← The a16z Show
The a16z Show
Stablecoins, AI Agents, and The Future of Global Banking
00:37:05
Angela Strange speaks with Dileep Thazhmon, founder and CEO of Jeeves, about building a global financial operating system for enterprises across Latin America using stablecoins and AI. The conversation covers the challenges of building localized financial infrastructure across 25 countries, from regulation and payments to underwriting and compliance. They also discuss why stablecoin adoption is accelerating in Latin America, and how AI is helping Jeeves scale billions in payment volume while automating underwriting, customer support, reconciliation, and KYB workflows.
Loading summary
Transcript68 lines
- [00:01]
Dilip Tasman
The goal with Jeeves is really to build a global business bank that can function in multiple countries. Stablecoins is a lived experience. It's not something that's a theory. In Argentina, 60% of the population use stablecoins. Our revenue has grown 10x. Our volume has grown 8x. This just wouldn't be possible without AI. Money is going to become programmable and hopefully we're at that nexus of all of this happening.
- [00:22]
Angela Strange
Often when companies start in financial services just because there's so much product to build, the tendency is to start with smaller guys, grow with them, and then eventually earn the credibility to get with the bigger guys. You took the opposite approach. How did you become relevant to relatively large companies fairly quickly in the journey?
- [00:38]
Dilip Tasman
Our underwriting team today is four people and that team is doing 2. 3 billion in TPV. That just wouldn't have been possible two and a half years ago. You need 15 people just to get that off the ground. That's what's changed. If you are not AI pilled, you're not going to make it.
- [00:53]
Podcast Host / Narrator
Most financial infrastructure is still fragmented. Country by country. Different banks, payment rails, currencies and compliance systems make it difficult for global businesses to operate efficiently, especially across emerging markets. Jeeves is trying to rebuild that stack using stablecoins and AI. What started as a corporate card company in Latin America has evolved into a broader financial operating system spanning payments, treasury and expense management across 25 countries. A16Z's Angela Strange speaks with Jeeves founder and CEO Dilip Tasman about stablecoins, AI and building global financial infrastructure.
- [01:34]
Angela Strange
Dalip, welcome to the podcast.
- [01:36]
Dilip Tasman
Thanks for having me. I'm super excited.
- [01:37]
Angela Strange
Before we dive into Jeeves, I don't think a lot of listeners know your backstory, which is pretty interesting. You grew up internationally, you've got a technical background, you started another company, sold it for over $100 million before long before this.
- [01:50]
Dilip Tasman
Yeah, yeah.
- [01:50]
Angela Strange
So maybe tell us about you and how that led you to start a. Geez.
- [01:53]
Dilip Tasman
So I was actually born in Nigeria, so I grew up in three or four different continents actually. So we moved from Nigeria to Doha, Qatar. So that's where I spent most of my childhood. And this was when Doha was like a one camel town. I mean, it's funny looking at the airport now, which is massive. And it used to be an airport, like two stalls and plane A or plane B, and you get on one of them and that was the airport. So it's very small. And then we moved to Florida outside of Orlando, probably about 25 years ago. And so what I've kind of noticed is the company that I'm working on now, Jeeves is also a global company. So it's almost like a manifestation of my childhood and the fact that I grew up in all of these different places. I then went to school for engineering, so I have actually two engineering degrees, which now I feel a lot more unleashed with all of the stuff with AI, because I can start building. But the funny story is my master's thesis was actually on using genetic algorithms to mimic money markets. And that was basically LLMs, which is basically, you'd have a end goal and it would come up with test cases, rank them, and it's just wild. This was in 2006, and it's amazing seeing that full circle 20 years later with everything that's happening with AI. After that, I started a company. I was in the marketing automation space, and right before that I was at Stanford gsb. And that company was Spyro Inbox, which we sold for about 106 million. And then I started Jeev. So basically, at this point, if I don't start a company, I'm unemployable. So. But yeah, it's been a great experience, learning a lot of different things. I think what I love about Jeeves most is it's just such a steep learning curve. One day you're doing banking in Brazil, next day you have to be an expert on stablecoin. And if you don't like the learning curve, it can be so exhausting. But if you like it, it's very exciting. And that's the biggest thing that I find interesting about the day to day.
- [03:33]
Angela Strange
Amazing. And so we're going to. Maybe we're going to talk through the backstory, but just to ground us, we're five years into the company. What is Jeeves today?
- [03:40]
Dilip Tasman
So Jeeves is a stablecoin native financial operating system for global enterprises. We have two core products. The first one is expense management. So we sell corporate cards that we offer to our companies and we can do that in 25 countries. Our core markets are Brazil, Colombia, Mexico. And then our second product is on the payment side, so we do full local payments and international payments. Today, our international payments, more than 50, 60% settles directly on stablecoin, and that's been powering a lot of the growth that we've seen in the last year.
