The Agentic Travel Agent with Glenn Fogel, CEO of Booking Holdings

Episode Transcript

Glenn Fogel: Travel is actually like a game of dominoes, one thing falls over and all of a sudden a whole bunch are going over. And you think, "Wouldn't it be easier if everything was connected?”

Sarah Elk: Today we're excited to welcome Glenn Fogel, the CEO of Booking Holdings, whose brands include Booking, OpenTable, Kayak, Priceline, among others.

Andrew Ng: Travel's been widely discussed as a sector facing disintermediation risk from the launch-owned company. So look forward to hearing from Glenn about how he's navigating this as CEO.

Sarah Elk: Thank you so much for joining us, Glenn. We're so excited to have you.

Glenn Fogel: Oh, thank you so much for having me.

Sarah Elk: A lot of people are talking about agentic travel, and certainly there are examples of some people who have tried to one-shot their travel online. How do you think about that?

Glenn Fogel: Well, it's an interesting issue that very few people in the world have ever one-shot their travel. Travel is a complicated thing, and most people are going to spend an awful lot of time starting to explore, starting to look at different ways to discover. Then they try and narrow down [to] “What's it going to cost?” And one of the very interesting things is it's not linear at all. So your one shot that you just mentioned, that sounds like a linear process.

And this is really important actually for making sure you have systems that have memory because you go down part of it, you have to remember what you already thought maybe I want to do and then go back up and look at some other options because of something new that you just discovered. It's almost like a Chutes and Ladders game. Very complex. And I'll tell you, that's part of the problem. Why nobody has yet developed some sort of perfect technology to create what I am talking about as the connected trip that we're working on. We're making progress, but it's not perfect yet.

Andrew Ng: No, I think that's great because an AI travel agent’s job is not to just do what I tell it to. It's to share context and information to shape what the final trip turns out to be.

Glenn Fogel: And it's interactive and that's the great thing, because most people when you put together your travel, you're using different vendors, different times, and you're dealing with a lot of different people. And that's very frustrating. In fact, all of us have traveled and all of us have experienced terrible frustration, whether in the beginning trying to put it together or in process when something is going wrong. And the way I really look at it is I look how travel was done before there was digital commerce. When you go back when I was younger certainly and my parents would set up travel with a human being, that human being was called a travel agent.

And that travel agent knew some of you. They knew a little bit about your background. They knew what you could afford and you probably had put your travel together with that person in the past. So they were able to help you figure out what you wanted to do. They would show you a couple of alternatives. They'd put it all together, maybe give you an option here or there. You gave it one check and then it was all done. And if anything went wrong, you called that person and that person fixed it. That was great. That's what we really want to do is we want to create that experience only so much better using technology, using the capability that technology can give us so much better than any human could have done.

A human being wouldn't have access to all the data that we now have. Certainly won't do all the processing of all the different permutations, certainly wouldn't have the ability to make changes on the fly dynamically based on sensory information that comes in from throughout the world. Those are the things that we are creating using technology to make it so much easier. Our mission is to make it easier for everybody to experience the world. And using technology, we'll be able to do that. Now it's still very early. We are seeing a lot of progress, which is good, but I know that we are nowhere near where we need to be such that down the road people will no longer be talking about how terrible the frustrations they've had in their trips. They'll just be talking about how wonderful the trip was.

Sarah Elk: I look forward to agents figuring out how I optimize travel with five kids in a 14-year age span, keeping the teenagers happy at the same time as the six-year-old. That is definitely something I want.

Glenn Fogel: But that's exactly it. Look at the number of permutations in that and then come into if you put into a multi-city thing, it's going to be incredible. I mean, probably not as the number of molecules or atoms in the universe, but I will say it's fairly complex.

Sarah Elk: No, absolutely. And then your flight to Paris gets canceled, so you go a day later, then your train has to be rebooked, your hotel has to be rebooked. It's all the things you're describing.

Glenn Fogel: It's horrible. It's a horrible thing. And that's why you almost never get somebody putting together the verticals. We call them verticals, like an airline, and then a hotel, and the transportation. You don't do it at once because when you finally have gotten one part finally done and booked, the last thing you want to do is start exploring the other things. What you want to do is just go to the bar and get a drink and relax.

Sarah Elk: Jack Daniels or Pinot Noir in your bar, Andrew?

Andrew Ng: Sparkling water.

Glenn Fogel: That's not going to help.

Sarah Elk: I'll take the Pinot Noir.

Andrew Ng: In terms of how the market evolves, so we've been seeing in e-commerce that different retailers have different views about making their websites and their catalogs easy for agents to navigate and even transact automatically. Where do you see this going for travel? Do incumbents want to own the discovery experience or start to publish APIs so that the agent book hotels, restaurants, travel for consumers?

