What It Takes to Put AI Into Production in Travel

0:00Benjamin Cany: Hey everyone, Sania Lali here coming to you live from the SCIF Data and AI Summit in London. In this segment, I'm joined by Benjamin Cany, Vice President of Product and Business Strategy for the AI Delivery Hub at Amadeus. So we've spent the day in a room full of different travel companies at various stages of their AI journey. We've heard how AI is reshaping the travel stack, what investments are paying off, and what it takes to deploy the technology under a fast changing set of rules. So Benjamin and I get into why moving from an AI pilot into a live operational environment is such a big challenge.

0:33Benjamin Cany: And what changes as AI moves from recommending in action to actually taking one. Check it out. Hi, Benjamin. Great to have you with us today.

0:42Sania Lali: Thank you for having me.

0:43Benjamin Cany: So you spend the day in a room full of travel companies at various different stages in their AI journey. And you were on a session as well. What have you heard that tells you that the travel conversation in terms of AI is different now than it was a year ago?

1:00Sania Lali: What's for sure is that the level of excitement in the room is comparable to what we were getting the previous couple of years. But now let's say, while last year we were wondering, are the travelers going to adopt the new AI surfaces? The question has been answered. And yes, they are using it massively. most of the surveys are telling us that the vast majority of travelers are using the AI surfaces, the like of ChatGPT, Gemini, Cloud. New ones, by the way, because in the room we heard a couple times about instinct, which was not really top of mind even a few weeks ago. So adoption is here. Now the question will be how the travel actors are going to be present in these surfaces where the travelers are already.

1:47Sania Lali: In my role in the AI delivery hub at Amadeus, I'm connecting a lot with these large AI companies. And what I got like only a couple weeks ago, one of the leading AI assistant companies is that they are getting more than 330 million requests a week, which are with travel related content. So no more question on the adoption. Now it's time to execute and and bring actions into these surfaces.

2:17Benjamin Cany: So one of the themes that we've kind of heard about is, you know, this difference in demonstrating something in a pilot and kind of trusting it inside a live travel business. And your session was called What It Took To Go Live, which gets to the heart of that. So what has to happen before a promising AI pilot can be trusted in a live travel business?

2:38Sania Lali: So you're mentioning the panel. I was with a hotel operator and I think both of us agreed that the best way to start exploring agentic commerce use cases is really to start from the business problem we are trying to solve. It can be a small business problem we need to start tackling, but that's the best way. It doesn't start with the technology, it really starts with the business outcome we are looking for. So that's number one. And second, I think the consensus in the room and very core to what we do in Amadeus is how do we, right from the start, put in place governance, a legal framework that will enable these use cases to be not only piloted, but then deployed later on.

3:25Sania Lali: And that's really what we are doing in Amadeus with the AI Delivery Hub, building AI and agentic capability that can then be reused throughout the different divisions of Amadeus so our hospitality, airlines, travel sellers, so that we can feel confident that these capabilities are built with the right level of compliance and governance.

3:50Benjamin Cany: I wanted to pick on something that feels like an important kind of dividing line in the AI conversation today. A lot of the first wave of generative AI in travel was mainly around kind of answering simple questions or like pointing people towards certain information. But now we're talking about AI that can actually take action. So, you know, whether that's making a booking or like resolving an issue. So what changes when AI moves from that role of just making a recommendation to actually, you know, taking an action?

4:18Sania Lali: Yeah, that was also one of the theme of the day. We are in a complex industry with a lot of fragmented systems, sensitive data, complicated workflows, data that are also critical and changing very swiftly for the availability of a flight, the pricing of a hotel room. So all this needs to be taken into account when we move from this pilot to deployment more at scale, specifically as we are moving from search and getting closer to taking action. So the closer we go to booking payment servicing, the higher the bar is being set when it comes to the standards that we need to deliver. So that's what we are working on as well in the AI Delivery Hub to ensure that this is done the proper way.

