a16z Podcast - From Copilots to Agents: Rebuilding the Company Around AI

Episode Date: February 18, 2026

a16z's Angela Strange and Gabriel Vasquez speak with Carlos García Ottati, founder and CEO of Kavak, about building Latin America's largest online used car marketplace across Mexico, Brazil, Chile, Argentina, and the Middle East. They discuss why building in emerging markets means constructing four businesses underneath your business, how Kavak replaced copilot tools with AI agents handling 90 to 95% of customer interactions, and what it took to go flat for a year during the transition before growing four times on the other side.   Stay Updated:Find a16z on YouTube: YouTubeFind a16z on XFind a16z on LinkedInListen to the a16z Show on SpotifyListen to the a16z Show on Apple PodcastsFollow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

When you're building AI, the first thing that you need to build is the brakes of the system. It's the understanding on where you're going to deploy it and how you're going to deploy it. We had thousands and millions of information lines about our users. And the first thing that we did was we just built the system that allowed us to understand everything about every user. We built these co-pilot tools, and we realized very quickly they didn't adopt them. So we went funnel by funnel, putting agents in front of critical aspects of the business. We put the agents in front of our users really quickly. It was very painful.

If you see the best companies out there, 97% of their value gets created after year 15. Year 1 5, 97% of their value. 40% of used car transactions in Latin America ended fraud. Not the lemon problem you find in the United States. Kidnappings, forged documents, stolen vehicles. And in Mexico, only 5% of the used car market is financed, compared to 90% of the used car market is financed, compared to 90% in the U.S. Carlos Garcia-Otati founded Kavak after being defrauded twice trying to buy and sell cars in the region.

The company he built handles everything, buying, reconditioning, selling, financing, warranties, and logistics. Four separate businesses underneath one consumer experience, operating across Latin America and the Middle East. By 2021, Kavok had 10,000 employees and was growing at 300%. Then capital disappeared, and the company had to rethink everything. Instead of retreating, Carlos bet on AI, replacing copilot tools that employees wouldn't adopt with agents that now handle more than 90% of customer interactions. The transition took a full year of flat growth. A16Z's Angela Strange and Gabriel Baskas speak with Carlos Garcia-O-Tati, founder and CEO of Kavak. Carlos, for those who aren't familiar with Kavak, can you introduce the company and tell us a little bit about the scale of where you're at?

Yeah, sure. Well, first of all, thanks for having here. It's amazing. Kabake is an all-in-one marketplace to do anything related to your car. So we work out of Latin America. We work in Mexico, Brazil, Chile, Argentina. We also have operations in the Middle East. And what we essentially do is we manage the lifetime of our users from the moment that they want to sell their car to the moment that they want to buy a car. If their car breaks down, we're there as well. So we manage the end-to-end process of a user's journey while they own their car. And we do everything.

We buy these cars. We re-conditioned them. We then sell them online. We provide the financing to our users as well. We manage their warranties, their services throughout the process. We pay their tickets for them if they want us to do that. And we just make sure that we're just building that relationship with our user as they go along. In terms of scale, well, the business is probably the biggest e-commerce auto business in Latin America.

We do around 10,000 transactions every month growing very fast. We have around 3,500 people. Yeah. Fantastic. I wish I had a service that would also pay my parking tickets. Yeah. Well, Carlos, when we first met, like, there was this particular intensity that you brought to the table immediately. And I think we'll see that across the interview.

But I think there's something unique about your background, having been born in Caracas, Venezuela, moving a lot. I think you did 14 schools. Is that right? That's right, yeah. And we're just curious how that early upbringing kind of shape, Carlos now. Wow, it's a good question. Let me try to unpack that, right? There's like a lot of the learnings and a lot of what I am today come from those years and from my family. So yeah, you know, I lived in, up until I was like 17, 18. I've lived in Venezuela and the U.S. moved a lot.

And I think three things sort of like came out of that that still stay with me. The first thing is getting used to change, right? And changing 14 different schools meant that sometimes in the same city, I would go to a different school. You're always moving. My, my parents are, were always trying to find the best things for me. So the first thing that I learned, I think, about change was how can life be at a certain point in time really great? And then all of a sudden, it could stop being great. Because you were in a school, you had great friends,

like you were enjoying everything, and that you had to just move and start all over again. And I started to see sort of like that pattern of everything could be great one day and then all of a sudden not great. And the same way and the other way, right? You get to a certain place where they're not having fun, it's not good, but all of a sudden you move. and you get an opportunity to re-embet yourself.

And this is something that I've been unpacking over the years, right? Because when you're building a startup, when you're building a business, it's exactly like that. You sometimes are on the top of the world, but the next day you're in the bottom of the world and vice versa. Learning that and having gone through that so many times, changed the way I see things. And I think that shaped my ability to build companies

and build them with all these changes that happen as you do it. I think that's one. The second thing, when you're changing, you always have to have something that doesn't change. And that was for me, my family. And the inception moving, you know that he had his totem to understand whether he was in reality or not or in a dream. For me, that totem was my family. It was my mom, my dad, and my sister. Throughout all of those moves, up until the time I was 17, that remained constant,

and that allowed me to value what was important. And my dad was in the military, so we had everything that we needed growing up as kids, the basic things. My dad and my mom just gave us so much love and they put ourselves, me and my sister first on everything that they did. But I think the biggest thing that sort of like shaped me back then was just watching them. Like every time we changed from a place to another place, they were excited about it. They were very optimistic about it. And they were always excited about what's next. They were always dreaming.

