20VC：Uber总裁谈Uber如何为AI编预算：AI如何帮助并伤害Uber｜自动驾驶为何关乎Uber生死｜如何击败DoorDash成为外卖第一｜Andrew MacDonald讲述Travis Kalanick、Dara Khosrowshahi与中国的未曾披露故事

We're doing 300 million trips a week. We were burning $52 million a week in China. We were competing in China with one hand tied behind our back. Autonomy is as bad as it's ever going to be today, right? Every single day, it's going to get better.

In the end, distribution wins. We could do everything we do today with fewer people in 5 years because of the power of AI. No one has been with the company longer than me at this point.

This is 20VC with me, Harry Stebbings, and I'm so excited to welcome one of the greatest operators of the last two decades to the hot seat, Andrew McDonald. He's the president and COO at Uber, where he leads all businesses initiatives for the company across mobility and delivery. He is Uber's longest-tenured active employee. And today, Uber's an absolute monster. They have a market cap of $160 billion, revenues of $52 billion in the full year of 2025, on over a quarter of a trillion in annual gross bookings. They have 200 million consumers that use the app monthly. This was a behind-the-scenes on Uber like we haven't seen before. Mac was one of the greatest operators that I've been fortunate enough to have on the show, and I think that really comes out in this discussion.

You have now arrived at your destination.

Mac, I am so excited for this, dude. I've wanted to make this happen, and we've been DMing for a long time. It's so good to do it in person.

So great to be here. And you're right. I remember the first Twitter DM from you, and I was a bit of a fan from afar, as you probably hear often, but great to be here now and so glad to do it in person.

I spoke to Dara before the show and I said, “What's his superpower?” He's been here for over a decade, 12, 13 years.

14 years. It'll be 15 in May. No one has been with the company longer than me at this point.

My God. I said, “What's his superpower?” And he said, “Oh, very simple. People really like him, but he is an execution machine and he is very good at driving people.” I suck at that. So can you—no, seriously, how do you do that but retain people liking you?

1. Building Trust Through Execution

I get the question in the context of career advice. People join Uber and say, “You've been very successful at Uber. How do I—how should I be successful here? What did you do to get successful?” It's hard because every formula is different, but there are a couple of things I say.

One is—and I think the most important thing is—if you genuinely are trying to do what you think is the right thing for the company, and that is your filter, and you build trust that that's what you're optimizing for all the time on every decision, big and small, that you are using the lens of what is the best thing for Uber. You're not always going to get the decision right, but if people know that you're filtering on that, then I think that builds followership and trust over time.

Then you can move people, because if you're pushing on something, they know it's because you genuinely think it's the right thing to do. If you pair that with a deep knowledge of the business—and I've grown up in this business, so I know ride-hailing better than anyone in the world at this point—those two things together are pretty powerful.

You do literally know it better than anyone else in the world.

It's hard. It's only been around since 2009. I've been working on it since 2012, and most folks from that time period are not working on it anymore.

You said when you genuinely believe it's the right thing for the company. What did you genuinely believe was the right thing for the company and it turned out you were wrong?

I am wrong every single day—big things and small things, right? I think it's a Bezos quote: “If you want to be right most of the time, you have to change your mind a lot,” or something to that effect. Effectively, you're going to be wrong a lot, and the people who are successful over time are willing to change their minds. I think that's true. I'm wrong a lot.

2. The Membership Mistake

This is probably the most common running debate that Dara and I have had over his tenure as CEO. Thematically, it's the tension between short-term levers and long-term levers for the business.

A short-term lever for ride-hailing is price, right? Every dollar that we can put back into lower pricing, I think, is valuable. Even if, long term, there may be other things you want to do—acquire new users, build a membership program, or build new business units—you should be weighing those investments against the fact that I could just put a dollar back into price.

I think I have been too short-termist on certain issues, like membership—Uber One, for example.

Well, I'm smiling because I said to Dara, “What is the single biggest disagreement that you've had with Mack that comes to mind first?” And he's like, “Say it's from me. Say it's from me.” Normally, people anonymize this. He said, “Say it's from me.” He was reticent about membership programs.

Yeah. Totally. And I've turned out to be wrong.

The reason I've changed is because, first of all, Uber One is, on many metrics, one of the more successful membership programs in the world. We're not at Amazon Prime or Costco levels, but we're getting within spitting distance. From a company-lever perspective, it's highly efficient.

When we look at efficiency, we usually look at, if I put a dollar in, what am I getting back in terms of top line? It's one of the best levers we have. The farther out we can measure it, the more efficient it gets and the better it stacks up against other levers.

So you're like, “Okay, Mack, how did you get that wrong? If that's what the data showed you, then why weren't you all in on membership?”

When I say I wasn't all in on membership, what I mean by that is that I would constrain the capital envelope we had in the mobility business to invest in it. If I had $40 million next quarter to invest, my gut was always, “Put as much of that into pricing as you can, or put as much of that into driver supply to improve the health of the marketplace so that service is more reliable.”

Ridesharing, at the end of the day, is price, reliability, and safety. That's all it is. That's what it was 10 years ago. I think that's what it's going to be 10 years from now, even when there are autonomous vehicles. It's price, reliability, and safety.

Putting money into something like membership, where people get a suite of benefits, part of which is price but the rest is a whole other host of things, means you're explicitly choosing not to put that dollar back into price. That's just the tension. I probably was short-termist in my thinking there.

You said that about dollar leverage—putting in dollars and seeing what you get out. What is the single most efficient dollar-in-to-dollar-out business for you today?

I think membership is the most efficient long-term consumer lever that we've got.

The reason for that is that, ultimately, we are looking at IGB as a critical input metric for any dollar I have to put in. So—

What is IGB?

Incremental gross bookings. Think of it as incremental revenue. If I put a dollar of incentive into the market—if I give Harry a dollar and I give a million other consumers a dollar discount—how much incremental revenue do I get back from that?

By the way, the ROI on that is different. If I get $2 of revenue back from Harry by offering you a dollar, you might be like, “Okay, that’s great. That’s a two-to-one ratio.” But actually, we only make 7.5% of your dollar from a profit-margin perspective, so you’re still negative ROI. You make those sorts of investments to grow the platform over time because I’ve increased Harry’s engagement, and then your LTV goes up over time.

We’re typically looking at a very baseline IGB-type, or incremental-revenue-type, metric for any dollar we’re putting into the marketplace. Membership just gets better over time. The reason it gets better over time is, if Harry becomes a member, not only do you ride more next month, but that cohort of members we acquired in that month tends to ride more over time.

