So revolutioniert Künstliche Intelligenz die Reiseindustrie

0:00 [Music] Hello. Hi, James. Let's have a seat. Thank you. So, full house here. Welcome to Hamburg. Thank you very much. Good morning, everybody. So, you've been with Booking.com for 16 years. I have. Yes. And Booking.com is one of Europe's biggest companies. If it were a German company, it would be the second largest company. You guys do 1 billion room nights annually. It's a new KPI I learned. Um, you do 24 billion in revenue. At least that's what you did last year. And you have around 20,000

0:48 employees. So booking is huge. Originally Dutch, now under a US American flag, right? Correct. Yeah. And tell me how how did you get there? Uh totally by chance is actually the answer. Uh I was a student. It was my holidays. I had registered with an agency just to do a little bit of work here or there frankly just to pay my rent and they sent me to cover for a sick receptionist at a company called Booking.com. Um and I went for the day. The day became a week. The week became a month. Um I went back to university. I wasn't sure what I wanted to do. Uh, and I really like the the culture in the company. So, I basically picked the phone up and said, "Hey, you were growing. You were recruiting. Would you give me a job?" They said yes. Um, I

1:35 said to them, "Uh, fantastic. Just so you're clear, I don't think I'm going to come back for very long." And I've never left. Wow. So, from receptionist to CBO, that's quite a journey. I've been through a few things. I've been very lucky. I've I've been marketing, I've been sales, I've been customer service, I've been product, I've been tech. Um I've never been in finance and I've never been in legal but I think that's fair enough you know and throughout your whole journey like what techwise what's the biggest disruption you've seen or the biggest change in the past 16 years aside from AI we'll talk about AI in a minute I mean if you're aside from AI it's probably the as mobile appeared and

2:20 changed the way that consumers search consumers by consumers uh view content etc. That was in some ways I guess a little bit akin to how we sit with AI today. It it was crystal clear it was going to be huge. It was crystal clear it would change a lot of things. It was clear what some of those changes were and it was a big unknown for a while about exactly how that would work. But that certainly changed the way we thought about our business quite a lot. Probably also the future, right? I think uh mobile usage will go up. Desktops will eventually die. So um yeah, I truly believe that that that will be the that was the biggest disruption before AI. I desktops will be down, mobile will be up, and I imagine something else will be up. What the something else is, I'm not

3:05 as convinced. Yeah. But how do you So you guys are are disrupting your business with AI. You're implementing and and bringing AI into travel. What are you ultimately trying to achieve? What's your biggest vision for AI and travel? So our our core mission as a company is to make it easier for people to experience the world. Um and the first part of that mission is actually easy. Um human nature does this for us. Humans like to travel. They like to explore. They got like to go to new places, like to meet new people. Just look at how quickly travel came back after uh after co. There was no gradual buildup of confidence to travel. It was super simple. if soon as the regulations in a given country allowed people to travel, boom, they did. Um, so that's

3:52 why we talk about making it easier. We don't have to convince them to to want to. Um, travel's complex. It's full of the unknown. You're going places you don't know. There's an uncertainty there. And uh, there's also your travel is generally a kind of high financial investment, but it's a really high emotional investment as well. And so the way we like to think about everything is through the lens of what we call the core customer journeys. It's basically what the steps that someone needs to take to travel. I need to choose where to go. I need to see places I can stay. I need to pick from the options. I need to pay. I need to decide what to do. Duh. And for as long as I've been a booking, those questions haven't changed. And so when you say, what are we trying to achieve with AI? We're trying to make each of those steps of that journey smoother, easier, simpler,

4:38 more intuitive. take the work away. Um, like when I joined, booking, what we used to call breadth of selection. More hotels online was like a differentiator and an advantage. Now the problem's different. There's millions of places you can book online. How on earth do I take it from this big down to the a set that I can deal with? Those are the kind of problems we we look at. So, we're looking to build contextual experiences. Um, reason I say contextual, we all talk about personalization. You're a business traveler. You're you're a family traveler. Yes. You're a leisure traveler, of course. You're taking weekend trips. You're taking a two week summer vacation. Like, so personalization only gets you partway there because you've got different elements. So, personalization, contextualization, adapting the experiences to the needs. That's what we're we're looking to deliver. And what