- [04:08]
Angela Strange
Amazing. All right, so I think one of the. One of the most interesting ways to understand Jeeves is through your moats. And now they're moats. I would say they were massive challenges to building the company and getting scale. And now we can classify them in the realm of defensibility. And so two big ones, infrastructure and regulation.
- [04:25]
Dilip Tasman
Yeah.
- [04:26]
Angela Strange
So let's start with infrastructure, which is my personal favorite because it looks like unsexy plumbing. But then the decisions there are really what drive the products you're able to offer. Specifically, specifically the differentiation. And I think where you started in Latin America, an added challenge is unlike the us, there's no as a service for all of these different pieces. So maybe talk us through, like what did you splice together, what did you build? And then like, how does that enable you to drive?
- [04:54]
Dilip Tasman
It's a really good point. So internally we have this saying which we say difficulty is very defensible because it's difficult. It's actually defensible. And we've done a lot of difficult things from day one. And so just to give you an example, our core ip, what we consider our core product, so to speak speak, is not the front facing product, it's the layer in between. We connect to 50, 60 different partners, different vendors, and we abstract all of that information into the central layer. We call it a cms. And so just to give you an idea of what it could do. So as an example, we maintain our own ledger in 25 countries. And that is really hard to do because you have to do that across currencies, across FX. We issue statements in 25 countries in less than 24 hours. That took us 18 months to figure out. And that's a core skill set because then the ledger sits directly with us and we sit directly on the card auth stream. I think a big difference where we play on the card side is we're full principal members, for instance with MasterCard. And so when we issue the card it says issued by Jeeves Mexico, issued by Jeeves Brazil. There's no bank in between. And so that obviously requires a certain level of sophistication to be able to run. And we've kind of learned that over time of how to build that. But we try to go very, very close to the infrastructure stack in all of these countries that we play in. I'll give one more example on the card side because I think this might open up how we think about it. When I think about card issuing, I see there's four components. So you have the top of the funnel, customer acquisition, you have the processor, you have the program manager for regulation and compliance, and then you have the issuing bin from Visa or MasterCard that gives you that license to issue the 16 digits. What we do the first three. That sits on our own stack, even though we have different partners. If we were going from Mexico to Brazil, we only have to change the bin. If you tomorrow wanted to launch Indonesia, we only have to change the bin. And now with stablecoin, even the bin doesn't have to change because you get one bin and you carve it up by different countries. So it's a super exciting time to be building. And part of the reason we're seeing this acceleration on stablecoin is because we did the hard work of building our own infrastructure stack for three years.
- [06:47]
Angela Strange
Yep. And so for people who are maybe less in the weeds of infrastructure, like, why would it matter to someone like nubank that you have a Brazilian bin versus a US Bin?
- [06:58]
Dilip Tasman
Yeah. So I think the value is that the customer experience is very seamless. So one of the things that BCEL is if you log into Jeeves in Brazil, Mexico, Colombia, it looks the same. And you're used to seeing the same cards in the same places. And you can't do that unless you sit directly on the rail. And so when we started, which I think you might remember, we actually were shipping U.S. cards.
- [07:21]
Angela Strange
I remember.
- [07:22]
Dilip Tasman
Yes, we were shipping U.S. cards because we didn't have the setup yet. It takes time. It takes time to get the licenses. It takes time to get the processors, all of those things. Today, in every country we operate in, we have local issuing and local licenses. And then what that gives us is the experience. For companies like a hotmart for a Burger King, it's seamless. It's seamless in the sense that you don't have to train the team three times for three different logins for three different countries. It looks exactly the same. And you can't do that unless you sit on the rail.
- [07:46]
Angela Strange
Yep.
- [07:46]
Dilip Tasman
And then just kind of building on that as well. Our acceptance is super high because we are sitting directly with our own bin. In Mexico, Columbia, Brazil, we have our own bin that's just for Jeeves. And then the second part of that is we can also build in much better card controls because again, we do the actual authentication. When you swipe it, we say yes or no. We also do some of the fraud detection components, obviously with partners like Sardine, but all of that sits directly with us.
- [08:11]
Angela Strange
Okay, so infrastructure, the next critical but often tedious topic is regulation.
- [08:17]
Dilip Tasman
Yeah.
- [08:17]
Angela Strange
So maybe talk us through. We've been at this for sort of five years at this point. I think you've got a lot of your licenses. You've got another roadmap. What have you Built there.