Part of that is because they’re consolidating more of their mobility business onto Uber. Part of it is because you get some Uber Eats benefits with your membership program, too, so now you start using Uber Eats instead of DoorDash or Deliveroo. The LTV of Harry just goes up over time with membership. You’re less likely to churn, and you’re more resilient from a market-share perspective.

There are all these downstream, long-term impacts that multiply the value of that first dollar I put into membership. With shorter-term levers, like price or promotion, there’s some tail. If I give you a dollar to take a trip next week, there’s some value in the following weeks, but it tends to dissipate faster. That’s often the debate.

You said it’s not quite Amazon or Costco, in terms of membership dominance.

Yeah.

I think that’s fair. They’ve had a little bit more of a head start than you. When you look at that chasm between you, what do you not have that you would need to have to reach their dominance?

From my perspective, we need to put more consumer value into the membership program. Today, I think especially if you’re a mobility rider, typically on mobility you’re getting 5% cash back, which is our standard offer. The consumer comprehension of that is still relatively low. For a membership program as big as we are, I think there are still a lot of people who have Uber One and don’t fully realize the benefits they’re getting on mobility.

The other thing we need to do is look for features that are high perceived value and low cost. That’s the sweet spot of any membership or rewards program, for example. With our business, that’s tough, because I don’t have a lot of free things to give away on the platform. If I want to give you a ride because you’re a loyal member, either through a membership program or a rewards program, I still have to pay the driver to provide that ride.

It’s not like a hotel, where you might have excess inventory and your marginal cost of giving away a room night is pretty low. The beauty of our model is that we’re primarily a variable-cost model. That means when demand drops, our costs scale down with it. But it also means that we don’t have a fixed capacity to give away, and so it makes the challenges of building a membership or rewards program a little tougher for us.

3. Incubating New Growth Bets

What line of revenue do you not have today that will be very significant in 5 years’ time?

I mean, it’s hard, because “significant” for us is really fucking big at this point.

Right.

We’re approaching a quarter trillion dollars in GB. If you think about that top-line metric—what our GMV or GB number is—we’re not far off from being a $250 billion company. So for anything to pass the significance test, it has to be a multibillion-dollar business in terms of transaction volume.

I’m thinking of a new product I want to stand up, and it’s a mobility service that we’re going to offer through the Uber app. For that to even be interesting, I have to see a path within a few years to multiple billions of dollars of GMV. It actually constrains your thinking a little bit.

Yeah. Does that—do you worry that that prevents you from trying new things?

Yeah, totally.

Do you like a Google Labs? Go on, try something—try Gmail, Paul Buchheit. Do you know what I mean?

Yeah. We do. We try to set up structures to solve this problem. It’s a classic innovator’s dilemma problem: the thing you’ve already built is so big that it swallows up your organizational capacity to do anything else.

Even if you’re able to stand up other businesses, it’s impossible for those businesses to get the resourcing, attention, distribution, marketing dollars, and engineering capacity. Whatever it is, it just gets swallowed up by the whole. Part of it is even management focus. It’s very hard to focus on the new thing when you’ve got this $225 billion blob that you have to manage over here.

So how do you solve that? We run a program called Growth Bets, which is intended very much to incubate new businesses within Uber.

How does that work?

Basically, what we try to do is, first, create dedicated resources. If I’ve got, say, 2,000 people who work on our mobility business, I want to try to have 100 to 150 of them working on the new stuff—the small stuff, the stuff where we don’t have product-market fit or unit economics figured out, but that could be a big future business.

It requires dedicated capacity and thinking. If you try to do it as 5% of your job—if you run the marketplace for UberX in the US but you’re also trying to incubate this other thing with 2% of your time—it’s really hard.

Do you know who the best in the world is at this? Nik Storonsky from Revolut.

Oh, interesting.

I’ve interviewed 1,000 founders.

Yeah.

He’s the single best founder I’ve ever interviewed.

Yeah.

It’s because he runs 26 product experiments at once. He gives them $2 million, tells them to run for a year, and every single week he checks in for 20 minutes with each of the leaders. Then he determines whether to fund their next round or not.

Yeah, I love that. We have a version of that. It’s not the—I love the cadence of that, by the way. Operating on weeks, not months or quarters, is how a new business should run.

I also think having to sing for your supper and come back and ask for money is important. Again, the other challenge of standing up a new business within a big company with a big P&L and a big balance sheet is that people just get fat on the resources. You don’t build it the way you would build it if you were a startup, because you have more resources.

You end up moving slower, consuming more cash, and getting more heads than you otherwise would if you were starting from zero to one. As a result, it’s not that you necessarily build something better; you’re just slower, and you’re constantly chasing the people who are doing it from first principles. I think that’s hard.

You have advantages. We have distribution. Distribution is—

Which is the mother of invention.

100%. If you can actually build something interesting and then plug it into 200 million consumers who use our app monthly, you’re going to be able to scale way faster than anyone who’s doing it without that distribution advantage.

Even figuring out how to do distribution right—the 200 million number is attractive, but even within that 200 million, there’s tons of internal discussion and debate around how we spend our pixels. Every new product wants CRM support. Every new product wants to be featured on the masthead of Uber Eats or to be in the product selector for rides.

How you make those decisions as an organization is tough. But you still have this built-in distribution that is super interesting, and so it’s an advantage. You’ve got to figure out the other stuff, which is how you stand up new products in a company.

Other than time, what is the one inhibitor to getting to 500 million users? You mentioned 200 million.

I would say our IR team is not going to love this answer, because I would say price. The reason our IR team won’t love that answer is that when you start talking about price in the context of public markets, people are like, “Oh, you’re going to get into a price war, margins are going to come down, and it’s a less attractive business.”

But that’s not really what I mean. What I mean by price is that when you think about the businesses we operate—primarily mobility and delivery—the vast majority of transactions in delivery, or the vast majority of transactions in transportation broadly, happen at a price point that is much lower than our core products.

Taking an UberX to and from work every day in New York City for $35 a direction is still a luxury product. The vast majority of transportation in New York City is not happening at that price point.

And so if we want to get to 500 million users, and we want to go from people using us on average 6 times a month to using us on average 25 times a month, the average cost of that transaction has to come down. There are all sorts of ways of doing that. You have more modes that are cheaper. You can get trains on Uber here in London. You have alternative modes like bikes and scooters, et cetera.

Because once you deconstruct car ownership, it's not just about UberX; it's about all the other things you do. But you have to get the price down.

À la poubelle. I fucking hate these bikes that litter the pavements. Oh, my God, they're so annoying.

You do.

