How Expedia’s CTO is using AI to transform work for 17,000 employees—and travel for millions

Extractive Summary of the Article | Read | Listen

Ramana Thumu, the chief technology officer at Expedia Group, says there’s no “AI Center of Excellence” at the online travel agency that sits in an ivory tower and mandates how everyone should be using artificial intelligence.  “We are democratizing AI across the entire company,” says Thumu. “Every employee, every team, and every workflow.”

Some examples of how this plays out include the creation of Expedia’s “AI playground,” which gives employees access to more than 60 different large language models—including from OpenAI, Google’s Gemini, Meta’s Llama, and Anthropic’s Claude—to build their own AI agents. Since January 2025, employees have built more than 1,500 different AI agents and around 6,000 monthly sessions occur within the secure AI agent builder environment on a monthly basis. 

Around two-thirds of Thumu’s software developer workforce have embraced AI coding assistant tools including Copilot, Claude Code, and Cursor, often resulting in an estimated 20% productivity lift. Thumu says he wants to bring more “joy” to coding by giving them a broad set of AI tools to infuse in their workflows. He vows that more efficiency won’t necessarily mean fewer jobs. “That’s not how I see it,” Thumu adds. “It’s an improvement which means you can get more work done, much faster, and higher quality work.” 

And last year, Expedia began to embed AI “squads,” teams of around four to six AI engineers that work with various business divisions ranging from legal, procurement, human resources, and marketing, collaboratively working alongside AI “champions” for each of those segments to figure out where AI can possibly be implemented to handle some manual tasks. 

Every one of these AI investments are being measured, ranging from velocity and cycle time for the AI coding assistant tools, and within customer service, tracking how AI is speeding up the time it takes to resolve a customer query. For the broader employee population, Expedia is measuring usage and the impact on workflows. If the company doesn’t see the right adoption or outcomes for these tools, Thumu says Expedia is quick to reassess and adjust its approach.

“We are absolutely testing a lot of experiences and placing a lot of bets to make sure which ones stick and where we scale, and when the test is not wildly successful, fail fast and take those learnings to do something different,” says Thumu.

External applications of AI include Expedia Trip Matching, which debuted in June and allows travelers to share any publicly available travel reel that caught their attention on Instagram and then share that reel directly with Expedia. In October, Expedia also served as a pilot partner—along with Booking.com, Zillow, and Spotify—for what’s known as “Apps in ChatGPT,” a feature that aims to further integrate the chatbot with external brands.  The company also developed an AI customer service agent that’s built on Amazon Web Services, which can help travelers with simple tasks including cancelling a hotel booking, getting a refund status, or changing their books so they don’t need to call a human agent. 

As Expedia moves forward with embracing more agentic AI capabilities, Thumu stresses that the company will prioritize stitching together the technology in a manner so one AI companion can help customers effortlessly through all touchpoints. He doesn’t want to build one AI companion that helps book hotels, another to handle discovery for excursions, and yet another to address customer service questions. 

“We are trying to bring those together,” says Thumu. “To make sure that from a customer standpoint, it’s one agentic experience. There is a lot of work to be done, but we are in the initial stages.”

show less