How Booking.com uses AI and data to create connected trips for its customers

Anyone who’s planned a trip knows that planning a successful holiday is a challenge. Even in an age of AI, holidaymakers can find it tough to find quick answers to important questions.

That’s where Booking.com’s plan for the connected trip aims to make a difference. The online travel specialist is using Snowflake’s data platform to consolidate information and create better customer experiences for holidaymakers and hotel partners. Huy Dao, Director of Data and Machine Learning Platform at Booking.com, explains what the connected trip means:

If your flight changes, or you need to upgrade your hotel, or you need to arrive a little bit late, we want all those things to be dealt with in one place. To deliver that connected experience, the data also needs to be joined up. So, a lot of the work we do is about making sure that the data is high quality and the models are set up in the right way, so that they can be connected.

Dao joined Booking in August 2023 and discovered the company was using Snowflake as a fully-managed, cloud-based data warehouse. While this platform was proving its worth, he knew the technology provider offered additional capabilities, particularly for exploiting data and creating AI-enabled services.

Today, Booking uses a range of Snowflake components, including Snowflake Intelligence and Cortex AI, as it explores pioneering developments across Machine Learning, gen AI, and agentic AI. Rather than just providing a consolidated and trusted source of information, Dao says Snowflake provides an enterprise-wide solution for intractable business challenges:

It’s not just the data warehousing component, the catalog, the natural language interface, or even the OLAP functionalities, but also potentially the OLTP – the real-time data access to applications. Snowflake is working to provide them all. And then the challenge for people like me is to keep up with what they have and pick and choose what to use and what not use in the moment.

Exploiting data

Booking began to work with Snowflake when it moved from its on-prem Hadoop system to a cloud-based managed service. The technology provided a range of benefits, including reduced maintenance requirements, highly governed data access, and elasticity in terms of computing. The Snowflake platform forms part of an integrated data stack with other technologies, including ThoughtSpot for analytics, Astronomer and Airflow for orchestration, Immuta for access control, Arize for Machine Learning observability, and AWS for cloud services.

Dao says the developer experience is significantly improved with Snowflake and the supporting technology ecosystem. Something that might have taken nine days to complete in the firm’s Hadoop base now takes only two or three days, including building the pipeline, making it operational, and creating a much better user experience than before.

A core component of the company’s strategy for connected trips is the Booking Data Exchange, an end-to-end data and machine-learning platform powered by Snowflake. The platform is used by 1,500 data practitioners, serving petabytes of data and hundreds of billions of predictions daily.

As Dao and his colleagues have begun to explore ML and AI, they have started to exploit some of Snowflake’s additional features, such as using Snowflake Intelligence, the tech company’s AI-powered conversational agent that enables users to query structured and unstructured data using natural language:

Snowflake Intelligence reduces the barrier to entry for internal users when they access data. The technology is not just about core tech analysts. We leverage Snowflake Intelligence as a solution to empower our internal users, even those with low data literacy, to get access to the data.

Booking also uses Cortex AI, which provides models and capabilities for staff to access data safely and securely for business projects. Across the organization, Dao estimates Snowflake has helped the company scale its data operations and eliminate workload delays, with connected trip transactions increasing by 20%.

Developing features

Booking is already using data and AI to develop the next generation of connected trips. One example is a partner-to-guest agent that facilitates communication between Booking’s guests and hotel partners. The agent suggests responses to each guest’s inquiry. Depending on the message, the agent uses an existing template or generates a tailored answer, helping partners reply faster and more accurately.

This tool is Booking’s first agentic development. Crucially, Dao says the Snowflake platform and its additional features are helping the company prepare for the next stage of customer service, where AI-enabled agents use data:

The semantic layer on top of the data we have ensures that, even when an AI agent can look at the data, it knows the actual meanings, metrics, and measures that we have. As Snowflake is the place where we choose to serve a lot of our data, it’s very strong in that area.

What’s clear is that the way Dao’s team uses Snowflake now differs from the past. Rather than focusing on connecting the platform to the rest of the data ecosystem, the team today concentrates on ensuring the business makes the most of the technology’s AI capabilities:

We need to make sure that, when we work with something new, we test it, report issues, and then ensure that it works well before we roll it out. So now, in terms of moving forward, instead of working on the basic, fundamental components, it's about making sure that we add more value to our users.