We were always talking about all of these dreams of the things that we wanted to do and how we want to spend our time. together. Everything that we did, my parents sort of like took it upon them to just make it great and fun. And that continued, right? You know, in 2002, my dad was an admiral in the Navy and didn't want to be a part of what was happening in the country. And he was against the government and he decided just stayed his position there and that it got him prosecuted. It took his freedom away. He lost his job and he eventually had to move from Venezuela. And I saw him as well and my mom do the right thing every single time. And if it wasn't convenient for them and I stayed in Venezuela. They had to leave the country. And the next chapter of my life came without that constant on a physical constant.

We still had that fortune that we could see each other every year. But I saw that, right? I saw them just do the right thing no matter what and be optimistic about the world that they wanted to live in no matter what. And the third thing about those years is the pattern recognition that you develop. You know, like I have the advantage and the privilege that I went to public schools in Venezuela and to Oxford. So I lived in, and I had a life where I've seen different socioeconomic ecosystems and that sort of like provides a level of empathy that few have.

But on the other hand, you develop, when you're changing so much, the only thing that's constant is human beings is people. You start to learn how to read people and you start to see like the common threats around your world and I've developed sort of like

it probably would have been a defense mechanism but I developed a formula on how to make sure to create win-win scenarios with everybody around me, right? Like how do go that formula of building relations where one plus one equals four

and not two? And as I got older, I became really good at making that formula work. So I started to take more and more control of the narrative than I thought I had. Carlos, so how did you get into entrepreneurship? When was your first business? What were you thinking and what got you excited? Yeah.

Well, this zero, like, no technology into the mix, right? But it was actually at 12. My dad had been sent to Newport Rhode Island. And I remember in the U.S. when, like, if you wanted to work at McDonald's or Walmart packing back, you need to be 14 years old. I wanted to have some money to buy some Michael Jordan cards. This is 1996, so you have to remember the whole Chicago Bulls phenomenon, and you're going through that. And I remember that the house that we moved into had a lawnmower.

And I just went and just knock on a door, ask the person that owned the house that I can mow their lawn. And she said, yes. And I started to do that. And at a certain point in time, I didn't have have enough time to mow all the lawns that I was getting and I was getting paid like 20 bucks to do this. So I went back to my house and asked for my dad to dad, I need you to loan me some money so I can buy a few more lawnmowers and I needed to recruit a team to do this. So I went back to the basketball court. I convinced a bunch of kids to mow some lawns. And for me, the business there wasn't mowing the lawn was just knocking on doors and asking people to mow their lawns and just getting a big enough demand and figuring out how to fulfill that demand.

And I think that was very early learning. But what really hit me was that most of these kids didn't show up at the time that they needed to mow that lawn. So I was always under-delivering to my users, to my customers. And I ended up, like, changing the whole system to becoming sort of like a subscription system where I would tell the owner that their lawn would get covered every month for a $20 buck fee and that I will just get there and mow it and every month it will be in the right sort of condition and that allowed me to think about it very differently and it ended up becoming

something really big for me at that time so you know it started there I went back to Venezuela and like at 14 and I had like $20,000 in my pocket and that sort of like it was crazy you know 14 year olds it's a lot of money right in Venezuela I don't know it was 14, like I don't know, the exact amount, but it was a lot, like, it was a lot of money. That, I just seated into the next business and so on and so forth. Some of that capital actually came into Kabak, right? It just, I never stopped building after that. I just, I learned that, you know, you need to put that money to work.

And that was the same year that I also, like, this was 1990, 6, I opened up my first email. It was magic, you know, like I went to my father's office. I'm moving a lot. So I'm writing letters to my grandmother, to my cousins, and all of a sudden, you know, you go from letters to email. That was a magical moment for me as well. That also, like, seeded another obsession that was intern at it. And I think I was really early for a Venezuelan guy

into seeing how magic worked. And the combination of those two things, of, you know, operational intense, understanding what your product actually is, which is not mowing lawns. It was just making sure. that people were satisfied with how their garden looked and having the workforce incentives

in a place where I would go, it led it continued to compound and to work out into what we're doing today. Thank you for that background. You could have done many things with this background. What was the moment that you decided, ah, I'm going to start Kabak? Yeah, well, it all built up to sort of like that moment.

You know, I left Venezuela, first went to Italy, ended up doing an NBA in Oxford, worked at Amazon for a while in the UK, and I think that this is the first time that I see a digital business at scale and how it's impacting the users. And I returned to Latin America, obsessed with Amazon.

I went to work for McKinsey for a while because I needed to pay for my MBA, but I ended up leaving to join Lino and build the marketplace side of Lino, which was, you know, it's a business very similar to what Amazon had, where you had retail and your marketplace business and we launched it out of eight countries in Latin America

and while I was, you know, and we can unpack that and talk about that and that journey. But while I was doing that, I was moving from Colombia to Mexico and I needed to sell a car. When I go out and try to sell my car, there was no place where I can just go and sell my car. It was very informal market.

So I had to, you know, join a classified, put my car online and then meet up with a bunch of people to get that transaction done. I ended up settling something with a person there. We sold the car to that person

and I got defrauded in that process. It was a nightmare and I lost a car. And I moved to Mexico to launch Linio and that sort of like it was in the back of my mind.