Glenn Fogel: So the key thing is you want to be wherever the traveler customer is, so to speak. And certainly in the past, and way back when SEO was very important for somebody to get free traffic from a search engine, you're optimizing for that. Now, of course, now people will talk about something very different saying, "Well, I want to make sure that I'm showing up when a large language model is going through and coming up with a recommendation. Somebody prompted, I want to go to Paris and I need a hotel." Well, every hotel in Paris wants to say, "How do I end up showing up in that answer?"

And the issue is nobody really has a good sense. There [are] different ways. How do I make my hotel content more optimized so it shows up more? And lots of different things. From our point of view, there are lots of different things that we need to do to make sure, one, working with all the frontier players so we are in a way that we can be seen, but even more so, so we can work with them to help them provide income revenue. And that's why, for example, with ChatGPT, OpenAI, we have a relationship with them in terms of advertising. So what we are able to pay to make sure that we are showing up as a way someone could book whatever the recommendation, whatever's coming out of the content.

There are other things too being played with the different frontier players, but that's just one small portion. What's most important to me is developing a process that people want to come to us direct. People want to begin their travel discovery with us, stick with us. Right now, we have about 65% or mid-60% of our customers come to us direct in the B2C area. That's really important. I want to maintain that percentage, but I want to even increase it. I want people to think the best way to do my trial process is going from thinking “I want to take a trip” to come directly to booking.com and go through it from start to finish as opposed to going to say Gemini, or Claude, or OpenAI's ChatGPT and starting there and then switching over to us.

It's interesting, many people are using the large language models as a way for discovery. Not a lot of people are actually then directly booking. Even though we have APIs, we have ways for people to do it, as do our competitors, as do suppliers. People have not yet developed a sense of trust with those models that they want to put their credit card number in there and actually execute a transaction. And I think that barrier, I think that's going to be there for a while.

Sarah Elk: Glenn, how does that impact your path to agentic? You made some recent announcements on big investments in your journey towards being more agentic. How would you describe where you are now, where you're going? Any lessons learned along the way?

Glenn Fogel: Well, obviously there are a number of different parts of our business where trying to create agentic systems is important. So we talk a little bit about wanting to come up with the connected trip for the consumer traveler, but we're a two-sided marketplace. So we have a different customer that's called the supplier or partner, the hotels, the airlines, the car rentals, attractions because nobody goes on a trip to sit in a hotel room. You go to do stuff, that's our attractions. And then there's the issue of how do we make our own internal organization more efficient?

So we have all of our developers making sure that they are doing what they need to do in terms of using the most recent, the things that are the cutting edge in terms of using agentic coding, making sure we're getting that out productivity. It's not just coding by the way, it's throughout up and down the organization. I'll give you an example. We're already seeing tremendous progress and that's in customer service. So I mentioned that travel can be very frustrating and you reach for the phone and you call. And unfortunately for many companies, many vendors, you end up on hold for a very, very long time waiting to talk to somebody, and it's incredibly frustrating.

Well, with us using agentic systems and such, people call and we can have an answer to that immediately. And it's a voice answer. It sounds like a human being is giving the right answer. One of our companies is OpenTable. So you can use our, the system is called Concierge. And you go on and it's like you're talking to a hotel concierge about restaurants and you're having that conversation. And that is really powerful use of GenAI to make things easier, faster, more efficient.

Andrew Ng: Actually, I've used the OpenTable Concierge. It's a pretty good restaurant recommendations. It's not bad.

Glenn Fogel: Thank you very much. I'm glad we have a customer. And if you have any problems at all, I'll give you the name of the people at OpenTable. You can talk to them.

Sarah Elk: You can talk to his agent.

Andrew Ng: Yeah.

Glenn Fogel: Thank you.

Sarah Elk: The pace of innovation in the consumer segments certainly seems to be faster than other industrial businesses or other B2B businesses. I'm curious how you think about the pace of innovation at Booking and how do you know if it's fast enough?

Glenn Fogel: Well, you never do know if it is fast enough, but it's something that I've said a lot to people and it sounds cliche-ish, but it's true is that speed wins, but speed also can kill. And the problem is you have to be safe when you're going fast. It's always a question, are we going too fast so fast that we make a mistake? We do something that actually is going to cause tremendous damage. And actually that's kind of almost the question of all technology when you get it up to scale. The issue is how quickly and how well are people considering what the potential consequences of it are?

If we just keep on developing new technology and pushing out very fast, we're not spending enough time thinking about what ramifications for society are going to be. We've been making a serious error. The benefits of a new technology, the shiny new things are seen instantly. The benefits are seen instantly, but the effects, the long-term problems of them aren't usually recognized until way late. And the problem is you've now gone that far down with this new technology, it's very hard to pull it out. And we can come up with a whole bunch of technologies that we may wonder, gee, maybe we should have thought about this a little bit more before we started actually implementing them.