- [08:26]
Dilip Tasman
Yeah. So where we sit, we sell directly to mid market and enterprise companies. So we do not really sell to the long tail. The reason that I bring that up is it's really hard for us to gain the trust of these companies without the right licenses. So we are a regulated entity in Mexico. We have filed for our money movement license in Mexico as well. In Brazil we're actually filing to be a full deposit taking bank. So it's a very high bar for a license. But if we don't have this license one, again we don't have the full customer flow. So we can provide some version of the cards, we can provide some version of the payments. But to really be kind of the global business bank, enterprise bank financial operating system, you need to own accounts as well. And to own accounts you need these licenses. So we have always looked at licenses as core defensible mode. As in the product we want to build cannot be built without licenses. And so the issue with licenses is it takes time. And so part of the way I think about building is how do you get into market quickly and get a lot of market share and then you just have a team that just every day we're just building layers and layers of this license. And so that way the customer experience hopefully is seamless. They see the same thing, but on the back end the product is a lot better because now the licenses sit directly with us. And then the second part that is really important with licenses and with infrastructure is it helps us to expand the margin. Our margin has expanded from 40% two years ago to north of 80% right now. You cannot do that unless you own the infrastructure and unless you have the regulatory coverage because then you get the whole lion's share and then you pay it out to partners. So that's the positive side of having all of these things. But it's really hard. It's really hard. And you need a team. That's what they're doing. And we go country by country. And even now I would say we probably need to get four or five more licenses in the regions we're in to get to the stage that we want to get to.
- [10:07]
Angela Strange
So I want to go to. You mentioned you're a little bit more enterprise versus smaller company where I think often when companies start in financial services, just because there's so much product to build, the tendency is to start with smaller guys, grow with them, and then eventually earn the credibility to get with the bigger guys. You took the opposite approach. And so how did you become relevant to relatively large companies? Fairly quickly in the journey.
- [10:32]
Dilip Tasman
So it's funny because we started with a little bit of doing everything for everybody. And then, you know, 2023 was kind of the year that forced focus and we were realizing that we were just not making that much money in terms of revenue from smaller businesses and we needed the same support for them. And we're a very lean company. We have less people today than we did two years ago and our revenue has grown about 10x. And so what I'm trying to get to is we're not built to sustain 10,000 small businesses. That's just not our model. What we're built for is these real mid market enterprise companies that we sell one product to and then we cross sell six others and we increase the retention, we increase the gross profit that we have at these companies. And so it was a little bit of we did everything and then we made this really hard decision in 23 to say, okay, which is the audience that really resonates with the product and what's the product we're looking to build. And at that point we did have to make the hard call to debank some of the smaller businesses that we had. But I don't think we'd be here today if we didn't make that hard call in 2023. And that started putting the focus on what are we offering. So before 23 we did not have a payments product. When you go sell this to enterprise, Payments is a huge, huge part of their business. In fact, cards volume is actually much smaller than payments volume because if you look at share of wallet, AP accounts payable is a much bigger box than corporate card spend. And so we started with cards, we started with these versions of credit products. And so we turned all of that off. We were like, what's our focus? It's enterprise, it's mid markets. What do they need? They need accounts, they need payments. And then how do we make sure that we only sell into that subset in the sense of we have requirements where if you're below a certain revenue size, and that goes from somewhere between 10 million reai all the way up to 100 million reai. Those are the two segments we have. We just wouldn't onboard you, we can't onboard you. And so it's a hard decision because it also then focuses on companies like Y Combinator where we came from. And they've been incredibly supportive and we've had to find ways of like, okay, can we squeeze that in? Because those are the companies we want to build in, but they tend to Be smaller to start.
- [12:32]
Angela Strange
Yeah. So maybe a special YC exception here and there. Maybe just a good way to understand jiu's product suite is to take a company like nubank, which is Brazil, Mexico, Colombia, you landed with cards. Do you want to just walk through you land with this and then where does it go?
- [12:51]
Dilip Tasman
So I think what's interesting about our product suite is we are global, but we are fully localized. So what I mean is it took us two years to figure out how to sell in Brazil. And a big reason we were not selling in Brazil, and Brazil's actually going to be our biggest market this quarter. And it really started turning in September of 24. And the reason was we started localizing the product. And so we made it obviously in Portuguese. But it's more than that. It's tying it into the actual way people use the product. Brazil has an incredibly sophisticated financial ecosystem. I would say, in some sense, even more sophisticated than the U.S. there's no cash, everything is digitized. It's instant. In fact, like when they come to the US And ACH takes two days, it's almost like, what are you guys doing for sale?
- [13:32]
Angela Strange
Exactly.
