There
are about 14 places named Birmingham in the U.S. In the new world of travel booking using
artificial intelligence (AI) a traveler conceivably could be sent to the wrong place.
The scenario was presented by Travelport CTO Andrew Jordan during
a session at TravelTech Show in London last week on travel industry
preparedness for agentic commerce.
Jordan said the industry is confronting huge change in how
products and services are discovered and distributed as it moves from being
deterministic to probabilistic.
Search and booking pathways were once form-based, with trips built
around origin and destination, now they are becoming more conversational with
the advent of AI.
He added that the huge variations in prompts and what the
platforms return create a clear role for companies such as Travelport.
Examples Jordan provided included ‘I want to book a cool holiday in
Thailand,’ to ‘the least risky way to recover a stranded business traveler on a
business trip that stays within policy.’
“Those instructions need some way of being diagnosed in a
completely confident way, because the risk of you getting that wrong is pretty
significant,” he said.
“Imagine you've got a firm line here, and on this side of the line
is everything to do with discovery, itinerary building and inspiration. And
then that line is basically where you commit, you hand the money over and
things have to be precise. The thing that you have to get right, and this is
where the sweet spot sits for us, is that you can't get it almost right.”
Figuring out where that hand-off takes place is a big question for the
travel industry currently.
Agentic adoption barriers
Harry Herbert, enterprise GTM lead for Anthropic acknowledged
how hard it is for the travel industry to move to agentic because of legacy
systems and siloed data.
“There are tasks that are being made agentic and workflows that
are fully agentic within travel customers I work with. But there are a lot of
systems, there are a lot of layers, there is a lot of legacy and tribal
knowledge when it comes to codebases. What I would say is think about how can you use the
tools we give you to get to a resolution faster—to understand those codebases,
to upskill your workforce, to create better processes and integrate better with
codebases.”
He
added that safety and security are also critical in the booking journey and
stressed the importance of building trust through guardrails. And while the jury
is out on whether brands will disappear and consumers will just trust agents,
the landscape is in for a significant shake up.
“That suggests that there is an almost perpetually high level
of trust with the two big OTAs—and that whole world has just been turned on its
head already,” Jordan said. “The library of trusted brands is going to really
change. And as people get more comfortable with using agentic tools to do that
work for them, I’m sure if we walk round here in two years let alone 10, the names of the companies will look quite
different.”
He
also touched on what intermediaries including agencies and the GDS need to do
to survive going forward. He said that for three decades GDSs had provided
content coverage and incentives but that is no longer enough.
“If
you can unpack where value gets created and take the heavy lifting and place it
where it belongs, where economies of scale can apply which is where we sit," said Jordan. "You’ll
naturally see the people who don’t do that start to wane and die off.”
Consumer confidence
But neither
trust nor technology is viewed as the major barrier to agentic commerce, according to panelists. They believe consumer readiness is a bigger challenge. Jordan
compared it to autonomous vehicles and technical capability versus the
confidence levels.
“The
tech is good,” he said Jordan. “It’s because on the whole society is not quite
ready yet. I don’t think it’s the tech, it’s society that needs to adjust its
confidence level that what it’s telling you is correct.”
Herbert also
said the technology capability is good but that businesses need to “dissolve
the noise,” to understand what it means for them and find use cases that make
sense.
Meanwhile, Jason Cincotta, CEO of Kismet, a PhocusWire Hot 25 Startup for 2026, said consumer behavior is a barrier.
“You
could probably book travel right now agentically on some hotel websites but that’s
not how people plan trips,” he said. “They don’t go to one website; they don’t
look at one hotel. The chatbot you are going to interact with to plan agentic
travel or executive agentic travel will be your own. A core belief we have is
that consumers want to bring their own AI. Right now Claude won't let you check
out in Claude—for good reasons. The first one across the finish line is
probably going to be Google.”
Cincotta added that independent hotels need to be ready for consumers' own agents to book travel not the hotels' agents. Hotel marketing teams currently think the rate of AI conversions is around 5%. Then they implement technology which reveals that it can be up to 50%.
A further barrier lies in the “plumbing” or protocols such
as MCP, ACP and UCP that enable agents to act for consumers according to Cincotta, adding Kismet wants to be the "pipe fitter" for hotels and short-term rental properties.
“They're not the same protocols your existing website runs on—the
important takeaway here is that there is an emerging parallel internet that runs
on a completely different set of protocols and that agents will operate on.
Everybody in this space is going to have to decide whether this is buy or build
for them,” Cincotta said.
In the meantime the AI platforms will develop further and deliver better capabilities. Herbert shared that there will be a shift in where users interact with AI models such as Claude.
" Not just a chat bar, the web, different channels— becoming an agent across platforms, not siloed in one environment."