Andrew Ng: AI certainly has problematic applications where you look very broadly across society. So we need to squash those and put up guardrails. I'm curious, focusing on the travel industry, are there any outcomes you worry about?

Glenn Fogel: I worry about the standard ones that aren't necessarily because of the technology, though technology can make it easier to be abusive about it, but all the things about fairness, equity, the ability to not do something that is harmful, whether it to be to individuals or communities, and it's easy to use technology the wrong way. And that's not because it's bad technology, that's because [of] how people are using it. But I do recognize that in the travel industry, it's one of the things where how do we make sure that we are maintaining a fairness issue?

Particularly when you think about making sure that all people, all consumers are given the same opportunities. One of the issues being debated right now in the States is the issue of surveillance pricing. And you can use technology to individually price people. Is that fair depending on who they are? It's a big debate. And then in democratic societies we'll have debates about and our elected officials come out, what should the rules be for surveillance pricing? Some things are good about it, some things are not so good about it.

Andrew Ng: No, that's a great point. And I think also about fairness to consumers, also fairness to small businesses. And if it's good for society, for large businesses and small businesses to be supported and to the extent that large platforms have the ability to steer traffic, making sure people are treated in some sense as fair. So I guess there are a few things that have to be thought through.

Glenn Fogel: Well, that's a good point, Andrew. And just following up on that. So one of the things I haven't seen too much data on it now, but when somebody is looking for a travel answer, they're prompting for travel information, let's say it's hotels and you're looking for hotels in a particular city. Well, the way the data has been in the models has gone in, do some of the hotels that are just much more prominent because they're so much more prominent in the data set.

Let's take the largest hotels, the ones that are well known, the chains, et cetera. And so they come up more as an answer to the prompt versus the small individual one. So we end up in a situation where the small individual hotels are never showing up as a potential solution unless somebody's prompting, “I want small hotel.” If they just prompt, “I need hotel,” you never see the small hotel coming up at all, or is that okay? And that's another thing we just have to think these things through.

Sarah Elk: One of the things that I think is unique is the number of different places that you're working on AI and agentic solutions across your businesses and across your portfolio. A lot of other companies are struggling to get started, struggling to find leverage points. How do you think about best practice sharing? Are there ways that you are intentional about applying learning across the portfolio, for example?

Glenn Fogel: One of the key things this early in this process of learning about GenAI and how to do it and how to use it is not trying to constrain or centralize and say, we're all going to do it this way. Instead, what I'm doing is I'm saying, everybody go out and learn on your own, come up with the best things are. Then we'll share what are the things that are winning and working and what are the things that actually aren't working at all? And we do that by having the different tech groups have discussions and being able to talk about what's going on, what's not.

And that helps particularly for the more sophisticated, really in the guts of technology who are doing that. But even more so, one of the things that I've been doing, and I've been saying this a lot over the last two years, it is incredibly important for every single person who works at our company, 25,000 people, every single person is responsible for learning how to use AI to improve how they do their job. And I say don't do it just because it's good for the company. Do it [because it’s] good for your career because you need to become as literate in using technology and AI as you were in learning how to use spreadsheets, how you're [learning to] do email, doing Slack, whatever.

Sarah Elk: That sounds like you, Andrew. Everybody needs to harness the power of their computer more.

Andrew Ng: I do warn those people that ChatGPT had come out years ago, I still see people, they're prompting like it's 2023. And the tools have advanced so much that even if someone's been using it for three years, if they're using it like they were three years ago, that's not anywhere near what you can get all of them now.

Glenn Fogel: And the [thing] that's unfortunate is that if you want to learn how to do it better, it's not that hard because you can use the large language model to help teach you how to use it better. It's a great teaching tool. And even then people said, "Well, I don't know how should I even begin?" I said, "Well, just start prompting.” “How do I do that?" And it'll eventually develop a course that you can do it.

Sarah Elk: What are the ways that you reinforce that? I know Tobi Lütke at Shopify issued the memo that came out maybe a year or so ago, and that was sort of a moment and everybody stopped and paused and was like, "Wow, he set expectations pretty clearly." Is it just cascaded through your leadership team? Are there ways that you talk about that internally? How do you reinforce the intrinsic motivation for your employees to learn?

Glenn Fogel: So it's trying every single meeting, every single time I'm saying it, I'm saying it loudly, and then having my leadership team cascade it down and then have every single person such that it becomes part of the culture. You really want to have it be embedded in everybody's mindset that this is really important. Now it's helpful because if you keep on pushing the point about how important it is for your career, that helps get it going a little bit more.

Sarah Elk: I guess what I want to know speaking to somebody who deeply knows the travel industry is where should we go?

Glenn Fogel: Well, we want to personalize that. I don't know you well enough to be able to give you the [answer].

Sarah Elk: Oh, there we go. Personalization.

