Travel companies are beginning to discuss artificial
intelligence (AI) costs—think tens of thousands of dollars monthly—but that is
likely a drop in the ocean of what’s coming.
While having teams begging for more tokens at the end of the
week can be seen as a positive, at some point the investors putting
billions into AI platforms will want their money back.
At the Short Stay Summit in London this week Steve Schwab, founder and CEO of Casago and Graham Donoghue,
group CEO of Forge Holiday Group discussed the challenge during an AI-focused session.
Donoghue shared that the company is not far off spending between
$30,000 and $50,000 a month with Anthropic on tokens and that he is considering
creating a role in finance for a token optimizer.
Schwab believes there is an “entire token economy” coming.
“Tokens right now are probably 20% to 30% of the cost it takes
to cover that compute cycle. We’re going to be in a situation where investors say, 'we have to make money,' and the compute cost of running these cycles
is going to be really expensive,” he said.
“There are ways to optimize that, you start
thinking about arbitrage and different ways we’re going to afford this. But
unless there is a deep change in the way the architecture of AI works, we’re
going to struggle to make it affordable for everyone,” Schwab added.
Despite the current expense and likelihood of it growing, the
rental companies are seeing value from their AI investment.
Tokens right now are probably 20% to 30% of the cost it takes to cover that compute cycle. We’re going to be in a situation where investors say, 'we have to make money', and the compute cost of running these cycles is going to be really expensive.
Donoghue said Forge started its AI journey more than a
year ago. He wanted it to be used across the whole business and not viewed as a
technology project. In the beginning, value was derived from “repetitive
and monotonous processes” such as image retouching and call transcriptions, but
that has evolved.
The company developed an AI tool called Vivid Stay where it
trained the model to turn static images of its properties into videos and then into
drone-like footage. Donoghue described the product as a good example of how it
is leveraging AI.
“AI does that for us; it takes 25 minutes. It would probably have cost $2000 or $3000 to do it before.”
He also said the company has moved from using AI for
simple things to “complicated algorithmic processes.” Tasks that
would have taken weeks and months in the past now take days, he added.
Learning curve
But it’s not all success. There have been some bumps in the
road and learnings along the way.
Donoghue shared that there have been lots of failures, including an assumption that users would love conversational search, which the
company is not really seeing so far.
A further learning, he said, has been ensuring employees are
fully on board with using the technology.
“I got quite annoyed as the CEO when people could not see
the future as I can. I expected that if we gave people the tooling they would adopt
it, but in many cases we had to say this is non-negotiable. Culturally that was difficult.”
For Schwab, the learning was around structured data, which
meant outputs were “sub-optimal.” He added that while internally the company has made changes and improvements with AI, it has struggled more with customer-facing initiatives.
He also shared some advice for smaller operators.
“Don’t spend a lot of time vibe coding something that you
think could be useful in the future. We’re seeing people get the vibe code up
to 80% and then just can’t finish it up because they don’t have API access or
access to the proper data. It’s a waste of time sometimes to build these cool
vibe codes unless you have an engineer to get to the last 10%,” Schwab said.
As for how AI might shape the short-term rental and hospitality
industry, both executives felt disintermediation of the big players could be in
the cards.
“I genuinely think one of these big models is going to find
a new way of monetizing it and provide you with much more relevance. There will be
groups of consumers who will be so comfortable using AI that it will end up disintermediating some of the large players in
the marketplace,” Donoghue said.
Schwab had similar thoughts.
“I think there is someone out
there right now working on code that is going to dislodge the OTA traction. We’ve
all been looking at the OTAs for a long time and thinking how is anyone ever going
to compete with this. Somebody is thinking about search in a way no one is talking
about, and that completely changes the search game in a way we can’t imagine
until we see it.”