You know, like I just been through this terrible process. And I started to work on a solution to sell my car in a different way. So when you think about the market in Latam, you have, it's very informal. 90% of the market is informal. It doesn't happen in dealers. It doesn't happen in a formal space. All transactions happen between peers. People like you and me that don't know what they're doing and have no information.

The problem with this informality is that 40% of every single transaction, ends up in fraud of one way or another. And this is not the lemon issue that you find in the U.S. This is literally getting kidnapped, killed, being defrauded the way that I was, buying a car that has forged documents that is stolen. So you have all this fraud that exists in the system that make it very unsafe for, like, normal users to go out

and just transact among each other. And I was part of that trying to sell my car. But the interesting part was that when I landed in Mexico, and after a year, I was probably the worst type of citizen than Mexico. You know, a Venezuelan, an entrepreneur, so no bank wanted to give me access. I already was going to underwrite you. Yeah, nobody was going to underwrite me. But I still needed a car.

Like in Latin America, you need cars. So when I landed and after a year I bought a car, I went into one of these classified, got a car that I liked, it was a used car, bought it, and after two months, the police stopped me and they were to throw me into prison because all the documents were forged. So the whole process of discovering Kabak was

just by me, you know, being probably the most naive person in Latam. And that's sort of like the thing that I needed to figure out at that point, you know, whether I was an idiot and I got defrauded twice or there was a bigger problem. And that's when I learned sort of like those stats

that level of informality, that level of fraud. but the most interesting piece about it was, yeah, imagine a world where you can solve that fraud, where you can, you know, figure out how to get the information and make sure that, you know, people don't have to go through this fraud. And imagine a world where you can also back the lemon issue, you know, if a car breaks down. And one thing that's true about use cars that I've learned after the years is that they're going to break down.

Like there's no way around here, right? Like, it's like humans. At a certain point, we're going to break down. The real question is, how do you show up for your, users when that happens, right? I don't make these cars. I just provide the proper warranties for somebody when they buy them. I can be there when it fails. But imagine in that world you're able to also solve that. The biggest thing that we saw was that nobody was financing this asset. So to put it in perspective, in Mexico particularly, only 5% of the market is finance of the

U-S. market. That, when you compare to the U.S., the U.S. is closer to 90%. So this means that in the U.S., out of 10 people, seven have a car. In Latin America, out of 10 people around two have a car. And that was the real sort of like opportunity that I saw that it could relate to us well. When you live in Latin America and you use public transportation, you take two hours to go into work, two hours to get back, you're using this really unsafe public transportation system, that the moment that you get yourself in a car,

you change the life of your family forever. And I went through that myself. I bought my first car in my 20s. And the biggest, you know, I've built Kabak, I've built all of these companies. And the biggest change that I saw in my life was when I got myself into that first car. My business started to do better. My social life started to do better. And in Latam, when you're thinking about these economies that are so polarized and, you know, there's no middle class.

There's just people that are struggling, people that are doing really well and very few people in the middle class. When you think about what's that one thing that gets you into the middle class, or that separates you, it's a car. Yeah. And so you've described, or I've heard you described recently, Cavac, as Spotify, meets Amazon,

meets Toyota, meets Citibank. Was that always the original vision of Kabak? Or how did that evolve? Maybe unpacked that analogy a little bit. When I saw this opportunity, the opportunity was not to just go and try to solve the buy and sale process of it,

solve the sort of like the lemon or the warranty coverage part of it. It was very importantly find ways to get people into cars. And to get people into cars, you needed to make sure that the transaction was safe, that you could provide the proper coverages, and that you can get financing for your users

to be able to buy this car. My whole goal became, how can I get people into the middle class through the car? And having built Linium, I learned, this is a business that started in that time in 2012, right? This was very back in the day where...

It was the first... Marker, place. Yeah. So when you're building, when we went out to build that, we were very young. We were in our 20s. I was one of the oldest one or the oldest one. And we went out to build Amazon. And what we realized very early on, that if you're going to build a business in Latin America,

you need to build like 10 different businesses to make your business work. And I'll give you some examples, right? We needed to buy these products and sell them online. But to do that, you needed to have payment rails. and you didn't have stripe back then so we had to build our own sort of like payment systems to be able to transact and you had like 70%,

you know, rejection rate and you needed to take that to where it is today, right? Half of the people didn't have credit cards so they wanted to pay cash. So we needed to build cash on delivery options which meant getting to people's homes with the product

getting the cash and getting that cash back into the system which was very tough. you didn't have trucks. You didn't have DHL FedEx in this market. So we needed to buy our trucks, build a software, build the systems to be able to do this. We basically didn't have anything, right? So we spend a lot of our time building the basic infrastructure

to do that last mile thing that's selling online. And when you're in the U.S., you're not facing that issue. You're just trying to make one process that doesn't work 15 times better. I learned through that experience that you need to build a whole, infrastructure behind it. And when I was thinking about Kabaka, I was a little bit older. I have gone through that experience,

and I knew that in order for me to solve this fundamental issue that is how can I get more people in my lifetime into the middle class and sort of like be a part of their life through their car, the biggest question that I had back in 2016 when I started the business was, why do I need to build? What are those pillars that I need to build

to make this business work, right? And I didn't want to do what we did in our previous company that was, you know, figured it out as we go. I needed to strategize really well because this is the time where, you know, the Uber without cars, slide is out there, Airbnb without hotels. And I'm telling the ecosystem, hey, I'm going to have infrastructure. It wasn't popular back then. And still it isn't really popular.