5:23 we call we call it the connected trip. When we do that, right, across the different elements of of travel, the how you get there, where you stay, what you do. So what do you have a clear example what customers see really like uh if you go on booking.com like as a as a user as a customer what sort of AI application or service what do you get in touch with so we've had AI in every page of our website for many years obviously what's changed is the gen AI element so perhaps a really good practical example of what you see today is our website used to ask you give us a destination give us a Give us a number of people. You can now go and search with natural language. I want a standalone property with a

6:09 swimming pool, with a sea view within x hours of Amsterdam, whatever it is you want to do. And the website can understand that, translate that into the needs, and give you results that relate to your sort of more human search. That's very different than it used to be. um in a number of markets, English speaking at the moment, we've got a an assistant that's there for you as well that works through a chat interface that will help you discover where you want to go or help you manage your bookings, etc. So, those are two of the more visible examples that we get on the on the traveler side. Also, a lot on the partner side. Partners get lots of questions like suppliers, hotels get lots of questions. We've got AI there to suggest answers to those traveler inbound questions as well and help the the supply side manage their business. Okay. So are you sort of trying to replicate the travel experience we all

6:56 used to have when we had travel agents if you like. So basically you know the lady who knows your family very well who who knows you like to stay in little cutesy hotels in in in Cornwell. Um absolutely you want to recreate that sense of you know the travel agent who's known me for 20 years in the sense of how relevant it is for you but I also think you want it to be a little bit easier like arguably the technology can be simpler. Um, my background is anthropology. Uh, one of the drivers of human behavior is we don't like not being in control, by the way. So, if we can give you the right personalized content, the fact that you're making the decisions will actually make that a more beautiful experience than relaying a bunch of information to a travel agent to do it for you. So, yes, but I think there's even a build on top of that that we can deliver. Yeah, I can I I can envision a a very good experience

7:44 because mostly what frustrates people I think about travel is um the lack of transparency sort of you're trying to rebook a flight because your flight has been cancelled and you get so frustrated because you hang in in some call center somewhere far and you can't get you can't seem to get your message across sort of. But imagine a fleet of agents that are individualized and specialized on you as a person. They know your credit card info. They know where you fly from. They know how you want to fly, where you want to travel. And that whole like that whole service or AI will create, I think, hyperindividualized, hyperpersonalized experiences, right? I mean, absolutely. And you're you're sort of describing what we call disruptions to travel. And you picked a classic

8:30 example and a massive example. My flight got cancelled. Yeah. Imagine how many of those micro stresses you have when you travel. You go somewhere you don't know and the first train is canled. That's not as serious as a flight because there's going to be another one. Yeah. But hey, we immediately tell you without you having to go look, look, there's another one you can take in x period of time. The ticket you've bought will still be valid, whatever that might be. All of those micro stresses coming out allows people to enjoy the travel and not have to worry about the logistical challenges of it. And that that's, as you said, as you're describing it, that that's where our investments into AI go. It's not only reacting to but also preempting and predicting what it is that someone might be interested in, what someone might need. Yeah, that's interesting. So, you have a whole different set of systems. So, you have predictive AI, you have machine learning to understand where they're coming from,

9:16 you have generative AI in the chatbot section and voice. So, that's that's interesting that you imply all of these different technologies. But how how do you like do you build this in house or how do you decide do you do like buy and build or do you develop your own? Yeah, we've got a a pretty strong capability I think in this space compared to most companies. We don't have we're not going to go build our own LLM. So we partner with OpenAI, with Perplexity, with Google, with AWS, with Meta. Um, we have what we believe is some quite smart middleware that allows us to work with models from all of those companies for like they're not all equally good at all things. So, we can test them against each other and decide which model we want to use for what use case. We can

10:02 change that down the line if we see a a change in performance. We're not going to invest in the very base line of those LLMs, but we do have a lot of a IML practitioners. We do have a lot of technology we build inhouse. Our build versus buy thought is very similar to most people's. Is this something that's differentiating or not? If not, probably more inclined to buy. Um, and where we think we're super well positioned to sort of drive this transformation, if you like, is we've got an awful lot of bespoke understanding and data of the travel vertical. The big players we talk about like Open AI, Google, Plexity, I'd say they play generalist. Like they do lots of things and as they get smarter, they start to build slightly vertical specific things. But our expertise on how people travel, what travelers want,

