Agoda has rolled out a multi-product booking engine that lets travelers bundle hotels, flights, and activities into a single transaction — no separate checkouts, no juggling multiple confirmation emails. The feature, called One Checkout, went live on May 13 and is accessible via Agoda.com and its mobile app. The checkout combines payment processing, fraud verification, and supplier confirmation into a single flow.
According to Agoda’s CTO Idan Zalzberg, what sounds simple on the front end required the company to rebuild how its backend systems talk to each other.
“To date, on our platform a traveller could find a great flight, a great hotel, an activity they wanted, but it still meant they’d have three separate transactions, three separate confirmations, three separate experiences,” Zalzberg told WiT. “We know that’s not how anyone thinks about a trip. One Checkout closes that gap. You put together the journey you actually want and confirm it once. Sounds simple, but when you’re running separate booking workflows per product, things like charging the customer once require a completely different architecture.”
Rethinking the data model
The architectural challenge Zalzberg is referring to wasn’t just a matter of stitching workflows together. Agoda’s systems were originally built around individual products, each with its own booking flow, payment step, and fraud check. Making those work as one meant pulling apart and rebuilding a lot of what was already in place.
“Yes, absolutely,” Zalzberg said when asked whether One Checkout forced a rethink of Agoda’s data model. “We started with separate booking workflows per product, which worked fine in isolation. But a unified checkout means you can’t run independent workflows and charge the customer twice or ask them to complete payment verification twice. We had to introduce a way to dynamically combine product-level workflows into a single integrated one, where shared operations like payment happen once. That required phasing out a lot of the structures baked into the original single-product model.”
Idan Zalzberg at the most recent WiT Singapore event, talking about the efficient adoption of AI and new tech
The sequencing problem
Beyond the data model, Agoda had to solve a harder, less visible problem; in what order do you confirm bookings across suppliers with different rules, cancellation terms, and hold windows?
“The sequencing,” Zalzberg said. “Every product has different cancellation terms, different hold windows, and different risk profiles. If you fire all confirmations in parallel and one fails, you might be stuck with a confirmed booking that you can’t roll back without a penalty. We had to build a product with logic that looks at the combination of products in the cart and recommends the safest confirmation order based on each product’s terms. Getting that right across combinations took real work.”
In other words, the system doesn’t just confirm all items at once, it evaluates each cart’s specific mix of products and figures out the lowest-risk sequence to confirm them in. And when something does go wrong mid-transaction, the goal is to contain the fallout.
“We’ve deliberately ordered things so that we’re in a position to cleanly handle and resolve any potential issue,” Zalzberg said. “This is a complex process and the ultimate goal is that the traveller never ends up with a booking they didn’t choose or a penalty they didn’t expect.”
Running it all across four data centers
Agoda operates its own on-premise infrastructure across four data centers rather than relying on public cloud providers, which raises the question, does that make building something like One Checkout harder?
“Having a multi-zone architecture adds complexity regardless of whether you do it yourself or use a cloud multi-zone setup,” said Zalzberg. “That’s just the nature of it. But we believe it’s worth it because it means we can keep customer data safe and keep the service running continuously even when things go wrong. Over the years, we’ve built the tools and practices to handle multi-zone systems to ease the operational path to building One Checkout.”
What about AI agents?
The obvious next question is whether One Checkout sets up a pathway for AI agents to handle trip planning autonomously on a user’s behalf.
“An AI agent can do this, but One Checkout was already designed to be a single click,” said Zalzberg. “Where we think AI agents can significantly support our customers is around planning and answering questions; our AMA Bot and Booking Form Bot are good examples of that. Those are areas we’re actively exploring.”