But in Latin America, it was needed. So for us to sort of like solve that issue, we needed to build. four different businesses below our business to make our business work. The first thing is we needed to build that e-commerce platform so people can come, sell their car, and buy a car.

We needed to build their operational or factory re-conditioning warranty machine where, you know, we could come at scale and really fix these cars or recondition them, get them into hand of another user

and when it fails, make sure that we were there. And this was a big undertaking. You know, use cars is usually anywhere between three to seven times bigger than new car sales. So if you're successful, your factories are going to be three to seven times bigger than all the OEMs combined. And you're dealing with edge cases, right? You're not dealing with a linear manufacturing process. So we needed to figure out how to build, you know, a machinery that could do that at scale for hundreds of thousands of cars.

You know, the regular mom-em shop, you know, does 20 cars in a given month, right? We needed to build the financing engine. How are we going to underwrite these clients? How are we going to get the capital to finance them? How are we going to collect? How are we going to do this all online? So we need to build a financing business. And we needed to build a logistic business.

When you think about cars is that they move. All of them, very, very easy things to do. Yeah, well, back, I was very naive, right? like when we started this. Like we... Now we call that defensibility from when you're starting, yes. But we knew that,

we needed to do that to really provide that amazing solution for the end user and we knew that our life was going to be living hell for the first few years because nobody was going to appreciate the infrastructure piece of it.

And this is a great segue because, you know, obviously Kabak is a very complex business. If you think about one country but you being very intentional since the beginning of saying that Kabag is going to be a global company and you know there's very few examples of companies in Latin America that have expanded successfully across the region but also abroad like you know you guys have operations in the Middle East as well so

like talk to us a little bit about that like what is that you know what does that mean to build Kabakh as a global company and how do you translate this you know kind of complexities that you find in one market across the different markets as well because you can imagine that that requires a lot of effort. Yeah, well, the whole logic was the complete opposite, right? Like on how we think about sort of like expanding, how we think about markets. So first, we're not thinking about markets. We're thinking about the problem that we're solving.

You know, we wanted to like make sure that we could provide as much impact as fast as possible for our users. So the way that we think about markets and why we launched in Mexico City is because Mexico City is the largest city in North America. The market size of Mexico was huge. And in Latin America, people concentrate around the biggest cities. So 60% of the opportunity of Mexico was in one city. So when you're starting a business, the last thing that you want to do as you're going for that product market fit, as you're building all of these infrastructures, go too broad, too quick. So we actually, when we started the business, we went, no, let's go into the place where we can just spend a lot of time trying to understand the problem without expanding and just building everything.

And that was Mexico City for the first four or five years. And we didn't have a lot of money to begin with as well. So we couldn't build all of that infrastructure that I mentioned that I was going to build out. We just had to put all of our time and building sort of like the data capabilities to be able to scale that. So at the beginning, we actually just spend a lot of time in Mexico building this, building a product that had product market fit, profitable product in the first four years and growing. And then we realized, you know, was that for us to launch this in any other market

and do it well, it takes you four years of just building the basic little things so then you can think to start to scale them. So like when we started to expand, we expand them. under the same philosophy is, okay, our home ground and what we're building and what's getting big, it's Mexico. But then, you know, we saw cities like Sao Paulo, cities like Buenos Aires. And how do you, I guess, when you think about expanding, like, what led to those, you know, like it was the density of the cities or like, how do you guys pick, you know, the cities

that you guys were going to expand? It's, you know, one thing that was really relevant for us is the size of the overall market. So we were looking for cities that, you know. It was, you know, bigger than a $20, $40 billion opportunity as a city as its own in terms of transactions. Two, how deep the problem is that we were solving, you know, whether it had, what was the fraud rate, what was the financing penetration. You know, the more problems that it had, the more frauds, the less financing availability, the better for us and for what we were building. And also the size of the ticket of the car, right?

Like economics play a big, it's really important at the beginning that you have the right unit economics to sort of be able to continue to fund your growth. So any ticket size that was below $10,000 didn't make sense for us as an expansion city at the beginning. So when we mapped out those cities, you have very few cities left that are going through all these things. And that's how we mapped out the cities that we're in. And sometimes we went in really not aggressive. We just went in slow and learned. Other times we went in a little bit more aggressive.

And we got the learning through sort of like just doing this in several cities throughout the past years. But it's always been about not just establishing a global presence. It's about getting, investing and seeing these markets for the first four years, follow the data and try to make sure that you build the right data capabilities to scale those markets. And what we see in every single one of them is that after a certain amount of years that you build all of this infrastructure below it, that skillability becomes a lot easier and replicates what we saw in Mexico. So it's never been the ambition to build a global company. The ambition is to solve this global issue.

And we're trying to pace ourselves into that. All right. I want to jump back to 2017. and you put out a company-minded memo, which started with, I'd like to start complementing our people pipeline with a robot pipeline. And just to ground all of us, 2017, the transformer paper had just come out.

It was still five plus years until there was chat GPT. And then you were pretty specific. You're like, we're going to have Robot 1, the VP of sourcing, Robot 2, customer satisfaction, Robot 3, cross-selling, and so on. And so today, many companies are starting to implement that.