10:48 etc. We believe is really really deep. So one of the big questions for us is can we leverage that? If we think we're in a position where that data is powerful, we're more inclined to go go build something. So we we invest heavily ourselves, but we also make sure that our partnerships are strong with those those key players. Yeah. I think you were one of the first travel companies really having a close partnership with OpenAI, right? And you what did you do? you fine-tuned the models with your own proprietary data or done a few things there. So, we're part of uh the uh part of the OpenAI sort of first launch with brands of buying experiences that are coming in there. So, that's very visible to you on the OpenAI surface. Um we also use a lot of OpenAI models behind the scenes on our side. Some of them

11:34 sometimes localized. So if you uh if you come in and you use our itinerary builder, our trip planning capability, that's got open AI behind it, but it's got booking data built on top of it as well. Um and there's also places in full transparency where we just use the the off-the-shelf models. Um particularly for sort of internal efficiency cases often that's a a more straightforward. So you've got things you can't see, things you can see on our platform, and things you can see on Open AI that all represent that that partnership. And what's how's the adoption? Do you see like a lot of usage or do you attract that? Absolutely we do. I mean at the moment it's the small minority of our travelers who go and use those products on our website. Okay. Um I think that's likely to continue for a period of time and you see a lot of curiosity and people playing and then transitioning

12:20 back into what I'd call a more traditional e-commerce experience. But just look at the growth of users on those platforms. Like it's only going to go up. People will get more comfortable. It is a minority, small minority today, but it's it's going to grow and it's going to expand and it already helps us learn a huge amount about what people want from those those systems. I know talking to you backstage, there's things you wrestle with as well. I mean, one of the things for me is at times we almost say that Gen AI equals AI equals a chat interface. But those equals don't apply. Like for example, we offer a chat interface for the planning phase. I'm not convinced that people will go through and book in what looks like a chat interface. They may use that to explore to a certain point and then pivot back into something that looks more like a a search result experience. Those things to me are very

13:07 much to be seen. Yeah. And but how so if you have like all these different models underlying, you know, or using different sort of technologies, how how do you choose or how do you select the providers? Like do you have a process or why would you work with open AI versus anthropic versus Google like I mean we don't think of it as a versus we work with all of them. We've got the capability to plug models in for all of them and generally we've got a I think fairly sophisticated testing structure on our side that if we've got a new use case we would actually test multiple models to see which one performs best. Okay. And we'd be able to track the outcome of that and we'd simply use the the strongest performing model. And that might, by the way, as we think about the Gen AI transition, that might include

13:53 testing a complex Gen AI model against a much simpler, much cheaper, much less power and compute intensity uh intense model as well, just to see what the difference in performance is because I think one of the traps we can fall into is everything becomes a Gen AI problem and in some cases it's fractionally or not better than a a simpler setup. Interesting. And do you ever run into problems? Did you have any like what are the biggest challenges s of you encounter with that? I think we probably run into the same challenges that that many people do. Um you've got a massive knowledge lift to generate across a huge number of people in your company because 24 months ago if we what was your AI knowledge a gen AI knowledge sitting in the room probably very very small. So

14:39 there's an education element. Um, we believe that it's very important to be thoughtful and upfront about the ethical approach, which is why we signed the the AI pact, which is basically the AI act enacted early by choice. Um, so we had to put in place the right AI principles, the right ethics, the right bias checks, etc. That takes time and especially if you're a big company, finding that balance of innovation and challenge and disrupting yourself versus the risks that you take and the responsibilities you hold is always challenging. um driving the culture. People are worried about Genai, right? There's a sense of it's going to take my job away. What am I going to do? I'm a believer that won't happen. I think uh people thought the industrial revolution would end employment. It didn't. People thought the internet would end employment. It

15:24 didn't. It pivoted to service. People will go and solve problems that the Genai is not good at. But it will replace some of the work we do today. So, there's a big transition moment there for all of us to sort of embrace the the reality and the learning we're going in. So, that's a big challenge. Um and then trying to work in the right way with the regulators like our belief is and thank you for the introduction and the the very generous introduction earlier. We would like to think of ourselves as one of the European tech scene success stories. Absolutely. Yeah. Reality is for most of the 16 years I've been at booking the leading tech space is in the US. Yeah. The fastest moving and evolving tech space is in Asia. True. Europe right now is at risk in my opinion that if we don't get that balance right between responsible regulation but driving innovation will

