Uber CEO on AI Travel Agents, Robot Drivers, and the Future of Travel

0:00ZS is a global consulting and technology firm that partners with companies to improve life and how we live it. Working with the world's top hospitality brands, ZS brings together data, science, technology, and human ingenuity to transform ideas into customer loyalty. They're ready to help you elevate guest experiences, improve operations, and uncover preferences and motivations through their AI-powered solutions. With more than 1 ,700 clients in over 170 countries, ZS has built the expertise, fortitude, and vision to help you deliver impact right where it matters. Hey everyone, it's Kelsey, the producer of the Skift Travel Podcast.

0:41This week, Seth and Sarah are at the Skift Commerce Summit, so keep an eye out for an exclusive look into their live session for next week's episode. But this week, we're taking it back to Skift Global Forum, where the CEO of Uber took the stage. He gets into AI agents booking rides for travelers, why his biggest competitor is the car in your driveway, and his prediction that within seven to eight years, every car sold will have a robot driver built in. He also explains why Uber is willing to break even on hotels. If this sounds interesting to you, we hope you enjoy this look into the live session from Skiff Global Forum.

1:17Let's get into it.

1:41Please welcome CEO of Uber, Dara Khosrowshahi, in conversation with SCIFT CEO and founder, Rafat Ali.

2:02Thank you, Dara. Thank you, Mabee. Happy to be here. Let me start with softball. Softball. I like softballs. This is an insider joke. If you don't remember, you may not remember this, but you said to Dennis, come on, softballs, when you were expediated. And that became for us a viral moment for us internally. I heard you took a bike here I did take a bike here The traffic But it was not available? Traffic during UNGA drives me crazy So sometimes it's better to take a bike Do you work with cities on these big events The surge that happens Besides obviously the surge pricing, etc, etc Do you actively work with cities to say How can we help reduce congestion?

2:46Yeah, totally So we work with cities We work with venues We work with airports, etc And obviously, we've got algos that learn and anticipate what's going on to make sure that we are prepared for large events. So, for example, the World Cup here and a few other cities, we work very, very closely with events. But they have to know where to drop off people. Yeah, listen, it can't all happen randomly, etc. So events are a big part of what we do. we'll use incentives for example to get drivers to work during events when events are over etc and position them to where we want to go and then we'll work with the venues to have proper drop-off pickup locations as well and then program that routing into the algos to make it as smooth as humanly possible we don't get it right every time but I do think that we're by far the best in the business as it relates to the event area.

3:41Was the World Cup good for you? Great. Great. It was very, very good for us as it related to mobility, but also actually unexpectedly, or maybe not, the delivery business did better because a lot of people stay home, they watch games, and they order food. That's true. Let me ask you a question. So I came back from India 10 days ago. I was flying in three young kids. and sitting right. I'm sorry. Yeah, no. My son, one of my sons is somewhere there. All right, cool. So, and you know, been sitting on a plane for whatever, 15 hours. I wanted Uber to not wait for Uber. And I wanted to pull up as I'm coming out.

4:25So I used. You want the Uber to pull up as I'm coming out. Just like with the kids after this long flight. So I used the instinct. I don't know if you know instinct or muse. Instinct or muse are these new personalized AI agents that are sort of making the rounds and becoming very popular just in the last few weeks. And I used an instinct. My first time ever using instinct was on the flight because I had Wi-Fi. And it worked. As in, as I was landing, so it told me, as soon as you get the back, just text me to say Uber. And it worked out. So my question is, do you care? Two, advertising is the fastest growing business and I'm not seeing the ad.

5:11So how are you thinking about that? So I think, do we care? Yes, but we care the most about your personal experience. And so number one is we are working on making that airport experience for you better. So that you come to Uber and we anticipate, for example, call your Uber when you're a baggage claim. it'll take an uber uh you know seven minutes to get there so that's about the walk time etc so we are the airport experience is experience that we are still honing down and getting better so that you don't wind up at the pickup area you're waiting uh we have a pin dispatch now where basically the cars just line up really good they're they're lined up and all you need to get is a pin you just get a pin number and like a taxi line yeah exactly so by the way what's what's old is new again, right?

