Follow ZDNET: Add us as a favorite source On Google.
ZDNET Highlights
- Agent AI is often more about conversations than producing services.
- Smart professionals focus on use cases and assistive technology.
- They test processes, refine approaches, and look for new opportunities.
Conversations with digital and business leaders about agentic AI often revolve around a similar sentiment: We’ve discovered agents, but there’s nothing in production yet.
But while everyone talks about AI experimentation, no business can afford to run endless pilots without creating business value. And experts suggest that professionals who fail to take advantage of AI risk being left behind, making it imperative to deploy successful agents as quickly as possible.
Also: How to build better AI agents for your business – without creating trust issues
At online travel specialist Booking.com, Hugh Dao, director of data and machine learning platforms, has been charged with delivering value from AI, including agentic services. They have generated results by taking a structured approach to service rollout, creating targeted solutions to the challenges facing customers today and tomorrow.
Dao referred to this approach in a conversation with ZDNET as the “connected trip”, in which Booking.com strives to ensure all elements of a customer’s journey, whether flights, hotels or attractions, are treated as an integrated experience.
Also: Worried AI agents will replace you? 5 ways you can turn worry into action at work
Creating connected trips means working with disparate information. The data stack that Dao’s team has built has allowed Booking.com to develop new AI-enabled services, including the firm’s first Agentic application, a partner-to-guest system that facilitates communication between customers and hotel partners.
Here’s what they’ve learned so far, along with five key lessons for other professionals who want to turn agentic AI pilots into great production services.
1. Identify a business challenge
Dao said the key to harnessing emerging technology is to use it properly. While some professionals remain uncertain about AI’s potential, he said companies can use agentic technologies to overcome difficult challenges.
He said, “In my opinion, AI is not some day or even flavor of the year – it is the real thing.” “I see every day at work how AI can impact the way we work.”
Also: 5 Ways to Use AI When Your Budget is Limited
At Booking.com, Dao and his team recognized that responding to customer inquiries in a timely manner was a major challenge for hotel partners. He agreed that agentic technology could help hotels respond to queries faster and more accurately.
“Before we launched the agentic solution, whenever a customer wanted to connect with a hotel partner – for example, if you wanted to check if the hotel had a pool, or if you wanted to check in an hour or two later – you’d contact the partner and say, ‘Hey, can I get this information?'” he said.
“However, when hotel staff did respond, they often required more work to get the correct response. Additionally, they were sometimes unavailable when a customer asked a question. Therefore, it could take a few hours or more for a customer to receive an answer.”
2. Build a data platform
Dao said the data stack his team has built allows Booking.com to accelerate the adoption of AI and machine-learning technologies for use cases like the one outlined above.
Dao: “AI isn’t like some flavor of the day or even the year – it’s the real thing.”
booking.com
The Snowflake Data Platform forms part of an integrated stack that includes ThoughtSpot for analytics, Astronomer and Airflow for orchestration, Immuta for access control, Aries for machine-learning observability, and AWS for cloud computing. The data team also tests and uses AI models from leading providers like OpenAI, Amazon Bedrock, and Google Gemini.
Also: Why enterprise AI agents could become the ultimate insider threat
Booking.com’s exclusive partner-to-guest communication system was developed internally in Python, and the data team used Langgraph, an open-source agentic framework, to help agents understand guest inquiries.
Dao said effective agentic systems are not just about backend systems. His team also thought carefully about the user interface.
“We want to integrate technologies or AI capabilities where it makes sense for our users,” he said.
“And in this use case, our partners already had a web-based portal to view their messages, so it was clear that we should integrate the agent right there to help them.”
3. Test the use case carefully
With a business challenge identified and the technology platform perfected, Dao and his team focused on implementation, which took place in two phases.
In the first phase, they developed a trusted assistant to help hotel partners deal with customer queries.
The result was an agentic technology known as Smart Messenger, which collects partner, property and reservation information to support hotel staff in communicating with guests.
Also: 90% of AI projects fail – Here are 3 ways to make sure your AI projects don’t fail
At this early stage of agentic service, Dao said humans are still in the loop.
“We want to make sure that partners have the final say on how they want to respond to customers,” he said.
“But we give them an assistant, so instead of it taking five minutes to respond, it can be just a one-second click if they’re happy with what the agent provides as a response.”
4. Appoint representatives as confidence increases
Over time, Dao said confident hotel partners can begin delegating more work to agents — and this phase represents the second phase of agentic implementation.
Here, Booking.com’s auto-reply tool allows hotel partners to define custom replies and quickly answer guest questions, such as whether the hotel has on-site parking.
“This stage is where the agent says, ‘Okay, if you trust me enough, I can work for you,'” Dao said.
“In this use case, the partner may be sleeping when the customer asks a question, as it is late at night. However, the agent can respond on the partner’s behalf – and this approach helps in a few ways.”
Also: 5 ways you can stop testing AI and start expanding it responsibly
Booking.com reports initial experiments Partner satisfaction increased by 73% Compared to previous messaging tools. Dao said the agent continuously learns from past interactions and user feedback, adapting its responses for accuracy and relevance.
“Now, with the agent, we measure the answer based on what it does; we experiment with it, and then we compare the improvement in satisfaction,” he said.
“Because the customer gets the answers they need, they don’t need to contact customer support, and it also reduces breakthrough support costs.”
5. Look for more opportunities
Dao said that agentic exploitation should be linked to personal use cases. As their team refines the customer experience, they continue to improve the platform, creating a foundation to support other agentic explorations.
“We didn’t want to build a stage for the stage’s sake,” he said. “When we built the platform, we had the user in mind. We made sure we chose the right agentic technology.”
Too: Is Google’s new $8 AI Plus plan worth it? How does it compare to the $20 Pro subscription
Dao said his team learned a lot from the agent development process. He advised other professionals to heed these lessons.
“When you do your testing, you may think the agentic system is good,” he said. “But when you go into production, things like latency can become a problem you have to deal with. Then, you have to simplify your architecture and platform.”
Over the next 24 months, Dao expects to drive even further growth at Booking.com. “You should expect that, as a company, we will invest heavily in generative and agentic AI not for entertainment, but to enhance the user experience,” he said.
“People are now looking for a ChatGPT-like experience, and we want a similar experience, or even better, when it comes to the travel experience on our sites.”
