Rebuilding legacy technology to optimize for artificial intelligence (AI) is a massive undertaking. For travel giants, the journey is resulting in lessons in architecture, speed and corporate culture.
Executives from Hilton, Air Canada, Marriott and Evolve pulled back the curtain on what it takes to confront legacy tech.
Stakeholders shared insights in a string of discussions that surfaced several common themes at Skift’s Data + AI Summit in New York on Wednesday.
The value of the cloud
It is “quite a journey” transforming legacy, monolithic architecture, according to Michael Leidinger, SVP and CIO of Hilton, who discussed the brand’s multi-year efforts.
As Hilton sought to modernize its technology architecture, it implemented a microservices-based platform, built a centralized data lake and migrated compute and storage to the cloud.
The cloud journey has been transformational, according to Leidinger.
“The most important application we have is the central reservation system and distribution,” he said. “Prior to going live on [a] cloud environment, anytime we had increased capacity, it ran about a million dollars of capex … we peaked at processing maybe about 120 million transactions a day.”
Now, 90% of Hilton’s compute sits on the cloud. “All day long, our central reservation system, our websites, our application systems—they spin up and down—variable capacity—and what's great is there are days now we are serving over one and a half billion transactions.”
That’s a more than 10x increase, he said.
“Why that's so incredibly powerful [is] that just isn't shopping or booking or reservation modifications—that's also doing things like great recommendations, all of the revenue management system,” he said. “It is allowing us to again provide transactions and do things we couldn't do before. The cloud journey has been absolutely transformational for us, and it was worth every lesson we learned around it.”
Firas Al Osman, chief digital officer for Air Canada, said the company is shifting to become a “connected and intelligent” airline.
“We needed to make sure that our data and our systems [were] no longer fragmented,” Al Osman said, noting the airline needed modern systems.
“We are no longer on the mainframe as of last year, which is huge,” Al Osman said. “We're on our journey to move all of our applications off of our data center into the cloud.”
Soon, Air Canada will not just be cloud-first but cloud-only, Al Osman said.
Picking up the pace
At Air Canada, the mindset around speed has changed with AI, according to Al Osman.
Teams are currently building technology in the span of weeks instead of months and years.
Historically, Al Osman said a bunch of questions preceded the decision to move forward with building. Will the new technology pose a competitive advantage? Should it really be built? What’s the concern level in terms of maintenance cost?
That’s changed.
“Now I think the default answer is, if we can do it in a few weeks, then we're going to go ahead and do it, and we have some examples of that already in production today,” Al Osman said.
Arun Nagarajan, chief product and technology officer at Evolve, said the industry needs to embrace the discomfort that comes with learning and iterating the technology.
“AI is one of these things that kind of … breaks hierarchy a little bit,” Nagarajan said. “All of a sudden, your operations teammate is able to do data analysis faster than your finance team, so that is democratizing something—but it's also scary.”
The sector needs to confront whether that fear is acceptable and needs guardrails or whether some legacy thinking needs to be dispelled, Nagarajan said.
The importance of change management
Embracing a technology shift like AI culturally is just as important as implementing the technology.
Al Osman said one of the biggest realizations his team had was that if they didn’t have support for an AI transformation, they would not be able to make a meaningful impact.
But with AI, he feels like the need for change management has been “forgotten.”
Citing experience in business transformation throughout his career, Al Osman said change management has to be a priority. Companies have to influence hearts and minds to make an impact, he said.
“With AI, it's further exacerbated because there's always this issue with lack of trust, transparency and so on,” Al Osman said.
Another layer of change management is ensuring that people aren’t just using the technology for the sake of using it—particularly as AI costs rise.
Leidinger said Hilton is taking tokenomics into account as large language model companies are leaning into monetization. Planning for token consumption has to become a priority.
“We are keeping a very close eye on token consumption across Hilton and figuring out how we optimize,” Leidinger said.
The company is looking to make sure that people who are using the most complex AI models are using them for the right things. With that in mind, Leidinger flagged education as another important part of change management.
“As you deploy, this is part of the change management and education program you have to do throughout the entire organization,” he said.
Colin Coleman, SVP of enterprise data, analytics and AI for Marriott International, echoed the sentiment that the skilling of the workforce is paramount.
“The team implications are huge across the board,” Coleman said. “I would say that is bigger than the actual technical scholarship.”