6:00Pin dispatch is exactly like a taxi line. We've got to work with the airport authorities where I think pin dispatch is the best experience where the driver's just lined up for you, you get lined up, you get matched up to a driver and you just go. But number one is we have to make sure that that experience for you at the airport, because we've got more information than anyone else, peak periods, trough periods, et cetera, is best done through Uber. That said, the world is changing. And the great news for us is Uber has unparalleled content as it relates to your local logistics. We operate in 70 countries, well over 500 cities.

6:43If you're an international traveler, you're probably using us in one way or the other. So we are having, as you would expect, active discussions with the instincts of the world and the muses of the world and basically every other agenda company and every other AI company because they can have one conversation with us, secure content globally, secure content both for mobility and delivery, but it has to be on the right terms. First, we want to make sure you've got the right experience and the quality is there. And then second are kind of the terms of trade. We are very, very early in those discussions, but because of our unique access to supply, We've got 10 and a half million drivers and couriers on our platform.

7:30No agentic company is going to replicate that content. So we are at a very, very good position as it relates to these discussions. And for me, I don't want to block that perfect experience. That sounds like a great experience. I've heard really good things on Instinct. We want that experience to happen. I mean, I could have done it on the Uber app as well. Of course you could. Yeah, it's cool. It's just one of the things I wanted to test out and it worked. For those of you who don't know, yesterday's news, you may or may not know, but Amazon banned Muse, which is a Facebook personal agent, their boss from buying on behalf of users.

8:04And their rationale is security, I think credit card, the safety of the credit card numbers, et cetera, et cetera. Which is a real concern. Totally. Totally.

8:19So on the AV side, on the driverless car side, by the way, my first Waymo using Uber in Atlanta was amazing. It's like life-changing. Awesome. And then you come out, it's like a new person. So cool. And then I called my kids and my wife, et cetera. I showed them all this. Check it out. My kid is there back there. He's 11-year-old. And then I have two young, seven and six. Will they need to learn how to drive? Need? No. But, you know, my kid is 20 and it drives me crazy. He doesn't know how to drive. He just uses Uber, right? So I think that... On whose credit card? Just take a guess, Raf. So I think that the AV technology is incredibly exciting.

9:06It is going to make our streets safer, right? These robotic drivers don't get tired. They don't get distracted. They continuously learn they will be driving many more miles than any other human being could. But the technology is super expensive right now in terms of the hardware and the sensors and the compute. And it actually takes a significant investment in both infrastructure and services to take care of these cars. You know, the job of the driver is not just to drive. They own the car. They clean the car. They refuel the car. There's a lot that they do. They are their own fleet manager, so to speak.

9:46All of that work is going to have to be systematized. And we are in a great position to systematize that work and do it at scale in all the countries that we operate in. So our view on AV is promising technology. We want to introduce it into the network, just like I want every safe human driver on my platform. I want every safe robot driver on our platform. And because we already have a density of demand and supply, we can offer our AV partners and our human partners more demand than anyone else. That's why we are the leading partner as it relates to human drivers. And the same is going to be true with AV drivers.

10:32So we now have partnerships with AVs live in 15 cities, both in the U.S. and outside of the U.S. And London, I hear, is coming. London, we have Wave and Nissan, as well as other partners that we anticipate in London as well. We're live in Dubai, driverless. What's the holdback here? The holdback here is mostly regulation. And by the way, I do think that this transition is going to happen, but we have to be careful about this transition because at this point, robot drivers will make a mistake. There will be accidents. There already have been accidents. And to some extent, our view of the world is the danger of going too fast is getting the same kind of backlash that you see from the public in terms of AI, et cetera.

11:21There will be some job displacement, but to the extent that you are introducing this technology in a measured way, in a safe way, in a way that meters whatever job displacement there's going to be, that will be kind of a successful introduction into society and into the streets of New York, London, etc. But it's going to take some time and it's going to be hard because there are many cities that have said no to AVs. And it's more because of generalized AI fears that are translating into AVs as well. So the technology is great. It will take time. I do think that, and then over a period of time, it's going to commoditize.

