
greed and fears," Hilgendorf said. "We're going to see both
experiences over the next several years, so find partners you know are
ethically responsible."
As such, expect ethics around AI implementation in travel to
be more front and center this year, and be wary of announcements that appear to
over-promise AI capabilities on the TMC side. AI fundamentally is only as good
as the data it can access, and some of the data technology that will be crucial
in marrying AI and travel—decentralized identity, for example—is still a work
in progress.
Travel Tech Consulting president Norm Rose said truly
autonomous agents, that are truly able to manage requests and effectively book
travel without human intervention—either through calendars, messages or a
ChatGPT-style interface—are still several years off. While there are already
companies today are working with those capabilities, he said it is likely a
five-year timeframe before they are a significant player.
That is, however, the ultimate direction, he said.
"It has to be trustworthy, and it has to be accurate,
but we are going to migrate toward this," Rose said. "It's going to
take some starts and stops, but this is where the future is going."