Oh, my gosh. I'm an old man.

But you know Londoners love them generally. London is such a significant market for micromobility.

I know. It's why I don't get out much.

Would you rather have more cars on the road?

Would I?

Yeah.

To be fair, I live around the corner.

Yeah.

To be fair, I use Uber as an argument every single day because I don't have a driver's license.

Okay.

Because I have Uber.

I love that.

My girlfriend has a car that's, what, 20 grand, and then insurance is three or four. And I'm like, “Do you know how many Ubers I'd have to take to get to 24 grand?”

Totally. I mean, the individually owned car is the most inefficient asset that anyone owns, and certainly at any price point. It sits idle 98% of the day. It's depreciating, and the ongoing operating costs are actually high. Even if you're not driving it, you're paying for that insurance bill.

Which is why I do think, in some future world—maybe not 5 years, but 15 or 20 years—everyone's going to be like Harry. Nobody's going to own a car. Nobody's going to have their driver's license because you'll be able to get around.

I think bikes and scooters will be part of that. Autonomous vehicles will be part of that. I think public transportation will be a big part of that. But I don't think you need to own a car.

4. The Autonomy Imperative

And one way to bring down price is to remove cost, and one way to remove cost is to think about robotaxis and autonomy. You've said before it was existential. Why is it existential, and how do you think about that and how it changes the business forever?

I mean, it's existential because, at the end of the day, it's a better product than our core product in many use cases, and I think those use cases grow over time. Eventually, it's better in all use cases.

You can quibble along the edges on current autonomous vehicle experiences. In most cases, it is going to be slower than a human driver. The pickup point may not be right in front of your door, as it would be with a human driver. It's not going to work in all weather conditions, all geographies, or all pickup points. You can quibble on that today, but I think increasingly over time, autonomy is not only going to be safe; it's going to be safer.

I think it's going to be a better experience because people like the in-car experience—the privacy, and being able to work or sleep or talk with your partner, or whatever it is you want to do. That is better, and people prefer that for the most part.

So when you have a better product that's only going to get better over time—and autonomy is as bad as it's ever going to be today, right? Every single day, it's going to get better—then that's going to be the business, and that's going to be how people get around.

If Uber doesn't have autonomy on our platform—and we will; we are investing actively and aggressively to bring it to market—but if we didn't, then it certainly would be existential for our core business.

Is it the largest investment that you make?

It is. I think it depends a little bit on how you define it. If you look at our autonomy investments, we are making a mix of equity investments in companies, purchase commitments, building out autonomous infrastructure, and building out data-collection fleets. There are a lot of different ways we're spreading the dollars, and we're pretty confident in the long-term ROI on those dollars.

So yeah, it's the largest single standalone investment we make. Now, don't get me wrong: with a P&L our size, we're moving billions of dollars around every month. But yes, it's the largest single area of investment.

I'm a venture investor also.

Yeah.

Which means I love to pontificate, and I also love to say, “I told you so.” You and Travis kind of went down this road already. Do you look at that with annoyance—that you stop-started—and would you be materially ahead had you just been able to continue as planned?

First off, I'll say that when we started our autonomous efforts, this was a secret project within Uber. I was not involved in starting that. I don't want to take any credit for having that foresight, because I think it was foresight.

This was 2016?

No, well before 2016. I think Travis had—I don't want to get it wrong—but years earlier than that, he knew that this would be the future. Like many visionary founder types, he could see ahead of where the rest of us could see and started taking the company in that direction.

This would have been, I don't know, 2012, 2013, 2014, when we would have quietly started working on this, and I wasn't involved in it at all. At that time, the autonomy narrative was ahead of the reality by a lot.

Yeah.

Right? I mean, you can go back and read various prognostications—not just from Elon, but from many people in the industry—saying, “Next year, next year, next year,” and it was never next year.

But look, I think it would be rose-colored glasses to say, “Oh, see, if we just stayed in the game, we'd have the leading autonomous vehicle company, and this existential threat for us wouldn't exist, or we'd completely control our own destiny.”

When we ultimately divested ATG, which was our internal autonomy group, we were in the depths of COVID. Our mobility business had lost 84% of our top line in 3 weeks. The company was burning billions annually. We didn't have a core business producing cash. The billions were not coming from investments or other things. Our core was burning money.

We did not believe we were leading in autonomy at the time. We were trailing. You can debate about whether we were trailing the field or whether we were just trailing Waymo, but we were not in pole position.

Uber had a lot to prove: that we could lead, win, and make money in our core business. And so we divested ATG. We turned the core businesses into cash-flowing machines. We took the company public. We've grown the value and grown the business. Almost any metric you pick from that point in time is up and to the right.

On all those dimensions, I think the focused strategy played out. But yes, of course, today, if you could snap your fingers and say ATG would turn into one of the leading autonomous players globally and we completely controlled our destiny, I think that would be a good thing for us.

Can I ask you, when you fast-forward 5 years, what percentage of rides will be human-driven versus robotaxi-driven?

It's so hard to predict, for a few reasons. One is that our denominator is huge here, right? We're doing around 300 million trips a week on our core platforms. That is just massive scale. We do a few million trips in AVs on that platform today a month, but it's such a small part of the business that it's going to grow triple-digit percentages month on month, and month after month, and it will still be a relatively tiny drop in the overall bucket.

The second thing that makes it hard to predict is that our human-driven business is going to keep growing. Even in the largest AV markets where today we don't have AVs, like San Francisco and Los Angeles, our human-driven business is growing faster than the rest of the US. So it's hard for me to know what the target is. It's a moving target.

The third piece is that Uber is so global. In mobility, we operate across 75 countries. Two of our three largest countries by volume are India and Brazil. The average fare in Brazil is around $3.50 or $4 USD. In India, it's $2.50, $3, or something like that.

It's going to be decades until the cost of autonomy compresses to the point where it competes with the cost of human labor. Those markets make up the majority of our trips.

And so if you want to say, “When will the majority of trips at Uber be autonomous?” I can't tell you, because I can tell you it's probably not going to be until autonomy gets to Brazil and India, and I can tell you that's going to be a long time.

So what's really interesting there is that it could still be a very low volume of trips, but in terms of dollar amount, it could actually be significantly higher.

Dollar amount, yeah, because if it's in the US and if it's in the largest cities in the US, then that's where the rubber meets the road.

Exactly.

That's the counter to what I'm saying. I'm saying, “Oh, it's going to be more complex, and we have all this other...” But yes, of course, if autonomy starts to make up the majority of markets in San Francisco, Los Angeles, Washington, DC, Miami, New York, and Boston, that's a big chunk of our bookings.