In 2017, that was not something that was top of mind for most. Like, bring us back to your thinking there and what drove that? We knew that we were building this really complex business. We were building, as we mentioned before, a business that was vertically integrated,

had many sort of different businesses layered into it. And we were dealing with real edge cases. When you're trying to solve lifetime value for customers related to their car, you're always dealing with an edge case. It's not a straightforward solution for any given problem. You know, it's the most,

important purchase a user is going to do. They're probably financing themselves for the first time. It's probably their first car. And their car is going to break down at some point, and it's going to break down in places where you don't want it to break down. So you're always dealing with different situations. You know, we're

the master of edge cases. And when you're dealing with an edge case, what happens is that it requires a lot of information and communication from the company to be able to solve that in a way that the user is satisfied. And what I started to see in 2017 in our company was that that highway of information,

it worked really well when we were really small because we were all in a room when something like this happened and we can make decisions really quickly and we had all of the information. And we were able to develop an amazing experience, but it was really clear to me at that point in time that if we scaled and if we became a company that's doing hundreds of thousands of transactions every year or millions of transactions every year, that was not going to be true anymore. And that's the moment where we sort of like when I sent out that memo and that communication was just projecting ourselves into the future that this was what we were going to be, you know, trying to solve at scale. So for me, it was all about building from day one, the right ontology around all of our different businesses or business lines for them to be able to communicate together. And you see it today, right?

Like when you scale a business, what's the issue when you scale a business is you have many people. people everywhere, right? And they cannot communicate in the way that they communicated before. So if you're in the invoicing team and you need to invoice a car for a customer and you have a doubt with the auto parts piece of the invoice, you have no way to know who to call. If you have a doubt with the price of the car, you have no way to know who to call. And that was a challenge that I was that I was seeing at that time, you know, was like, how are we going to create this highway of information where we can sort of like talk to

everybody. And I think that there's a show that I just watched a couple of episodes last week that can help me land this, but it's this show called an apple called Pluribus, right? Is this a show about a virus hits humanity? And what the virus does is that it connects us all in hive thinking, you know, like I have all the knowledge that you have. I have all the knowledge that Gabriel has. So everybody is sort of everybody. And in real time, they can connect all of the information of humanity and just deliver an output or a communication or an innovation, right? And AI is a lot about that, right? And building a business is around like the best businesses will communicate really well within

their organizations to solve a determined issue. It could be an operational issue. It could be an innovation issue. But if you manage to get that communication flowing, the velocity that you can innovate, the level of service that you can deliver was huge. And we saw that because it was hitting us in the face every day. I don't think there's many companies out there that have to manage so many educases.

So I wouldn't give myself too much credit. It was just about facing that problem on a daily business and then just pushing, you know, towards building the ontology because the only thing that I knew that was going to get us on the other side was the data piece of it and how we connected all of this business. That's where when you think about my business, particularly, I only learn through trial and error.

That's my way of learning. And everybody has to have different ways of learning. And I think that you need to make a lot of bad micro decisions every day and show up to see what happened to have a great macro outcome after compounding sort of like these decisions. And in Kabak, we have what we call the four places where business will succeed and fail every day if we do it well or we do it bad.

One place is pricing. If I buy a car and I buy it at the wrong price or I buy the wrong car, I'm going to lose money. That's a fact. So we knew that. We knew that in order for us to do pricing well,

we needed to be able to learn every day in each car that we bought to make sure that we bought the next car better. And you have 60,000, unique SKUs with unique features that you need to price in a given time, with a macro happening around you. So we knew that we needed to put AI in data in terms of how we thought about pricing.

And we knew that we needed to do that on a daily basis. Humans cannot scale that decision. We also knew that cars will fail. And I didn't know anything about cars back then. So the only way for me to learn about cars was not about how I inspected them and then sold them, reconditioned and sold them, the only place where you can learn about a car was when it failed. And this was something that was not popular back in the day

because everybody was something called, well, why are you going to own the warranty? There's no, there's no sense in it. And on one end, there wasn't an infrastructure in the markets that we serve to delegate it.

So we needed to own it just to be able to solve the problem. But the most important thing is that by owning the warranty, you own the data at a moment of failure. And we built all of our systems, just learning sort of like from that failure and every day just making the inspection, the re-conditioning and servicing the post-sell of it better just by sort of like learning that process.

And that's a daily decision that you face that you needed to replace with data. Underwriting is the same thing. Which is notoriously difficult in this industry. So then if we take to your micro-decision point, you started like many companies, like, all right, we will have AI tools, we will give them to our smart employees.

It's going to make everything more efficient. But as you talk about, that actually did not work. Yeah. And so you pivoted more towards AI is just going to have to do this. And there might have been two strategies, tested a little bit here and there. But you guys pretty famously went all in, just assuming it was going to work and then working really hard to get it up to human level and then past human level. How did you manage to move the organization in that direction?

Because it's not just technology, it's also the people. challenges, as you point out. So there were different places where it worked more or less. You know, like pricing, it worked pretty well. Rear conditioning, it worked pretty well. Underwriting it worked pretty well. But the biggest challenge that we were facing was customer interactions, right? Like those edge cases.

The beautiful thing that happened was, you know, we started this in 2000, thinking about this and building the ontology and building our of our systems together in 2017. So by the time that chat GPT came out, you know, we were ready. We were really ready from an ontology perspective. We had all our systems connect like that. It was really, we had the systems in the right place to sort of like extract information to start to build agents. And we needed to make a decision on how we viewed our organization going forward with this technology. Right.