Glenn Fogel: Exactly. Exactly.

Sarah Elk: It's interesting though. So if we think about in the age of AI, what data would you capture now so that you could better give a personalized recommendation to me?

Glenn Fogel: And this is what is also interesting. In the past, we would've been doing things in cohorts. We would've been doing psychodemographics of who you are without really knowing who you are. But the wonders, the real wonder is that we'll be able to know everything about you to help come up with a better solution for you. Knowing where you have gone in the past, where has the rest of your family gone in the past? What are the ages of people in your family? Do you like hard pillows or soft pillows? What do you want in the mini bar? You want Jack Daniels or did you want some Pinot Noir?

You can personalize so much of that experience all the way down that it makes it so much better. Look, here's the way I always think of it is that I'll be honest, I like it when my spouse does all the travel organization for the family because it's a lot of work. She'd like somebody else to do that too. And the thing is that agent, that agentic travel agent that we're creating is going to do it for everybody and everybody's going to feel better about it. That's what we want and that's where we're going.

Andrew Ng: How much do you think the future row of agents will be doing the work of personalization based on my preconceived preferences versus shaping and taste-making? So for example, in the fashion industry, I have no idea what's supposed to be in fashion a year from now. Heck, I don't even know what's supposed to be in fashion right now, if I'm being honest. But it's not that my personal preferences on what will be in fashion a year from now drives my buying decisions, is that there's a process of society to come to conclusion about what should be fashionable.

Sarah Elk: You could change which blue shirt you wear.

Andrew Ng: Yes, I might actually buy a totally different blue shirt. So how much do you think AI will personalize existing preferences versus shape preferences?

Glenn Fogel: That's a great question. I imagine there's going to be a little of both happening, but society has always shaped people's preferences. So whether or not it's coming from an agentic system or not, people have always been influenced by what, let's say, the majority like or don't like. And fashion is a good example. I know nothing about fashion either, but Andrew, me and you together, we can hang out together because I don't know what to wear. But I do know that that is an issue of human society that everybody says, okay, this is what we should do now. And then a year later, that's what you're supposed to do.

In terms of travel, though, I think it is a very individualistic type thing. That being said, we see the impact around the world of mass social media where areas that were beautiful places to go, but as soon as people started taking Instagram photos of it and it became overrun with so many people because everybody saw the same photo, how beautiful it was, and now all of a sudden with that many people, it's not so beautiful anymore. And we're seeing over tourism because of I would call the Instagram effect. And that's happening in unfortunately too many places around the world that are having to deal with, well, what do we do? We have too many tourists to deal with.

But in regards to the personalization thing, I think that's an easy thing to solve in terms of it's not only looking at you from your own history, you get to choose. You have agency over what you want it to use or not. And if you want it to stop using, well, yeah, I went to Vegas, but please take that out. Do not consider that I went to Vegas ever. That's something I think is easily solved.

Sarah Elk: As somebody who's very far ahead on this journey, I think relative to other industries as we've discussed, what advice do you have for other CEOs?

Glenn Fogel: Well, advice for other CEOs. Everybody's in their own situation has to obviously personalize it for their organization. That being said, a couple things I think are always good. One is culture. And that's uniform across all organizations in terms of trying to create a culture that wants to achieve something. I really believe that there are different types of companies [and] that some are more successful than others. And part of the reason is because the people on the team want to do better, want to win. They have an objective they want to win. They're not there not showing up just to get a paycheck. They're there to accomplish a mission, to achieve something.

So trying to get that mission across so people feel that they're doing something that matters. So one of the things I talk a lot with our team is, look, I know we are not curing cancer. We're a travel company, but what we're doing is important to people. And the way I prove it to people is I say, “Think about the things that have been most meaningful to your life. And I say, maybe it's when you got married, maybe it's the birth of your kids, or graduation, but I'll tell you, I'll bet your travel is way up there. When you go with your memories, what are the memories you remember the most that are most fond?”

Maybe it's when you were a kid with your family traveling. Those things matter. And what matters is getting it right. It matters if we do something and there's a problem, we're able to fix it. And yeah, okay, so we're not curing cancer, but we're making hundreds of millions, not millions, not tens of millions. We're making hundreds of millions of people happier. And that's really important because the world needs more happiness.

Andrew Ng: Yeah. With the tragedies going on occasionally in different parts of the world, I think it's a good reminder of how small and how connected the world is so that travel not only makes people happier, it feels important for bringing people together as well.

Glenn Fogel: Thank you for saying that Andrew. I say that also all the time and I really, really believe it.

Sarah Elk: Well, thank you so much for joining us. We really appreciate it.

Glenn Fogel: Well, thank you so much for having me.

Sarah Elk: If you want to learn more about Winning with AI, please follow us on Spotify, Apple, or wherever you get your podcasts.