That's a big chunk of our dollars, and so that's the ultimate question.

Who do you think is a bigger threat, Waymo or Tesla?

I don't know how many spicy takes I want to have here, but I think there's going to be more than 2 winners. Do I think Waymo and Tesla will ultimately be winners? Yes, I do. I don't know who's going to bet against either of those, but I think there will be more winners.

I also think even in a world of strong winners, a very natural question—or often feedback we get from investors or smart types who follow our business—is, "So, yeah, I believe that there will be a few players that get to autonomy, and I think they're ultimately going to work with you guys, but they're going to have such strong leverage in the market that your share of every dollar is going to get squeezed. And so I just don't know, even in a world where you have access to autonomy, how are your margins going to look? Because today you guys benefit from fragmentation."

And that's true. I think it's a true statement, but there are a couple of counterpoints to that. One is, if you look at delivery as a comparable vertical here, McDonald's and Starbucks are also strong leaders in their individual verticals. They've spent billions building out fixed assets in terms of stores and all the infrastructure that goes into their supply chain. They have 1P channels. You can walk in the front door of a McDonald's, or you can order through the McDonald's app.

But they also ultimately work with the marketplaces and were able to come to a good economic agreement that works for both sides. At the end of the day, they have expensive fixed assets, and you want to drive as high utilization as possible. Whether that's a store or a car, I think that's going to be true.

And so I think whether Waymo or Tesla ends up being the bigger threat, I don't know. I think ultimately it's in both of their interests to put their vehicles on our network, even if they have their own robotaxi apps or their own 1P apps. Even if they work with our competitors on the rideshare side or on the delivery side, I think everyone will work with us because ultimately we have distribution, and ultimately they have expensive fixed assets that need utilization.

What's interesting there is you say, then, that distribution is more important than superior technology.

I think in the end, distribution wins. And look, of course, if only one player gets to the finish line on the technology side, then that is a problem for us. But that is not the future that I think will exist. And even if you look at what's happened in China, there's not one AV company that is emerging as a winner there. There are already 4 or 5.

So I don't know why China would have 4 or 5—which, by the way, will over time become 8 or 10—and the rest of the world would converge around one player. I just don't see it emerging that way.

You were at Uber when you did Uber China, no?

I was. We exited our China business in 2016.

You've got kids, right?

I do.

Yeah.

I have 3 daughters.

Okay.

Yeah.

See, with kids, you tell them story time.

Yes.

Yeah.

Yeah.

Right. I pretend like it's story time. What's the wildest story from Uber China?

5. The Uber China Exit

I was only over China for a few months before we ultimately did the deal with DiDi. Even just those few months felt like I lived years, right? Just seeing the deal process play out, all the regular emotional highs and lows that come with the deal process, but also the specific China-specific, Travis-specific elements—it was just crazy. And ultimately, we got a successful outcome that I think folks would say most Western companies did not have. Even though we didn't win, even though we took the silver medal in China, I think we got a better outcome than the vast majority of Western companies and the vast majority of Western technology companies that try to do business in China.

Do you have a crazy story?

The crazy thing in China was that you'd be negotiating. To be clear, others were running the negotiation; we were running the business. But the mandate behind the scenes in the negotiation was, "We had to push on investment because it gave you leverage at the table." If one side saw the other was gaining share as you were negotiating the deal, it gave you relative strength.

This was happening day by day, and both sides were so well capitalized. Travis used to have a saying, which was, "We need to raise more money than all our competitors in the world combined," because the basis for competition in rideshare—product-market fit—was clear, so it was just a land grab at that point, and money helped you solve the land grab.

On the other hand, so did DiDi. And the notion that we were going to be able to raise more than everyone in the world combined was just never going to happen past a certain point, because you had players like SoftBank investing in the market as well. And you remember those days.

Yeah.

That was the free-money era, and Uber was the best in the world at capitalizing on the free-money era, but there were many others that were good at it as well and ran the same playbook as us.

So all that is to say, I remember the last few weeks in the negotiation. We were burning $52 million a week in China just on price subsidies because there was this heated behind-the-scenes battle happening to get to the best economics and the ultimate surrender or truce. So that was crazy.

Another story I heard, which I thought was nuts—and this was not an Uber story—was that before Uber and DiDi did our deal, we were the 2 largest players. But before that, there was a third player. I think it was called Kuaidi, and DiDi and Kuaidi merged. They did a deal, combining their HR systems, and realized that, of the 2,000 employees here and the 2,000 employees there, about 200 employees were on both payrolls.

So you had this dynamic where you started to realize, "Oh, okay, this is real, deep, competitive, gnarly. You have employees who are wearing both hats." That was crazy to me to hear because that notion, just in competing in the US, is not something that I could see happening in a million years.

Feels like a frontier AI lab employee.

Wild. And so there was all sorts of stuff like that. Remember, we were competing in China with one hand tied behind our back. Because of the nature of the investor bases in each company, at one point we were not able to operate on the WeChat platform.

Trying to compete in China and not having access to WeChat is like trying to compete in the US without email or a phone number. It's very difficult to run your business. But we did have our own local partners that were helpful.

Were you pleased to get out?

Look, you're never pleased to take the silver medal. I don't think it was plausible that we were ultimately going to be the market winner. Even for geopolitical reasons alone, the notion that a US tech company would ultimately be the largest mobility service in China—I just don't think it's something that was ever plausible.

And so it was always going to be about some exit to a local player, and I think, all things considered, we got a pretty good exit.

I don't think Xi Jinping is going to give you an employee-of-the-month award, is he? No.

No, it's hard, right?

Very harsh.

It's hard. And I think the hardest part of exiting was not the market opportunity—that was obvious—or the growth, because that was exciting in our business but also heavily subsidized. It was the Uber China team.

These were people who bet on Uber, who joined Uber, I'm sure, when many of their friends and families were like, "What are you doing? Don't join those guys." They were heart-and-soul Uber employees.

And I think one of the awesome things we did at the time was try to give as many of those folks who wanted it roles in the global machine. Many of those folks are still at Uber today. But that was hard.

Travis is actually a pretty loyal guy to people who are all in on the company, and our Uber China team was all in on the company, so that was a hard moment for us.

Pretty wild freaking time.

Wild.

Wild.

Wild.

It makes today with AI look less wild.

It's all relative, right? When you're in it, some of these things are just your reality, so you don't quite realize. But then you have the benefit of 10 years of hindsight, and you're like, "That was crazy."

