Sylvain Roy has spent most of his career in technology roles at Amadeus. Joining as a contractor in 2003, he rose up through the ranks to become senior vice president of technology for platforms engineering in 2019 before becoming chief technology officer in 2023.
Roy spoke to PhocusWire about the company's technology priorities, its move to the cloud and how it's employing generative artificial intelligence internally and externally.
The interview has been edited for brevity and clarity.
As you look at your role within Amadeus, what are your priorities for next six months?
The move to the cloud, no question. It’s a major program, it's transformational. People think very different things about the move to the cloud. There’s a quick lift and shift of everything to the cloud where some companies move their applications and don't do any redesign. When you do that you don't leverage the cloud so much, and it’s not what we’re doing. It's a transformation for us in terms of the platform and in terms of the way of working. It’s top of my list in the next six months although we're not going to be done in the next six months. It’s moving well.
Generative artificial intelligence is number two because we're doing a lot. I truly believe it will have an impact on the world, on the market and on Amadeus in two ways: internally and to ensure we get our product to the next level with this technology.
[Priority] number three is always there in my job and is all about the stability of the system and ensuring that we provide a system to our customers at the level they expect. This never stops. You can ask the question five years from now and it will still be [a priority].
How do you go about setting priorities?
We don't do technology for the sake of technology, so the question is what's the business need? What do we need to do to better serve the traveler and better serve our customer?
Subscribe to our newsletter below
Another aspect is about technology. Technology is moving faster and faster, we see more and more things appearing. We have less time to actually select the right technology. Our technology strategy is based on those two inputs - internally what our business and customers need and externally what's happening in terms of technology. What can we leverage and which technologies don't matter so much? We have to find our way between the two.
The move to the cloud is a big part of that. You see all the new capabilities we give to our customers, for example, the ability to deploy our software anywhere in the world. Optimizing our costs would be a business need. You will see another need, which is the stability of the systems once we go to the next level. So, I truly believe that our technology strategies serve our business.
So when you look at those priorities, what do you think your biggest challenges will be?
The move to the cloud is the biggest program, the biggest IT program in the organization as we speak. Every single application in the system today is either in the cloud or engaged in the migration process. That means a number of developers all over the world are working on it, so just due to its size, it's the biggest challenge. It's a transformation and to succeed in a transformation is a huge challenge. There's the technological aspect of it, the platform we are building with Microsoft but the other part is the operating model - there is this change of responsibility where developers are more empowered.
Let's talk about AI use cases internally and externally that you think are interesting and where do you see developments going?
Internally, we have already deployed GitHub copilot for our developers and a [Microsoft] Office copilot for the entire organization, which means that we are already leveraging AI to generate code and generate better code. That works quite well. We're also putting in place our own large language model to get better access to our internal knowledge.
The other thing we are doing internally is to use generative AI for testing, to generate tests for us and for our operations, to ensure that we can apply the system better and when something happens, we want to sort out what the problem is. Generative AI is something that can help.
Where do I see things going? You might have learned about agent design where the idea is you use a generic model and you specialize it. So instead of just interacting with one generic model that will do it all, you have agents that are generic models specialized on a specific topic and they interact all together. So, the future is probably with this model that can do more and more, not because one will get smarter, but because if I want to deploy a piece of software, one agent is on software, one is on testing, one on deployment and they work altogether. I think the industry at large should be moving in that direction.
If I move to external world, there are several dimensions, the first dimension is generative AI generates content. The second dimension is that it’s a conversational interface, almost a new channel. The third dimension is all the problems it can fix, the solutions it can find to those boring tasks.
So if we think about that first dimension, think about shopping. We provide websites for airlines. We have to provide landing pages and those pages are generated for you when you're looking for inspiration. Airlines need content for that, it's difficult to acquire that content, it’s difficult to build those pages with the content - we can do it in a fully iterative manner for them.
You might have learned about agent design where you use a generic model and you specialize it. So instead of just interacting with one generic model that will do it all, you have agents that are generic models specialized on a specific topic and they interact all together.
Sylvain Roy - Amadeus
On the conversational interface, we have this generative AI model on Cytric Easy, which gives the corporate traveler a very natural interface in Microsoft Teams. He’s basically talking to a virtual travel agent who can answer every single question he asks and do the booking for him.
The third one is as a digital assistant. We have a complex product. There are things for our users that are complex to use, because the problem behind is complex to solve. How about you have a digital assistant which gives you insights about what you are doing, gives you insight about the data and the outcome of what you've done in the previous years and helps you to configure the two for current year. This is something which I really believe can make a number of our solutions more accessible.
The industry keeps coming out with these slogans - we want to be travel retailers, to create the connected trip, the perfect trip. We talk about customer-centricity, but we don't really do it. How far are we from any kind of better trip?
There's obviously a lot to do and a big part of the complexity comes from the number of actors. There’s no actor alone that can fix it.
If you look at your entire travel, you're going to be in touch with a travel agent, an airline, an airport, with a hotel. Then, you've got all the other players like the vendors plus you need to add on top of that local government and regulations. And, that makes things crazy complicated, because not one actor alone can make it better. I truly believe that Amadeus is one of those companies that has a role to play because we're present in so many touch points. When you do your [travel] shopping on a website it's likely to be Amadeus. Between airline and travel agency website, it might be an Amadeus touch point or it will be just behind, and when booking it will probably come to Amadeus. Then check-in at the airport, you’re boarding and all of that is in a way driven by Amadeus so we have a role to play across the entire journey.
How far are we from really making that better? I think we have more and more capabilities to actually align the sectors. From a technology perspective, what we're doing is introducing an event-driven architecture. Today we have a service-oriented architecture which is huge, we have thousands and thousands of micro services. We’re complementing that with an event-driven architecture.
Why? Because the thing with event is you can monitor the overall travel journey and react in real time. It’s not like services where it's two applications calling each other, it's events flowing into the system and allowing applications to realize what is happening. Then we can have an application that is able to follow the traveler, be more reactive in real time and more supportive.