12:06How do you prevent yourself from not becoming a commodity versus, say, Waymo or Tesla who have their own apps? Well, listen, there are going to be players who have direct channels. You know, Starbucks. Which is, by the way, reminds me of the hotel OTA, which is direct versus OTA. Exactly. So there isn't this, there's this kind of drama in the press, which is, oh my God, everyone's going direct. Oh my God, marketplaces are going to win. The fact is that I think the effect of AV is going to make it easier for people to use on-demand transportation. Our number one competition is a personally owned vehicle.

12:41And so I think that the loser in all this is going to be the personal own vehicle. And is Waymo going to have a great app? They will. Is Tesla going to have a great app once they figure it out? Because their approach is very, very difficult, different and more difficult. They will. But we also think that they will also be on our app as will NVIDIA, which is building open source technology. WeRide, AvRide, Baidu. There are lots of companies building this technology. And I do think to the extent you buy your own car seven years from now, just like you expect, you know, it'll come with Apple Play or Android Play, whatever you're looking for.

13:20It will come with a driver. And that will be built into every single car built out. And it might be WeRide technology. It might be NVIDIA technology. It might be Waymo technology. but every car sold, you know, seven, eight years from now is going to come with a driver installed or it'll be an upgrade. By driver, you mean driverless. Sorry. A driver. It'll have a driver. Yeah. You can be a driver or can we pull up this video? I just want to see if this can be a driver. Uh, just with the video that I, uh, Uh... This jokester from Tesla Bot. I've been a cyrus. Man, you've seen one of these before.

13:59Whoa! Warner, Sean. Uh, yeah, I'm Sean. What's up, man? But if you do that again, you're going to get hurt. May I take your luggage? Um, no, I'm okay. Okay, thank you. I said no.

14:15Hey, man, I got some... Very aggressive robot. Yeah, man, no problem. Oh, come on, man. What are you doing? Why are you Uber driving? Electricity costs are high. Electricity costs are high. Okay. Let's go. Okay. That driver would get a bad rating, but it's pretty cool. Silly video. But I don't know what my follow-up question was. There was some follow-up question. On the humanoid robots part, how real, from your perspective, to where you're sitting, the humanoid robots versus the AVs? And do you think they're converging at some point? I think that humanoids still have a long way to go. I think that the use cases, you know, there are players that are going for generalized solutions, but I think that the first use cases that you're going to see are more specific solutions.

15:10There'll be industrial solutions, and there'll be kind of personal home solutions, robots that fold, that fold, close, et cetera. I think for delivery specifically, the much more exciting use case for us is actually drones. And so there was a lot of hype and Jeff Bezos launched the drone stuff a year ago. It sort of petered out. Listen, everything early on comes with hype and then that hype turns into reality. It turns into, you know, magical reality, we think. We have, for example, invested in a company, great company called Zipline. Right. And it has, it's already doing drone delivery for Walmart in Texas.

15:48we're going to launch this year with them. And it will be drone delivery or food. And the thing that's great about drone delivery, you know, AVs are safer. People like the privacy. Some people like the privacy. Some people like the company of humans. They're slower. They are slower. They're slower. And they only drive at the speed limit. That's what I learned. They obey the law. Imagine, right? And then there are some pickups and drop-offs that they can't do at this point, right? So there are trade-offs on AVs, and that's why we think actually an AV inside of a hybrid network is a better solution, because sometimes a human is going to be able to, you know, if you're hurrying to JFK, I think you're going to want a human for that drive.

16:39Or there may be pickups and drop-offs that an AV can't do for whatever reason, World Cup, for example. Drones will have the same reliability, but actually they're going to cut your delivery time from 35 minutes, 30 to 35 minutes, to 15 minutes. And what we've seen in certain markets where quick commerce is in India or the Middle East where like 15-minute deliveries are a thing, the e-commerce, the on-demand e-commerce sector for food, for grocery, and then for general retail, it just takes off. Are you in the instant delivery in India? We are not in delivery in India. We're only in mobility in India.

17:26but we're getting into the instant delivery game. Cargo, right? Cargo? Truck? Trucking? No, we have Uber Freight, but we are going, we announced an acquisition of Delivery Hero and so that is, they are in instant commerce so we will be in the instant commerce game in a very, very big way, hopefully sooner rather than later. Okay, so let's move to the people who are sitting here in the travel industry so let's talk a little bit about travel. Sure. Original love, can I say that? Absolutely. You shipped hotels, a travel mode, voice bookings, and one search all in one year. Is that one bet about the future of travel?