6. The Elusive AI Payoff

Speaking of wild and crazy, and China letting US companies do well, China competing, you blew through a year's budget for AI in 4 months. Sorry. I'm just laughing because you go to this meeting and you're like, "So, how's the budget going?"

Well, first of all, it's not like that. It was a big spreadsheet reveal, and you're like, "Oh, it's gone." I think budgeting for new stuff is tough, right?

It's like me and my mother in Chanel.

I'm like, “Oh, the woman's gone.”

I'll stay away from that.

Yeah, yeah, terrifying. Is that evidence of incredibly effective tools, or is that evidence of a desperate need for guardrails?

I firmly believe multiple things can be true at once, so let's come to that. Let me give a little bit of backstory on this, because Uber had 2 big AI headlines in the first half of this year, both of which caught at least the people involved by surprise.

One was Praveen speaking at an event and generating this headline by saying we'd gone through our AI budget in the first few months of the year. Praveen's our CTO. Then I did another podcast and said it was hard to draw a direct line from our AI spend through to useful consumer features. Both of those comments caught fire in a way that I think neither of us expected.

Praveen wasn't making a comment about runaway spend, like we're going to bankrupt ourselves. He was just saying, “Hey, effectively, it's hard to predict usage. Usage has been more than I thought. We've been trying to drive usage, and here we are blowing through a budget.” But you're setting a budget number in November for a tool that's growing vertically in terms of usage. Of course, it's hard to pinpoint where you're going to be.

My comment, honestly, first of all, it wasn't insightful at all. It was held up as this insight, and it showed me the power of people reinforcing their preconceived notions—taking a statement that's fairly innocuous on its surface and using it to prove their point on one side or the other.

On the one side, AI skeptics were saying, “See? The Uber COO is saying there's no return on AI,” which is obviously not what I was saying. On the other side, if you were a fundamentalist AI evangelist, you were saying, “This guy has no idea what he's talking about. They're obviously doing it wrong because AI is God, and I don't touch my computer without engaging AI,” right? Obviously, there's nuance in the middle that's true.

The point around ROI, for me, is a couple of things. At the end of the day, we do want to get efficiency, or we want to get new and cool stuff built, and we are seeing examples of that every single day. We've stood up a pod of 30 of our best AI engineers who are partnered with businesspeople or folks in the G&A functions to go process by process and start from the ground up with AI: How do you improve that process?

If you can take a capital allocation process—every week we're allocating pricing dollars across thousands of markets globally—and take that from being a 15-hour process to a 2-hour process, which is what we've done, that is tremendous, tangible ROI. Now you get 2 days of someone's time back.

If you're able to take a forecasting process, where our finance team is constantly reforecasting every inch of our business, and turn that from 8 hours of work into 2 hours of work, you're able to do that not only with more precision, because you can add an additional layer of nuance into those forecasts, but you're also able to have your folks do other stuff. There's clear ROI there.

If you're able to take marketing QA from 2 weeks to 2 days, there are so many examples of that that we see. The way we've done that, again, is by pairing the business folks with the AI engineers.

But are you actually seeing that today? Alex Karp came on CNBC or CNN and said, “No, the ROI question is still there.” To validate what you said, to be clear, but outside of coding and customer support, with the greatest respect, I think anyone who runs a budget in a large enterprise today would say, “Yes, it's still not material at best.”

I think it's just hard to know. These things are hard to quantify, and so you do have to be a bit top-down and belief-based about it. I think the 3 examples I just gave—there are dozens more of those.

The natural question is, “Okay, great. How many of those people can I take out of my organization so that I get the cost back and that flows through to the bottom line, or I can put it into other things?” But formulaically doing that is really hard, because guess what? The 8 hours of value that was created, or the 8 hours of excess time, gets filled with some other activity, which is also presumably high value and maybe wouldn't have got done before, or wouldn't have been done to a similar level of precision.

So it's just very hard. I think the way companies ultimately have to extract AI efficiency, at least from a pure OpEx perspective, is in your target setting: hold the constraints tighter.

If we really believe that AI is making our employees 10%, 20%, or 30% more efficient, then next year we should just not increase head count. Or we should increase it by 2% instead of 10%, or decrease it by 5%, and say, “You all should be getting more done with less, and here are all these sub-examples of people doing that.”

But drawing the direct line between “I transformed this process” and “therefore, I need 2 fewer operations analysts” is really tough to do. So I do think there's ROI there, but being able to precisely quantify it is challenging.

How do you think about effective budgeting, then, having been through what you've been through with this—blowing through it in 4 months—with the difficulty of budgeting, and us both acknowledging that?

I think what you have to do is create combined pools of budgets and then let the people that you trust allocate where they see a high ROI. If you're talking about our CTO, I think it would be totally reasonable for Dara to say, “Your head count budget is X. Our compute budget is Y. Just add X and Y together and then spend it as you see fit.”

If you want to spend relatively more money on compute, on inference, or whatever, because you believe that's the highest ROI, do that, but it means you have less for heads. If you actually think it's more efficient to add more engineers because there's a compounding value to the new and novel products they will build, or it's not just about throughput, then do that. But if you combine the pools, I think that's an interesting approach.

The other thing I'd say is, remember, even at the beginning of this year, the idea that you would be doing things like smart routing internally in terms of which models you're using for which tasks—the idea that you would not only publish an AI usage leaderboard but also an associated cost leaderboard, just so people were aware; the idea that you might choose different models for different tasks from the outset, or give different levels of employees different models for different tasks from the outset—all these things were not really happening.

Do you work with providers like Fireworks to enable efficient routing?

Yes. We work with external providers.

Yeah.

We also do some of this internally. We've done things like build dashboarding so folks are aware. I mean, we have an internal—

Is that helpful, like usage and cost? I might be brilliant, but I'm number 1 on the cost, and I feel a bit guilty because I'm using an intense amount of compute. Is that good or is it bad?

At some point, I think it's wasteful. You do not need the latest and greatest model from Anthropic or OpenAI to ask, “Tell me who the president was in 1945,” and then run that again for the next 5 presidents and run it again for the next 5. You know what I mean?

But do the leaderboards help? I don't understand the point of them. Why would I create them?

I definitely think visibility helps for both the usage and the cost side of the equation. Imagine a counter in the top right of whatever tool I'm using that is just showing me the equivalent cost of what I'm doing. As that scales, it will make you more cognizant as a user. If you're at the grocery store and you—

But do I want to be number 1, or do I want to be at the bottom?

I think either extreme is probably wrong at this point, because it's a question of how much value you're creating.

Yeah.

And that's where human judgment still matters.

Well, then it's a bad leaderboard, man.