And we were, we knew that we couldn't scale the way that we were scaling. You know, we had too many people. The highway of information of things was not optimal, so we were not delivering a great customer experience or that customer experience that we wanted to deliver on these edge cases. And we knew that if we wanted to scale, we needed to change that immediately. So we started to think about more of that architecture

of how can we make sure to build agents and build them in a way that they can communicate with each other. And it started from the recruiting process. every time we had a friction or we had a need we started to think about well can this be solved with humans that have blood or can this be solved with agents that you know have electricity going through their veins and we started slowly just thinking about like

where do we need to put an agent where do we need to put a human but the first thing that we built out was that operating system that allowed us to know what to do next. And when you're building AI, the first thing that you need to build is the brakes of the system, is the understanding on where you're going to deploy it and how you're going to deploy it. We had thousands and millions of information lines about our users,

what they were navigating, how they were chatting with us, where we were failing them, where we were succeeding. And the first thing that we did was we just built the system that allowed us to understand. everything about every user, right? And that's something that you couldn't do prior to that chat GPT era where you had the information, but you couldn't

sort of like extract and automatically know, contextualize it, right? And we made the typical mistake that everybody's making, that is, you know, we build these co-pilot tools to give them to our teams so they could just use them and just provide a better customer experience or just have the information to provide, you know,

a better solution to our customers. And we did just that, you know, in late 2000, you know, late 2022 or in 2020, we spent all the time building these tools for everybody in our organization, and we gave them to them. And we realized very quickly they didn't adopt them. And when we saw that,

we knew that even if you had the right information, even if you had the correct highway or database for people to find out what they needed, they were not going to do it. So we went funnel by funnel putting agents in front of like critical aspects of the business. So it started by we didn't like every, most people attacked it back then just trying to like solve customer service. We didn't like we for us the issue was not customer service. For us the issue was underwriting your loan.

It was if your car broke down, what do we do then and at that moment? So we went and tried to solve the. hardest thing first. That was those edge issues that our customers were having. We had all of the information of our users. We just needed to build sort of like the right skill sets to solve that. And we created this process where we built agents working symbiotically with humans in the organization. And some things human solved, some things agents solved, depending on a complexity. And we had an orchestrator in the mix that was sort of like guiding who was going to solve which. and we started very slowly with one funnel, one process,

create a skill set, and we started to see how, you know, agents could perform better than humans on certain tasks and started to scale it from then. And for the past three years, we've just been creating all these skill sets that have been able to empower a lot of our AI agents and build the architecture that allows them to communicate with each other for us ultimately be able to orchestrate something simple for our user,

But we put the agents in front of our users really quickly. It was very painful. Yeah. No, I don't think this is what is unique is you had weeks or maybe even a couple of months where various functions were underperforming the humans. And you just had the sort of confidence that you were going to get there and resisted the temptation to hire more humans back or give up on certain tasks. No, we had a year.

Yeah. So we were growing like the business was growing 300%. Like 2022 was growing 100%. Yeah. In 2023, we were flat. You know, so the reason behind that was we were just restructuring everything below the line. And when you start to put these systems in, what typically happens is your customer experience starts to deteriorate at the beginning because you're not delivering at the same level that a human could deliver. Your sales is dropping because we're not doing customer service.

We're doing sales. We're doing purchases. We're doing financing. So you've seen all your output KPIs just deteriorate. I think that the resilience that we had at that moment was, you know, we didn't have for a certain funnel, not for the whole company, we didn't have a plan B. The plan A was we need to make this work

and we got our engineers, our teams, and everybody just making sure that these agents could perform at least priority to what we were performing before. And it was in a certain funnel you see it drop, up until it doesn't. And then you start to, it starts to go back up. And we got to a place in the first funnel that, you know, we were like, like 1.5 times better than our best human. And in a certain place.

And at that point, we said, okay, let's not focus on making it seven times better because the models will get there. And we have an amazing philosophy in the company that is we don't build for chat GPT4. We build for chat GPT 7, right? or we know that the models are going to move faster than what we can move. So we try to like anticipate what needs to be done. So like when that technology is there, our infrastructure and our architecture sort of like can can actually take advantage of it. And we started to work on market share of these interactions with our customers.

So we went when we saw funnel like delivering parity or a little better than parity than human, we went to the next friction, so on and so forth. And it was always the same pattern. It goes down before it goes back up. and we slowly started to sort of become more confident about that strategy and up until a point where we decided, well, you know what? We need to burn all the ships and just do this for the whole organization, right? Which is different.

You know, today we manage like 90%, 95% of every human interaction with our users with an AI agent in the middle. But that's not a reality for all of our backend, right? Pricing, invoicing, all of these things we have ML, you know, we have all of these systems. Now we're building sort of like the agents to manage these systems so they can also converse with all other agents. And it's just you just have to have the patience of it.

But it took us, you know, like it was a year flat, a lot of under a lot of pressure, you know. How did you manage, I guess, that pressure with the culture within the organization for such a long period of time? It comes ultimately to what we believe the world is going to look like in a few years, right? You know, we think that this is the biggest technological shift that's ever happened to humanity.