16:11 continue to be a a somewhat distant third third place. So the other part for us is how do we think about our role in that wider industry. How do we work with regulators because the core aims of the the regulations that are coming in make total sense. They're about transparency. They're about consumer trust. They're about privacy. They all make sense. How you implement those makes a massive difference to how fast we and other companies can move and we we see we've got a role there but it's a a very complex balance to play especially when we don't really know where all elements of that technology is going to go in the future. Absolutely. I think that's going to be the biggest challenge for the next decade. How to harmonize um innovation and growth with existing laws because it's AI is such a global phenomenon that

16:58 we are racing with China and the US and data protection is not you know the biggest concern of China and the US is slowly migrating into that direction as well as they're lacking training data. So the question as Europe like how do we not fall behind totally by um by really adhering to this very very strict laws and I find it interesting that you say that you adhered you you did an AI pact so you were even like preemptively you know like uh worried and concerned about that. Well, I think we've learned the hard way once or twice that if you don't think about complex things like that up front, it's even harder to think about them later. So, that's one part of it.

17:44 Um, the other aspect for us is think about, we've mentioned US, Asia, Europe. We're simplifying. Sometimes Europe thinks as Europe, sometimes it doesn't. Yeah. on our platform today. Right now, there will be someone who holds a French passport, who is currently in the United States, who's looking to book a hotel in Germany. Maybe they come to us from an advertising provider who's based in Japan, and the hotel that they're looking to book is operated by a company that's based in Spain and owned by someone who live doesn't matter. You get my point, right? Which of those constructs of law is going to is going to prevail? So, one aspect of it for for us and for me is if we did think more as Europe and we did try to harmonize a

18:30 little bit more, we'd probably make that life a little bit easier for ourselves relative to a US and a a Japanese base. I think so. Chinese base. Yeah, I I agree. I I think we need to we need to find a better balance like in everything in life, you need you need to find a better balance and the pendulum sort needs to move a little further. We always say, right, that um you know, solving a problem begins with the acceptance that you've got one. Yes. I think we've got to first accept that the data and the facts and the figures of the last decade suggest that we're moving much slower than others and we have to decide if we wish to change that and become competitive again. Yeah, absolutely. And when we speak of competitiveness, do you see like the travel so in the travel industry as a whole, do you see that AI is giving

19:15 smaller players a bigger chance to participate or give gives them a boost? There's probably two ways I would think about that. In the digital space, the more the cheaper that access to this technology comes and the smarter that it comes, the quicker I think you can get a smaller business up off the ground. Your entrepreneurial background, you'll have an opinion on that. But I probably think that that helps from a travel provider point of view. Um, one of the the key segment that we've always added lots and lots of value to as as Booking.com is actually those independent hotels. um they as opposed to large multinational chains because they just don't have the scale to invest in all of the languages, all of the content, all of the genai. So in that in that second example, I think

20:02 genai will help us to therefore help them. I mean one of the things you see today, the search I described to you earlier where you use free natural language that type of search tends to turn up more small unique independent properties than a traditional search funnel does. Why is that? Do you know? because Genai is very good at picking out the the unique piece of something. So, lots of small hotels or independent homes have got some unique feature. They've got a medieval fireplace, they've got a original um portico, they've got whatever it might be. We don't go on a website when we filter for thousands of properties and think about that one thing. We think about have they got Wi-Fi, have they got parking, like quite mundane things. Once you start to search what you really want, you say what really is interesting and then suddenly these little nuggets

20:48 get pulled out in a way that just structured data historically couldn't do. So it's been a a very powerful beneficial thing for that more independent element and I think it brings a new level of sort of human connection and experience to the travel as well because in the end of the day our mission was to help people to experience the world. Um that's what you want to do. Technology is a tool to help you do that and we believe we use that in powerful ways but we want to free people to enjoy themselves. Thank you very much. Yeah, I think I think AI will help help you demac democratize travel really at the end. Cool. So, thank you guys very much. Thank you for being here. Thank you for listening. Thank you for your time. It was a pleasure. Learned a lot. We'll definitely follow Booking.com's journey on the aspect and up to number to the next guest. Thank you.