I do agree that sometimes tools can be so blunt as to become useless, because folks are optimizing for the metric versus the outcomes.

I do think, though, there is value in everyone in our organization using the latest and greatest tools in their specific domain. I don't need every person in the company using Cloud Code. Not every customer support rep needs to be doing that. But for the AI assistant agent that's helping them be a better customer support agent, I want every single agent using that tool. An adoption leaderboard for that is helpful, right? And if you're not, I want to ask the question as to why.

When you think about the size of companies in terms of people, will you have more or fewer people in 5 years?

It's interesting, because I am tempted to say I think we'll have fewer, and I think one of the reasons I'm tempted to say that is when you look at the largest teams from a number-of-people perspective, you do have disproportionate head count in more producing-type functions, whether it's customer support, sales, content production, or analytics, where you're producing reports and dashboards and these sorts of things.

And I think those sorts of functions lend themselves well to first augmentation by AI and eventually, I think, at least partial replacement by AI. So I'm tempted to say less. The reason I won't emphatically state that is because I think that's sort of been proven wrong the last few years, as AI has rolled out and employment in companies continues to grow.

Does it? I don't think it really does. If you look at your Shopifys and the generation you're in, actually it hasn't. Headcount stayed flat, and the companies have just become much more efficient.

Yeah. No, I think you could probably find examples to prove any point that you want to believe, right? I mean, I'm not an AI doomerist from an economy perspective. I think there are going to be productivity benefits, but I also think there are going to be whole new industries and fields stood up that we can't predict today, just like every other industrial revolution that's happened. But I can't tell you exactly what that's going to be.

So within companies, I think if you took everything Uber does today and held it static and said, "In 5 years, you're going to need more or less people," I'd say, "Well, we could do everything we do today with fewer people in 5 years because of the power of AI, but we're going to be doing a whole bunch of new, interesting stuff, and so maybe we need more employees to do that stuff."

You said something about the AI ROI question, and that was one thing that Alex Karp mentioned. The other thing that he mentioned in this very pertinent interview was that the biggest companies would be nervous to work with frontier labs. Do you agree with that as someone who runs the P&L for one of the biggest businesses in the world?

I mean, we work with the frontier labs. I think I watched some of the same interview.

He's brilliant to watch.

Yeah. I mean, amazing and insightful. One of the risks that I saw him highlight was this notion that you feed all of your data to the frontier labs, and then they effectively stand up a competing product. So you're sort of—

No, this is getting cannibalized.

Yeah, you know, the fox in the henhouse. You're sort of opening the gate. For us, our experience of working with the frontier labs has been great, and I think we have experimented on multiple fronts. We are moving from experimentation to implementation and scale in a bunch of areas where we're seeing ROI. So I haven't seen that yet, but I certainly get that argument, and I think there are companies that have fallen victim to that. He gave a bunch of examples.

I totally get it. I think you put Uber in the less penetrable category for frontier labs. I would really argue—

Look, I mean, we have this physical-world component to our business that makes it challenging to do that. I don't know that I see OpenAI launching a ridesharing service any time soon, going around to tens of thousands of cities around the world, getting locally licensed, and then putting boots on the ground to run a physical-world service. I think many of the places we play just don't lend themselves well to that extensibility of their model.

See, I disagree. I've always known that Dario, in particular, was very passionate about last-mile delivery in Barcelona for convenience food.

Yes.

One of his big things.

Well, then I'll have to be on the watch out for that. Totally.

AGI and refreshments in Barcelona.

No, look, the physical-world aspect of our business is hard, but it also—

Oh.

7. The Agentic Uber Interface

It also means that I think some of the worst prognostications haven't come true, right? I mean, even at the transaction level, the big conversations we were having in our leadership team 18 months ago were, "What's going to happen to the consumer front end?" The disaggregation risk on both delivery and mobility is that people want to start their Uber ride with a plain-language query.

Where do you land on that line? I was speaking about this with really smart people, and they talk about agents and how agents will route you to a provider, and you have no customer loyalty. How do you think about the disaggregation of the UI and agent-led decision-making?

So this is where I think there are some interesting questions about what information you provide to the various consumer front ends, either from the frontier labs or others. Because at some point, maybe you're giving away that front end of the consumer experience in a way that is nonstrategic.

So I don't want to be aggregated on price. We've not participated historically in the aggregation apps where somebody will come to us and pitch and say, "Hey, give us APIs that give us real-time information on every car, every Uber car in the network, what the price of that ride is, whatever other characteristics you can feed us, because we're going to build an app, and then we're also going to put Lyft in there, and we're going to put other providers in there, and then that'll be incremental business for you."

I've been against that. I want to be the front end. I want people to start at the Uber app for the Uber experience. I think today we win that first look with 200 million consumers and growing every month, so there's a real question there.

We've been worried and discussing, "Hey, is somebody just going to put into ChatGPT or Claude, 'Get me my usual Uber'?" I could imagine a world where the query starts there. The challenge is that we have a very managed transaction, right?

Does that damage your business, though? "Get me an Uber" is still your transaction.

It doesn't damage the business, and I would of course fulfill that query, right? There's a question as to whether comparing the prices of Uber, Lyft, and Waymo and getting me the cheapest one damages my business. Well, no, if we're winning on the cheapest price every time.

But if today 80 percent of people just start with Uber, do I want them to migrate over to a service where they go to a comparison app, whether that's a—

Or they just say, "Get me a car," and then—

Get me a car, and then they're indifferent.

There's a time, a way, and a price, and then just—

Yeah. I think the challenge, though, is that it's a managed transaction, both on the delivery side and on the transportation side. All the little things that happen between saying, "Get me an Uber," and you being done with that ride—the interaction between the driver and the rider, the visual experience of the pickup, the fact that I left something in the car, your payment credentials—all these pieces that you sort of take for granted need to be figured out, right?

This is not an e-commerce transaction where you set it and forget it. You don't think about it until the package is on your doorstep. It is a managed transaction, and so that worst fear hasn't played out yet. I'm not saying it won't.

I mean, Brian Chesky got kind of roasted, but I thought it was an insightful point when he said it's not clear to him that the right interface for hotel booking is a chat interface. He was called a Luddite and this and that, but I think he was right. Some experiences are more visual, and some experiences are more managed.

Well, I think it depends if it's transactional.

It depends.

"Hey, get me a hotel for my trip to London to see Harry." You probably don't care about it having a sea view and being romantic for you and your wife, but you want it close to the office, efficient from a pricing perspective, and compliant with your HR.

Yeah. Yeah.

That denigrates the market.

Yep.