And our responsibility and as leaders in this time and for our company is to make sure that we're not only on the other side of the shift, but we're thriving. And I like to go back a lot to like Netflix. Like, I love Netflix because they started out as a videotape company, right? and then they went to streaming. Well, let's unpack that. You have Blockbuster, then these guys innovated, and you had videotapes delivered to your house,

so Stormen wasn't an issue anymore. And then you had streaming. The way that I like to think about Kabak is that incumbents or Blockbuster, Kabak was videotapes sending to your user, and that's probably traditional software. Yeah. And selling online and underwriting online and everything. And then, you know, moving to AI is sort of like that streaming,

transition. And you know, when you're doing a transition of this shift, you need to make sure that everybody understands that we're going to go streaming or die. Right? And one of the things that we had going for ourselves is that our market imploded. I don't know if you remember

what happened to the auto car, the aerospace market in 2002. 99.9.9.9. A lot of business went under. So it was literally nothing I can do or say to anybody. that made them feel comfortable, right? You know, like when you're, when your industry is down 99.9%

you could, like, everybody's going to be uncomfortable. So maybe set that backdrop of where you were. Like, Kavak, I think at that time, had 10,000 employees. Like, it was a 2021, 22, right? It was a very high-flying time. So 10,000 employees, like you were burning a million a day. Market implodes. We were building this, like, we were scaling,

we were building the infrastructure, a lot of infrastructure to scale. We had more than enough infrastructure to grow four times where we were at that point in time, growing really quickly from going to 200 people to 10,000 people, like in 24 months' time frame. The business went from zero to billions in revenue at that same speed with a lot of complexity behind how you deliver that revenue, right?

You're moving cars. So we had everything, you know, just ready to go really quick. And all of a sudden, markets employed, and you have to change your strategy, you know, you don't know. and more so if you're building in Latam you know like capital comes in in that time and well it never came in at the speed that it came in in 2021 and it never left so fast

as well right like you so you have to just rethink your whole strategy right it's like you know capital is not going to be here forever and we just need to make sure that we make the shift but when you're making that shift remember that it's like you know you're at full speed

going ahead and say it's like taking off on a plane right like that's the worst time to slow down you need to get yourself to like 10,000 feet up in the air before you start to make same decisions and we needed to slow down while we were sort of like going through that takeoff process you were facing so many things you know like should I just push for for growth should I slow down and just by time we were in a in a really interesting pickle because we couldn't just shrink to profitability, we needed to grow while becoming more lean and efficient at the same time. And the company from there, you know, we went on to grow four times to where we were at that place in time. And we went on from burning a lot of money to become profitable in that

frame of time. But it was, everything was very heightened. And when your market down, there's no decisions. When your market goes down 99.9%. There's no decision that's making anybody comfortable, Right? Because for us, we knew that we had something special. You know, we were not just building a business that people wanted. We were building a business that people needed. The metric that makes us more proud is that 40% of our customers are buying their car for the first time in their life. This is because... That's incredible.

We knew that we had something special. And we were also building these AI capabilities in the midst of it that we also saw is very special, especially as operators, because the same thing. scalability of it, the thing that we were suffering through the growth phase that's never ended. We knew that it could probably be easier in the other side. So I would say

like when you have everything coming at you with that level of intensity, expectations all of a sudden lowered. Right? Because there was nothing that I could actually do that would make everybody comfortable. More than just get profitable and go

through it and we were definitely working towards that objective, but the most important objective was not just getting properly, it was getting significantly profitable on the other side and being able to deliver more impact for our users. So we had that opportunity to, you know, make bets. And you made a big one. And we made a big one. We continued to make it, right? But we definitely knew that we needed to do this transition where we were streaming.

And I believe that any company right now, any CEO right now, if you're not thinking about, it in that way they're going to be up for a world of slow death and pain so we're forced into that transition we also have a lot of founders from all around the world that listen to this podcast curious if you have any advice that you can share with this founders that are building in emerging markets they say Latin America and you know how how to think about building a business now I think first you, a lot of founders, they go out there and they just want to be founders and build a business. I think that's the wrong take.

I think you just need to find a problem that needs to be solved and that you can probably at the early stages of it, relate to it as much as possible and that you care about. And if you find a problem that needs to be solved, not, in Latam, it's not about things that just make it better. you know, you're there dealing with that most of your population is going to be struggling. And they're suffering against, you know, they need to feed the table. They need basic needs are not covered. So if you're building in Latam,

you have to think about, you know, building products that people actually need, not that people necessarily want. And if you find a problem, you know, you have to go all in and make sure that you're solving for something that has real impact. I think the purpose behind it will get you through the highs and the lows if you find the right thing. And I also think as a Latam founder, you have to be also agnostic of region, right? You know, if you see something that can be solved in the U.S. that you're excited about, you know, just go there, right? One of the things that in Latin America is that nobody tells you that you can think big.

nobody you're not brought up sort of like in that way that you find you know when you talk to founders here and when you talk to VEC's here you know the level of ambition and especially in Silicon Valley

you know thinking big is it's the norm I think that's what I think you know founders need to understand that they can think big but they also need to understand that they need to solve something real right because it's going to take

it's going to take years like if you if you see the best companies out there 97% of their value gets created after year 15, to year 1-5. 97% of their value. I've talked and listened to a lot of these founders that have built this generational multi-billion dollar companies. And the insight that I get from zooming out and looking at these companies is that when you ask them, like, how can you be at year 20, still be growing 30%?

What happened? to make that happen. And like most of the feedback or the learnings that I got is that Carlos, it's just about taking frictions away from my users every single day, every week, making your business 1% better. And making your business 1% better means taking a friction away from your user and making their friction 1% better.