18:05Is that five separate bets? And what is the bet? So it's a generalized bet, right? I won't pretend for a second, Prof, that I know exactly what's going to happen. It would be nice for me to come up here and state that. But when you look at Uber, because of our position and because we operate in 70 plus countries with delivery, we'll be in 100 plus countries. The traveler comes to us. Often we're about 10 % of all airport pickups and drop offs. We are giving 100 million rides to and from airports. We give about a 1.5 billion rides to people outside of their home city. Right. So we know when you're traveling.

18:50Right. So we have a very, very large and very targeted travel audience inside of our ecosystem. Our U4B business is huge. We do, you know, corporate rides for many, many companies and often they're going to airports. So the hypothesis that we had is that we have a very large travel audience. And how can we serve that audience better? It starts with the airport experience that you talked about. We introduced train bookings in the UK. And now we're in France and Spain. And what we're seeing is this is not some aha. But the more people use us, the more they use us. So Uber equals on-demand in my mind.

19:39It does. But then we introduced reserve and that's not on demand. So more and more, we're able to extend the definition of what people thought Uber was, right? Uber used to be, when I joined, 95 % of our business was mobility. Now we're 50-50 between mobility and delivery. So we're demonstrating the capability to stretch the service. We started with trains. That has worked out really, really well. With hotels, our angle at this point with hotels is our Uber One membership program. We've got over 50 million members, growing 50%. Our Uber One members are - Just showing you from the hotel. Yeah, but three times more than non-members.

20:22And basically, whatever money we're making from hotels, we're handing back to the Uber One member. So you're basically losing money. We're breaking even on it. Okay. But basically, our members get 10 % off hotel bookings. There's a rotating list of hotels where they get an additional 20 % off. It is a spectacular deal. And Uber One members spend three times more than non-members. Travelers, international travelers, especially, spend three times more than non-international travelers. So we're finding a pocket of very, very attractive users. We're giving an unparalleled deal back to our members.

21:00will grow membership. And how long can you subsidize this? We're break even and we're not, you know, I'm not a prisoner to Google like I used to be in terms of traffic, right? We people come to Uber, Uber Eats, the vast majority of it is natural traffic. So because I don't need to pay for my traffic, there are certain products like hotels at this point that can essentially be a break even product. That's about deepening the relationship that we have with our customers. and then we will continue to deepen, I believe, our travel content because it's a natural thing for us to do. What's next at this point?

21:40We really want to focus on hotels and make that an absolutely great experience. Maybe, you know, with hotels, for example, we're adding to that, you know, we call it room service, right? Like room service through Uber Eats is way better than classic room service and we're trying to integrate with our hotel partners. Yeah, it's cheaper too. So we just want to be a better partner for travelers. And then where that takes us in terms of what other travel services we offer, that's a TBD. And so flights may or may not come? May or may not come. I want to prove that hotels works. I want to prove that there's product market fit between kind of travel bookings.

22:23Like you said, we are more of an on-demand, immediate gratification product. So it's one step at a time, order of operations now. The 80 % of our bookings in hotels are from our members. Our members absolutely love it. We're going to optimize that experience for a while, and then we'll think about what's next. Why did you not build your own supply? You know this. Because you got a ton of companies like Expedia's bookings and many others of the world who have incredible supply. And we wanted to hit the market fast. And I think duplicating what they've already done isn't adding value. Are you still on the board?

23:04I'm still on the board of Expedia, yes. So I guess you can't buy the company? I pulled out of the decision making and no, there are no discussions between the companies other than biz dev. So he's saying there are discussions happening.

23:22So it's interesting. I did a story on Skiff two days ago on Uber is coming from this side into hotels and Airbnb is coming from this side into hotels. Yeah. meaning you are known for what you're known for and adding hotels. And I guess now you're going to compete somewhat with Brian, which I don't know if you foresaw that. It's a big market out there. We do think we're going to add vacation rental content first with VRBO because of the integration we have with Expedia. And I think it's perfectly natural. You know, Airbnb is a great travel brand, And so it's not a surprise for them to look to extend their brand into other kinds of content.