I host a show that is very popular, actually, with 2 other investors who are much more insightful than me, and they talk about the Fortnitification of markets, which is just the shrinking of markets.

Yeah. Yeah.

And, don't get me wrong, Airbnb is an amazing business.

Yeah. No, I think it is.

But if you remove transactional booking travel, then it just becomes experiential booking travel.

Yeah.

It's more—

No, I think it's a reasonable perspective.

What do you do, then, to get ahead of it?

I want to be where the consumers are. Ultimately, we've chosen to participate.

Participate with companies like OpenAI?

Yeah. But really, with any of the large companies, if they want to do something interesting with us on the consumer front end, we'll have that conversation.

Can you participate, though, if you won't give them the data?

Well, I think that's always a negotiation or discussion around how much they need and where the transaction starts. There are 15 different flavors to this, right? The transaction can originate in different channels and then end in the Uber app. You have to define who has responsibilities for things along the way.

What is the consumer experience? If you go into ChatGPT and say—or even take this hotel-booking example—what happens if the hotel needs to send Harry a message because you asked for early check-in and they can't give it to you? Does that go back through the AI, or is that coming directly from the hotel? Who bears the cost of that?

These are operational elements to the experience that need to be sorted.

And I'm not saying this can't get sorted, but it's not as simple as that. You shouldn't picture the experience that goes right as the archetype of what this usually looks like, because it's the experience that goes wrong or requires some level of management that needs to be solved for.

One of my very dear friends is CEO of one of the largest airlines in the world. He says, “You have no fucking idea how hard my business is. If your baggage is 12 minutes late, I will have 50 fucking emails, and I do 5,000 flights every single day.”

Yeah, totally. And I'm not naive or saying, “No, no, our business is different. It's hard.” Every business is hard. But I do think these things need to be sorted out.

8. Buying Delivery Hero

Can I ask you one? You bought Delivery Hero. I know Niklas really well. I've interviewed him and really like him. Brilliant guy. I know Oscar from Glovo. I really like him. Why buy it, not just dominate? Is it not just a market maturation question, and you will slowly crush it over time?

So first of all, I'll say Uber has been on this journey, right? I often get asked the question of what business is going to be bigger long term, or where there is a larger TAM—

Food or mobility?

Food or mobility.

Yeah.

Yeah, because even the existential questions we get about AV tend to ignore the fact that we have basically an equally sized food delivery business that's in market-leading positions in most of our markets around the world. We get almost zero credit for that. But putting that aside, delivery has also been growing faster, right? So it's almost as big as mobility, has been growing faster, and has been more constrained from a country's perspective.

We actually did some rationalization of our country portfolio. We didn't launch as many of the frontier markets or emerging markets. We were more capital-constrained when we were scaling delivery. And so Delivery Hero presented a unique opportunity to, in one fell swoop, expand our geographic footprint.

It's not that we could never go launch and scale new markets. We have been launching new countries in delivery, but it takes a lot of time. It just takes time, and it's back to what I was discussing earlier: What's relevant scale? How quickly does it take for a new business line or a new country to get to a relevant scale that matters for Uber?

The other thing is that Delivery Hero has built a lot of local brands that are really strong, right?

And they have exclusivity and lock-in.

Whether that's true or not, they have consumer mindshare, right? They've built leading brands in Argentina, Korea, and the Middle East. These are brands that consumers identify with, have high household awareness, and are not easily supplemented. And so I think there's value in those brands.

They've also localized their services really well. I think the combined mobility and delivery offerings we'll now be able to offer in those markets will be really compelling for consumers. So it's scale, some of the local brands that they've built, and the platform. For us, when this deal goes through—we have to go through the requisite regulatory and shareholder processes—delivery will be a much bigger business for mobility, and that's an exciting version of Uber for sure.

Are you more passionate about one than the other? I know it sounds weird.

Which of your kids do you love the most?

The oldest one.

Actually, my baby right now. She's the most daddy's girl of our 3, so she's got a special-

How old is she?

She's 18 months.

Okay.

So—

She's doing 1 sleep?

Yes, 1 nap a day right now.

Like a venture investor. About 2:00 p.m.—call it siesta. Welcome to Europe, Mike.

Hopefully less grumpy. Look, I grew up in the mobility business. From 2012 to 2025, I spent 90% of my waking hours and most of my sleeping hours thinking about mobility and rideshare primarily, but also all the other mobility verticals we've built.

As I said, I don't think there's anyone in the world who's spent more hours thinking about rideshare. With delivery, I've managed teams over the years that have serviced the delivery business. It came into my portfolio, quote-unquote, 14 months ago. For the last couple of months, I've been directly running the delivery business.

Our leader of the delivery business left, so I took her role, and I've been doing 2 jobs: my day job and my night job. I literally had to schedule an evening shift because there's just no way to fit my operating cadence in.

Are you just a machine? You are an efficient executor. Even the way that you present, it's efficient.

Well, look, everyone is struggling to find enough hours in the day. At Uber right now, the teams are pushing hard, and I'm worried that some of our teams are going to run out of gas. You can only push above the red line for so long because we have a lot of opportunity, but also a lot of challenges.

We're best in a crisis. We're best with a challenge in front of us. We're best when we feel like we're up against the world a little bit. That's our DNA, and I'm inspired by that, but it's hard right now. Personally, as I said, I'm working 2 jobs.

But back to your question about which is your favorite business: I'm working in the delivery business and directly pulling the levers myself for the first time ever in my tenure at Uber, and I'm really enjoying it. It's a very complex business—a 3-sided marketplace versus 2-sided. I think there's much more complexity in terms of what the consumer actually values.

The inputs that matter on the mobility side are speed, price, reliability, and safety. It's a longer list on the delivery side of things that you have to nail. And so it's interesting. It's hard. We are not number 1 in the US, which also makes it harder, because I think operating from a position of strength just gives you a nice tailwind.

We're having to play the challenger role, which we relish, but it also changes the game a bit.

Some of my friends who are old Uberites, who I'm sure you know, but I'll keep them out, say, “If Travis were here, we'd be number 1 in food.” Is that true?

Look, I mean—

I love Dara. I love Dara. I don't know Travis, so—

No, I—

I don't know.

No, and look, the reality is that I think anyone who operates anything that says, “If X were this, this would be different”—it's a little bit of that “In the Arena” quote. If you're not in the arena, it's easy to sit on the sidelines and say, “If I were in the arena, or if so-and-so were in the arena, it would be different.”

It's fair to have that opinion, but when you're operating a business, it's hard. You have trade-offs to make, you get things wrong, and you get things right. You don't get credit for the things you get right. You certainly feel and take the blame for the things you get wrong.