And that compounds over time. And if you do that and you focus on that relentlessly and what you're doing, users really value, you're going to figure it out, right? My job is cleaning cars, right? At the end of it, we're building all of these AI. We're being in all of this, but at the end of the day,

is delivering a car that's clean. So even on the worst day, I could just go and wash a car and feel good about myself, right? So it's making sure that you find that friction and you take it away. Like great people, clean the car, amazing people build systems

that can do the scalable for millions of people, right? So I would say that gets me through a lot of it, just understanding that we're at year nine. So none of our value has been created yet. And we just need to take frictions away. And when you do that, scaling is easier. Four years ago, we didn't have fintech on in our business. It was very hard to do what we do because a lot of people needed financing.

And we couldn't control that piece of the business. Now we do. And it's thriving. And we're helping a lot of users out. And the reality is I make a decision today. and I need to factor the fact that I will show up in 24 hours to reassess that decision. Sometimes you make a decision

and you let it out there roll for 15 days. And I know when I do that, it's wrong. You know, if I make a decision, I need to get on a call the next day and say, how is this going? And reassess. And you have to be comfortable with being wrong. So I think enjoying the micro of it

is very important. and treat problems as a protein bar. Right? Like, our job is literally to solve problems. If you're not excited about solving a problem, if you're not excited about hitting complexity. That's a great analogy. If you're not excited about doing that,

what are you doing then? Like, every time you solve a problem, you make it, you build these barriers, this defense. So you need to enjoy it. That's the job. look at my story. You know, I come from Karak, from Venezuela. Like, I'm a glitch in the system. Like, I'm not supposed to be here talking to you guys about building this business. I'm so fortunate about having this opportunity and you need to acknowledge that every single day, you know. A little on the, on the zoom out, I think one of the things you shared when we were talking before that's unique about you is this annual process that you go through of self-introspection, which essentially, as you ask, am I still the right CEO for this phase of the company? Yeah, I just went through that, yeah.

Yeah. Do you want to talk a little bit about that exercise? Yep. So you have to balance your role as a founder, stakeholder, and as an operator, right? And we do this. I do it myself. We do it with our management team and team members. Every year or so, I fire myself.

You know, and I go through this exercise that is really painful, you know, firing yourself, it's really painful. Because your business, if you do well, is going to grow faster and it's going to be, than you can ever grow as a human. So the process that I go through is every certain, sometimes it happens every three months, sometimes it happens every couple of years, but I definitely go through it a lot. I just fire myself and lay out.

You know, if you're going to hire a CEO to build that vision that you laid out, what would you hire for? Like what would you ask him or her to do that's relevant? What are the risks that you know that they need to be taking? what are the things that you know that they need to protect and compound? What would you lay out like the job? And how would you guide them through the next year of their role

to make sure that they're successful? And I spend a couple of weeks just thinking about that. And then I go through the next, and this is hard because it puts all of your, you have to put everything, all of your emotions and everything in check, right? Like you're always, as a founder, you're sometimes doing things that you shouldn't be doing. You're trying to do too many things once

instead of sequencing things. You get impatient. You have all of these things that as a founder and operator you go through. And this is a time where you can, I take a time to like take a step back and make sure that I think about it a little bit more coldly. Listen more to like all of the inner voices of my stakeholders, my users, my employees, my family

and like all of the things that are out there. And I figured, that out first. Then I ask myself, can I hire myself back to do that? Right. And the pros of hiring myself back to do that is one, there's a huge probability that I'm going to care more than anybody else about it. And caring is important. Two, you have a pattern recognition that nobody else is going to have because you build it from scratch. So you can probably like see certain things and solve them quicker than anybody that you're going to bring in. But the most important thing is you're going to have to have yourself,

okay, in order to deliver what you want that CEO to deliver, what do you have to let go of? So you go through that process of letting go of certain initiatives, of letting go of certain pieces of yourself, you know, certain bully. Because in order to build businesses, you have to become a persona. You're like this, you're always like, that you know these actors that are always in Jared Leto, you know, that they're always in

character. You sort of have to get into a character because when you're doing this moment in time you have to get away from that character and realize what's the new persona that is needed for this new phase? And they go of a lot of the things that make you and build new things, build new skillsets, build new ways of thinking.

So I go through that process a lot. So far, I've been hiring myself back to do it And I really want to do this for the rest of my life. You know, like, but it doesn't mean that I don't have to go through that process. And it doesn't mean that, you know, the best person needs to be on the field operating. So that's something that I would recommend people to do. And you'll mature through that process. You'll change through that process.

But ultimately, you're always, that person that you are, you know, that kid or that founder for that certain stage, it's always there. And you're always going to have that as your instant. An instant is really important in decision making. But now it's guided towards a framework of what the company needs from you in this stage, what your employees needs from you, what your family needs from you, what your stakeholders need from you. For certain, if you talk to my employees, my family, my stakeholders,

they definitely want a different Carlos every year. Right. And they're excited about past Carlos, afraid of past Carlos as well. my job is to make sure that I get myself in a position where I can continue to deliver, you know, an amazing company for all of us. Thank you. I think that's actually a great note.

We love 2025, Carlos. Are excited to see 2026, Carlos. Thank you for joining us. Thank you, Carlos. Thanks for listening to this episode of the A16Z podcast. If you like this episode, be sure to like, comment, subscribe, leave us a rating or review, and share it with your friends and family. For more episodes,

Go to YouTube, Apple Podcasts, and Spotify. Follow us on X, A16Z, and subscribe to our Substack at A16Z.com. Thanks again for listening, and I'll see you in the next episode. As a reminder, the content here is for informational purposes only. Should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund. Please note that A16Z and its affiliates may also maintain,

investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16Z.com forward slash disclosures.