Christian Delbert - Fotolia
Agoda scales AI strategy, opens new APAC tech hub
The digital travel platform has set its sights on becoming an AI-powered travel companion as it changes how it builds software and moves its tech workforce into a new facility in Bangkok
Planning a holiday can be a logistical nightmare. Matching flight arrivals with hotel check-ins, figuring out transport to distant attractions and praying the itinerary holds together is enough to stress out even seasoned travellers.
Digital travel platform Agoda wants to fix this by transforming from a booking site into an artificial intelligence (AI)-powered travel companion that helps users manage travel logistics. To get there, the Singapore-headquartered company is rewriting how it builds its own technology.
Speaking on the company’s AI strategy during a media preview of its new technology headquarters in Bangkok, Agoda’s chief technology officer Idan Zalzberg dropped a bombshell about the reality of modern software development: traditional programming is effectively dead.
“We are essentially at 100% usage of AI-assisted coding,” Zalzberg said. “Nobody is coding anymore. Nobody is writing Python or Java – that practice is gone.”
Because AI coding assistants generate code so rapidly, Agoda soon found itself bottlenecked at the code review stage. In response, it built an AI code reviewer that now processes over 50% of all code changes with a 92% developer satisfaction rate.
Using an agentic platform, the company also injects Agoda-specific context, databases and deployment protocols into standard AI agents, so they operate like human engineers. But while Agoda is a strong proponent of open source, it prohibits the use of open source AI agents such as OpenClaw due to governance and security risks.
“Our security team has been hacking into bots, and I can tell you they’ve hacked into every chatbot out there,” Zalzberg told Computer Weekly. “I can’t really feel safe about what it’s going to do when I go to sleep and what it’s going to do in my name at night.”
Looking ahead, Agoda developers will focus on specification engineering that provides high-level instructions to AI, harness engineering to build validation environments and strategic architectural reviews, Zalzberg said.
Additionally, development environments will be moved off local laptops into Agoda’s private cloud, creating a “soft playground” where AI agents can operate safely. Zalzberg added that the infrastructure cost will be offset by no longer needing to purchase powerful laptops for engineers.
Agoda operates almost entirely on its own infrastructure, powering one million CPU cores that process 14 petabytes of data daily – including four trillion messages at a rate of up to 50 million messages per second. “Every second, 30 million prices are being checked worldwide, just to make sure every search comes up with the best offers for our customers,” Zalzberg said.
Managing token cost and proprietary models
Across its AI initiatives, Agoda uses half a million tokens per second on average, reaching as high as a million tokens per second across different AI technology suppliers, including the likes of OpenAI.
The company doesn’t manage token costs by handing developers a flat monthly token allowance, which Zalzberg quipped would result in engineers “going to the beach for a week” once they ran out of budget.
Instead, Agoda opts for fine-grained control. Developers might get a specific budget to use expensive, high-end models for critical projects but retain unlimited access to cheaper models for experimentation and pet projects.
Interestingly, despite the rising popularity of open source large language models (LLMs), Agoda prefers to use smaller proprietary models for specific tasks, which can be just as cost-effective as open source alternatives.
However, the company is not dogmatic in its preference and will use open source models where it makes sense. “We have one case where we found an open source model that we can run on our own GPUs [graphics processing units], and it will save 90% of the cost,” Zalzberg said.
The ‘inside-out’ approach
To manage AI disruption, the firm adopted an inside-out approach, focusing its initial AI deployments on empowering employees with AI capabilities rather than pushing untested tools to consumers.
“The lowest-hanging fruit is always in the people, and that’s the easiest way to see value immediately,” Zalzberg said, adding that it also allowed the company to test and break AI tools safely. “If I give my customer an AI tool that doesn't work, they don’t say, ‘This AI tool doesn’t work’. They’d say ‘Agoda doesn’t work’.”
Today, over 85% of Agoda’s workforce uses its internal AI platforms weekly. The company measures investment returns based on whether tasks are operational (efficiency gains) or generate new revenue. In marketing, for example, Agoda has been using AI to generate 500 million webpages localised for different markets and languages – a feat impossible with human writers.
On the consumer side, Agoda uses AI to enhance user experiences, including property page chatbots, real-time multilingual customer support translation and personalised review summaries. Agoda’s room recommendation engine even uses AI to explain to users why it recommended a specific room based on their browsing behaviour.
A physical space for the AI era
Zalzberg said the company plans to hire in the “low hundreds” for its technology teams this year, balancing growth against global macroeconomic conditions.
This week, Agoda announced its relocation to a new technology hub in the One Bangkok district in Thailand. Spanning more than 26,000m2 across seven floors, the built-to-suit office brings nearly 4,000 employees under one roof.
The campus-style hub is specifically engineered to support the operation of a global technology platform. It features a dedicated network operations centre, connected collaboration rooms for fast incident response and professional content production studios.
Agoda CEO Omri Morgenshtern said the new tech hub reflects the company’s investment in “creating an environment where people can do their best work and continue building world-class tech that delivers exceptional travel experiences”.
For Zalzberg, the intersection of AI tools and a collaborative environment will ideally spawn spontaneous conversations and innovation as Agoda looks to shape the future of travel tech. “When you walk around the office, the art on the walls and the different environments just get your mind going. That helps to bring more motivation and innovation in our work,” he said.
Read more about AI in APAC
- DayOne and Cortical Labs are bringing ‘wetware’ computing to Singapore, using living neurons grown from stem cells to support the demand for AI while addressing sustainability concerns.
- Singtel and Nvidia have teamed up on a multimillion-dollar facility to help organisations scale enterprise AI deployments, tackle extreme datacentre power densities, and prepare for the era of embodied AI.
- The Australian government has struck a major five-year volume sourcing agreement with Microsoft to speed up adoption of AI and cloud technologies across the public sector.
- Following the viral success of OpenClaw and product launches from Nvidia and Tencent, Alibaba has unveiled an agentic AI platform that integrates with DingTalk to orchestrate business workflows.
