Agoda’s inside-out approach to AI involves testing, refining, and validating new features internally before they are released to customers. According to CTO Idan Zalzberg, even a minor glitch in the company’s app puts its reputation at risk.
“If I give an employee an AI tool and it sucks, they will say ‘Hey, this tool doesn’t work.’ But if I give customers an AI tool that doesn’t work, they will say Agoda doesn’t work for them,” he explained.
Internally, this has allowed Agoda to experiment further and take more risks. The company’s recent move to its new APAC Tech Hub at One Bangkok supports this strategy, with around 1,500 of its more than 2,000 technology employees based there.
Building the future
Agoda’s strategy of building technology in-house rather than buying platforms is supported by its own data centres. The company runs about 1 million CPU cores across its data centres to power the Agoda platform and maintain what it says is 99.95% uptime.
“This allowed us great flexibility and cost control. We don’t buy enterprise platforms, because when they change their pricing, then you’re stuck with them, or if somebody has a great feature, then you cannot have it because you’re not on that platform,” Zalzberg said.
For Zalzberg, the lowest-hanging fruit for generative AI was Agoda’s employees. By automating repetitive and routine tasks, employees could focus on work with greater impact.
“Before Slack came up with AI in their platform, we built our ChatGPT-in-Slack tool almost two weeks after GPT came out, giving teams the ability to converse with AI in Slack to summarise a thread or a video, for example,” Zalzberg recalled.
Even before model context protocol existed, Agoda had built internal tools that allowed employees to access company-specific information. Query Assist, for example, enables employees to ask analytical questions about Agoda’s operations. The tool understands the company’s data structures and semantics, and can provide answers comparable to those of an internal analyst.
Zalzberg observed that software development has changed fundamentally: “Nobody’s writing down words of Python or Java and kind of clicks away. That practice is gone.”
He added that tools such as Cursor and Claude are a good starting point, but they don’t address the entire software development process.
“We noticed that because AI generated more code faster, we got stuck on the code review part, which is very important, because you want to make sure that the code is correct and matches your style,” he said.
To address this, Agoda built an AI code reviewer called Code Buddy. Zalzberg said the tool was once one of the most hated among Agoda employees, but has since achieved a 92% satisfaction rate.
“It goes to show that building correct AI solutions takes time, and takes a lot of feedback and processing to make sure that you do it the right way,” he said.
AI vs fraud
Agoda uses generative AI to identify fraudulent listings and suspicious activity. Zalzberg described the effort as a constant arms race between the company and fraudsters.
“The fraudsters are getting very sophisticated. They use AI to create fake properties. They can create fake messages so it allows them to scale something that they couldn’t scale before, so we have to do the same on our side,” he admitted.
Agoda’s fraud detection teams also analyse unstructured data, including how hotels and customers communicate within the app.
“We are looking at things like, ‘Why would the hotel ask this from the customer, like make them go to some YouTube link and watch a video?’ That’s weird, and this kind of stuff allows us to actually fight fire with fire, in that sense,” Zalzberg said.
Zalzberg argued that the challenge is not to stop fraudsters from being fraudsters, but to respond quickly as their tactics evolve. “You can’t make the fraudsters not be fraudsters, so we have to make sure that every time they change their practices, we will not only react, but react fast.”
Enhanced customer support
Customer support remains a big part of Agoda’s business, and the company uses generative AI to reduce administrative work for support agents. A case summariser automatically summarises activity on a customer case, eliminating the need for agents to manually document everything they’ve done.
Agoda has also introduced an auto complete and drafting feature to help agents respond more efficiently to customers.
Customers can also be overwhelmed by the volume of reviews on the platform. To address this, Agoda developed review summaries and personalised reviews designed to surface information that may be more relevant to individual travellers.
To complement this, Agoda’s Property Ask Me Anything chatbot allows customers to ask questions about specific property features and other concerns.
“We have so much data, but we don’t know how to give it to you. We don’t know exactly what you want to know. We don’t know what you’re missing. With the limited amount of space on the page, you can’t just read everything, so having this ability to ask us whatever you want, we will help you with all the data that we have, with the reviews, with everything we can do to make sure that you get the best, most trustworthy response,” Zalzberg said.
Approach to innovation
For Agoda, having no shortage of opportunities is both an advantage and a challenge. The company has more than 7,000 employees from over 100 nationalities across 50 offices worldwide.
“I always say our biggest challenge is that we have too many opportunities. Just everywhere we look there’s, ‘Ah, we can do this. We can do that.’ The only way we can achieve and get all this value is by having innovation coming from everyone. It can’t be just a top-down approach. We need to get everyone to be a part of this culture of innovation,” Zalzberg said.
Agoda now offers flights and travel activities in addition to hotel bookings. According to Zalzberg, the broader question is what role the company wants to play in travellers’ journeys.
“Are we still like an OTA that just allows you to select products, or are we becoming something more like a travel companion that helps you to plan and build everything related to your travel plans? That’s some of the thoughts we have today, and we are very excited about that future for us.”