24:03Just like we extended a brand from, you know, mobility content to delivery content. Now travel is another kind of adjacent ecosystem that we're getting into. So Brian is speaking tomorrow evening. What advice do you have for him? Because you've expanded beyond, as you said, from rides to, and he's trying to, and it's been hard to move from short-term rentals to experiences. He's tried a few times, and then now all this new stuff he's adding. I think one is that he's trying to add a lot of different services into his service all at once. And it takes time for a brand to stretch. Now, I think Brian is one of the great entrepreneurs in the world.

24:45So I think the advice that I give him is get signal as to what's working, and with things that aren't working, kill it. Which is the hard part of him. That's the hard part. Listen, it's all hard. Business is hard, but often the most difficult decision that companies make is, hey, how do I actually get rid of services that aren't adding to the whole? And you did a bunch when you came in. What's that? You got rid of a bunch of stuff. Of course. How many years has it been now? Nine years, over nine years. Yeah, it's been a while. Are you still enjoying it? Love it. Love it. It's the... This company and this job continues to just completely jazz me up every single day.

25:32I absolutely love it. Much harder than you thought it was. It's not fun if it's not hard. Like, it's just like I want it to be hard. What is the hardest problem you're trying to solve in 2027? 2027. Seven. I think the hardest job, and I put hard in quotes, is how do we leverage AI in a real way? Okay, like there's so much BS going on in terms of AI and AI this, AI that. And everyone says that they're an AI company. And actually integrating AI truly in how you work in a way that really moves the needle, that's not slideware. It's hard. Well, you, your CTO, maybe it was you who said you blew through your token.

26:21Three months. In three months, the whole years. For those of you who don't know, they're famous for saying that they blew through their whole token budget for the year in the first three, four months. Yes, yes. Which means what now? You had to reset the budget? Which means I had to redo my budget. But it, you know, it shows you the power of AI. And we continue to invest very, very aggressively in AI. but we are also investing in making sure that we are getting our dollars worth. And that means working with multiple AI providers. That means making sure that open source is a big part of what we do.

26:58And again, as it relates to the application of AI, and I know Brett is going to come up here in a little bit, to actually build out solutions that add value. For me, the definition of adding value isn't just taking out cost, but actually improving the service itself. is a lot harder than it sounds. And it's harder, especially at the scale that we operate in. So those are the discussions that I'm having. It is... What's the skunk quirks within Uber? Say this succeeds, then I'm going to use it in other stuff. We're doing it everywhere. I mean, it's... One is just larger models for us are working much better, right?

27:40So it's just like that is very, very simple AI. larger models based on transformer architecture that if you look at our system to some extent, it's a bunch of different systems that are optimizing for local optima. So we have a pricing system, we have a routing system, we have a matching system. All of them are trained on very small corpuses of data. They're optimizing decisions that they are making on that specific vertical, we're able now to build much larger models that take all of the context of the users and our couriers and our drivers and are able to make decisions based on the whole system versus the narrow slice of the system that we've trained the model on previously.

28:25That in and of itself is really powerful. Complex as well. Yeah. Right. Then you work on AI that's driving productivity, whether it's your developers or it's your operational analysts, et cetera. Uh, and then you've got AI that is, um, is, is hopefully revolutionizing customer experience, merchant experience, courier experience. And then you've got agents that we want to build so that next time on the, you know, you come to the airport, you can just talk to Uber and say, I'm a baggage claim, have an Uber waiting for me. I need an SUV and maybe we've got your hotel booking already there. Right.

29:04Right? That agentic experience, that agent is going to have way more context than other agents. Correct. And should be specifically tuned to our systems. And as a result, should be better. But that agent can also negotiate with third-party agents as well. That space is a really, really exciting but early space that we're working on. Thank you, Dar. Absolutely. We're going to have Brett Taylor come up now too. Okay.

29:44ZS is a global consulting and technology firm that partners with companies to improve life and how we live it. Working with the world's top hospitality brands, ZS brings together data, science, technology, and human ingenuity to transform ideas into customer loyalty. They're ready to help you elevate guest experiences, improve operations, and uncover preferences and motivations through their AI-powered solutions. With more than 1 ,700 clients in over 170 countries, ZS has built the expertise, fortitude, and vision to help you deliver impact right where it matters.