5:09Benjamin Cany: Right. And trust is a word that came up a lot today at the conference. And one of your points was that AI will scale at the speed of trust, which is a really interesting way to put it. But I feel like trust is a fairly broad word, isn't it? if I'm a digital leader, what does a trustworthy AI system actually look like for me? Or what do I have to see or measure before I can fully trust an AI system in a critical workflow?

5:36Sania Lali: Yes. So when designing systems, we need to ensure that some of these governance legal aspects are right at the core. There are also ethical principles. We have six ethical principles in Amadeus that we are following when we develop any new solution. As well, and let me maybe give you a very recent example. Last week, we launched Max for Travelers. So Max is the portfolio in Amadeus for AI and Adjantic. And Max for Travelers is a solution for airlines that they can deploy to better serve their customers. The first use case that's being put in production is around voluntary changes. You and I, we are in London, and tomorrow I need to see a customer.

6:25Sania Lali: It was not planned, so I will need to change my flight bookings. I will reach out to Max for Travelers if the airline has decided to opt for this solution. Max for Travelers will, in the vast majority of the cases, solve this challenge for me and rebook my flight. But we know that our industry is complicated. There are some use cases that cannot be solved without a human being. And that's where we are defining very strict rules when it comes to the handoff from a machine to a real human being. So in this case, the agent of the airline will be able to take the conversation and complete the use case if it has not been performed and completed by Max for Travelers.

7:06Sania Lali: I think we need to keep in mind these ethical principles, legal, and as well some clear rules on how we are going to interact between AI agents and humans.

7:16Benjamin Cany: So one thing that I feel I can kind of get lost in the discussion around governance and architecture is really the business case. And you've argued that AI shouldn't be the whole story and it's really the outcome that matters. So when you look at these use cases that are present in the real world today, what separates the ones that are creating actual value from the ones that stay as just interesting experiments?

7:41Sania Lali: So we are seeing clear business outcomes. I mentioned Max for Travelers, but we have also a company we acquired in the first quarter of this year focused on corporate travelers. Skylink, they've already managed to reduce drastically the time spent to perform a corporate booking. We have Max for Travel Sellers, helping employees of travel selling companies using our solutions to reduce by three quarters almost all the manual tasks they have to deal with. And there is also another aspect I'd like to bring is how do we bring to the engineers, to the developers, the capabilities to test and learn new travel use cases.

8:24Sania Lali: And that's at the core of the partnership we've launched with Anthropic earlier this year. We are going to bring to the developer using Cloud Code the possibility to use some of the Amadeus APIs to test and learn and develop the new news cases.

8:42Benjamin Cany: That's really interesting. Just to wrap things up, I have one more question. There will be plenty of companies that will be leaving today with lots of ideas around how to pursue AI. What's one decision that you think a travel executive should be making differently over the next, let's say, six to 12 months if they want to move from the pilot phase and into production? And then on the other side of that, what's one thing that you think companies should resist jumping into too quickly just because the technology makes it possible? All right.

9:12Sania Lali: So first of all, I think we should all continue to join in this kind of event. because I think for me that's the best way to start tackling business problems. Here we have all the players of the ecosystem from travel providers, tech companies, travel sellers.

9:29Benjamin Cany: There's problems in the room.

9:30Sania Lali: Yeah, for sure, lots of problems but also opportunities during the breaks to figure out what is the next use case we want to tackle together. And that's the role we want to play at Amadeus is really how we can connect the different actors of this ecosystem. to start with, I think, a concrete use case, not so complicated, bringing a little squad together with team members from the different companies and be focused on how do we bring this use case to life in the different systems connected to real data. And when this innovation starts to connect with the real system and the data of our travel ecosystem, them, that's when magic will start to happen in the Argentic Commerce era that we are entering in right now.

10:17Benjamin Cany: Magic, that's a great place to live with Benjamin. Thank you so much for joining us. And if you want to learn more about Amadeus, please head to amadeus.com.ai. Thank you so much.

10:27Sania Lali: Thank you.