So I don't know that there's an alternative history. DoorDash is an excellent company. I think Tony is a tremendous entrepreneur and founder. They operate really well, move quickly, are aggressive, take risks, and are well-capitalized. We have lots of competitors.

Did you ever have the chance to buy them?

You hear things. As I said, I was on the mobility side of the business. I don't know if that was ever a realistic possibility or not. Of course, there's always speculation, but I don't know.

Do you know one of the best answers I was ever advised on? My dear friend Shakeel Khan, who is Daniel Ek's right-hand man, says, “If ever you get a question you don't want, just go, ‘Hey, mate, that's above my pay grade.’” I'm just a podcaster, me.

That's a good way to handle it.

Yeah, it's just like, “Me? Nah.”

No, I honestly don't know the answer to that question.

Was Postmates a good acquisition? Because that seemed like a bit of a nuts one, to be honest. They were running out of cash. It was a challenged business. When I saw that, I was like, “Whoa, they've got balls at Uber.”

I think we probably get a harder rap on M&A than is deserved because, in many cases, a deal that from the outside you question or aren't sure what you got out of it actually leaps forward internal capabilities that you didn't know you needed. You learn things from the acquisition, get good talent, see where you have gaps, and so on.

I think in the case of Postmates—and again, this is where I'm a sideline observer—I could sit here and tell you that was the greatest deal in the world, or that we shouldn't have done it. But the reality is, I wasn't in the game at that time on the delivery side, so I don't actually know the answer to that.

Postmates has a strong brand, a strong following, and some strong geographic pockets. I think we've been able to build on those.

Are you ready for a quick-fire?

Sure. Let's do it.

Otherwise, I'm going to get in trouble with Nerf for taking too much of your time.

…and you actually have to be productive in London. What have you changed your mind on most significantly in the last 12 months?

9. Lessons From Two Uber CEOs

I mean, I’m a humanity bull, and I’m really becoming more interested and obsessed with longevity. I actually do think we’re going to solve all human disease at some point, and the idea of—I don’t know if we’ll live forever, but living a very long time—is going to be possible. I’ve gotten more conviction there.

I saw the Whoop. OpenAI or Anthropic?

OpenAI. For me, I use both. As we were chatting about earlier, I use voice so much. It’s my single most-used AI feature by a mile—probably 50 times more than anything else. I record notes, emails, thoughts, and lists. I’m constantly working via voice, and I find OpenAI’s voice engine incredible.

I write my investor updates by voice. Yeah, I totally agree with you. Okay, what’s the one thing that you most took from working with Travis? What’s the single biggest lesson?

I want to give you 2. He’s a problem solver. He will define what he is and what he looks for in others as creative problem-solving.

The ability for him to walk into any meeting on any topic, ask a few pointed questions, float a few ideas, and in 15 minutes change the minds or evolve the thinking of the people in the room who have spent weeks as experts on the topic is amazing. To then go through every day, every week, half an hour, half an hour, half an hour into the evenings, and just do that muscle over and over and over again is so valuable.

I think if, as leaders, we can play that role and do a microcosm of that—maybe not that well—you can move the ball forward a lot. Creative problem-solving as a skill that is valued in an organization is probably the top thing I took.

I think the second is the all-hands that he would host, where he would not only give an answer to a question, but explain his thinking on why that was the answer. I think that’s exceptionally valuable in leaders: to take people through why what you say is the answer. It helps them. It creates mini versions of yourself.

If you can do that across your organization, where you tell people how you got to an answer, you’re amplifying the power of the organization. The way I try to do that is by setting down principles for how I want to think about a given problem, a given solution area, or whatever, and then having my people try to use those principles as they think about the problem themselves.

What’s your biggest takeaway from working with Dara, on the flip side?

I think the most impactful quote I’ve heard from Dara that speaks to who he is, is, “Management comes from an org chart. Leadership comes from the heart.”

What he means by that is we can create rules and structures and hierarchy, and we can try to follow what the bureaucracy says, but at the end of the day, we have leaders at all levels of the company who are the ones who actually push the company forward. Those are the people who are leading with both the head and the heart. Those are the people who build followership, and that’s exactly who Dara is.

He will not ask you to do anything he wouldn’t do himself. He’s the first one over the fence. He’s the first one on the plane to go where the company needs him. Low ego, lots of heart. He pushes, but it generally comes from a good place, which makes people want to be successful for him and makes people want to say, “What do you need me to do?” That’s really powerful.

You worked with both. We both know the politics that were around. Very few people were able to work with both, and they were like, “I’m a Travis,” or, “I’m from the Dara era.” What made you able to be an OG with both?

For me, first of all, I think they’re both excellent—

Yeah.

—in their own domains, and I think Dara was exactly the right leader for Uber when he came in and continues to be exactly the right leader for the company today. They’re different, but it’s not any easier.

For me, it comes back to where we started, almost your first question in the interview, which is, when times have been hard at Uber, I’ve not wanted to leave because I felt like it was the wrong thing for Uber. I’d be leaving my teammates behind, and it just didn’t feel like the right thing to do.

When times are good at Uber, I want to stay because this is fun. We’re building. We’re conquering the world. It’s been hard for me through the worst times and the best times to ever think about leaving, and that’s true regardless of who the CEO has been.

I also firmly believe that people need to take what they can get from their leaders, from their managers, from their boss, and you’re not going to get everything from any one individual. I’ve been able to learn a lot from both of them, and I think that’s been really great.

Final one for you. What’s the best piece of advice you’ve ever been given?

We hired a woman, Rachel Whetstone, to run our communications and policy team, I want to say around 2015 or 2016. She sent the speech of a commencement address she gave to the whole company in her first week. The central thesis was, “Always say yes. Just jump at the next adventure.”

It really resonated with me because you can always analyze a career opportunity. Should I tackle this problem? I’m being asked to do X. I’m not sure if I’m going to be good at it. It seems like there’s a lot of risk. I always just tell people, “Just say yes.”

A, bet on yourself. You’re going to get in there. It’s going to be hard. You’re going to figure it out. You’re going to be better off for it, and the company will be better off. Or maybe it will be too much for you, but you’ll learn a lot from that failure, and you’ll just be a better version of yourself. I think just say yes.

Honestly, dude, I’ve really so enjoyed this. I do lots of shows, and episodes like this remind me why I love what I do so much. Thank you so much for doing it—

I appreciate that.

—and for being so brilliant. Honestly, amazing.

Thank you. Awesome. So great to be here.