- [13:32]
Dilip Tasman
And so it took us a while to figure out how to localize the product. And the reason I think that's important to your question of, like, how do these companies use us? We usually start with the card and then we sell second and third products. So second and third products could be different depending on the country that we're in. So in Mexico and Colombia, usually that moves into JeevesPay, which is our payment product. In Brazil, we actually have a very strong travel product. So if you think about Brazil, a big concern there is fraud. It has one of the highest fraud rates in the world. And so we had customers that would come to us and say, hey, if you have cards that are automatically turning themselves off, meaning they're single use, we would use that for all transactions. And so we started playing around with, okay, let's bring build out a cards API that now we provide to these different companies. And so when you think about nubank, you have corporate spend and then you have travel. So those are the two areas that we kind of work with them on. And that travel component is strong in Brazil, but might not be as strong in other countries. So I'm kind of getting to the point where we have this strong primary product, which is usually the card, and then the secondary product kind of changes depending on the region that we're in. And that tends to be localized.
- [14:36]
Angela Strange
So we were fortunate enough to get to lead your Series A. And every investment has a why now and the why now at Jeeves, when you started, which I felt pretty strongly as I had a broad portfolio in Latin America, all who had tens of millions of dollars in their bank account and they could not get corporate cards. And if they could, they'd be from incumbents. That made us incumbents look like the most taxable companies on the planet. Right. So that was a good why. No. Then I think now you've got two more which are even stronger. AI and stablecoins.
- [15:05]
Dilip Tasman
Stablecoins.
- [15:07]
Angela Strange
And so before we maybe dive into specifically for Jeeves, as you think about the region of Latam and the countries that you're in, like those tailwinds I would argue are even stronger there than in the U.S. yeah.
- [15:19]
Dilip Tasman
So stablecoin, let's start with that one. Because I think in Latin America, stablecoins is a lived experience. It's not something that's a theory. In Argentina, 60% of the population use stablecoins. Like that's not something you have to train people how to use. And so in the US it's a little more like, okay, why would I use stablecoins?
- [15:38]
Angela Strange
Don't quite get it.
- [15:39]
Podcast Host / Narrator
Yeah.
- [15:39]
Dilip Tasman
Because I have a stable US currency. Amazing. That's fantastic. A lot of countries don't have that.
- [15:43]
Angela Strange
Yes.
- [15:43]
Dilip Tasman
And so the fact that you have this dollar denominated coin that you can almost instantly transfer your local currency to is a huge, huge selling point. And so what we started realizing, and again, a big part of this was the genius act passing, which started making it more kosher to talk through stablecoin was that a lot of the infrastructure we had could now be adapted towards stablecoin. And going back to your question earlier, the reason we can do that is because again, we maintain our own ledger so we don't have to have two separate ledgers, one for usdc, one for USD. It all feeds into the same thing. And so as soon as the genius act passed, it was one of those things where I would say it was a little bit founder led. I was just like, this is happening. This is not going to go build this whole thing. And in fact, even today, I mean, you can ask our team, we have three standups every day, not every day, every week. And I run them directly because to me it's that important that we move everything onto stablecoin, because I see that as absolutely the future. And specifically for latam, which is one of the highest corridors of stablecoin usage, it just becomes something that people know how to use and the product becomes a lot better. I think the second part on stablecoin that I think is very interesting is when we sell it to these enterprises, we actually don't pitch stablecoin directly. We actually pitch it as Jeeves Instant Pay. But the inside is stablecoin. And the reason that's important is when you talk to a CFO at an enterprise company, they don't care as much. On the technology side they care, can I trust you? Is the money going to show up? And then is this something where the cost is better than what I have right now? And if they don't think the money's going to show up, it doesn't matter what the technology is because at the end of the day they're going to get probably fired. And so we have to win that trust. And we almost kind of guarantee it with the Jeeves brand of like, hey, you use us for all these other things, trust us on this and we'll make sure it shows up. And that's how we kind of sell the product in region. And that's been a huge driver of the volume that we're seeing. And our TPV has increased from about 400 million two years ago to north of 3 billion right now. And we're probably going to touch about 6 billion this year. So it's significantly increase. And a big part of that is because we have the stablecoin as the infrastructure in between.
- [17:46]
Angela Strange
So maybe contrast the and you've got some more innovative products coming out. But to start with the sort of classic cross border contrast, Fiat Rails versus stablecoin Rails because you were serving both customers.
- [17:58]
Dilip Tasman
So if you think about just where we were maybe 18 months ago before we had stablecoin and you wanted to do a payment between Mexico and the US you would usually have to have two different correspondent banks in between. You'd send the payment in Mexico, we'd have to make sure that we had a bank partner that could handle it in the US because in the US you need to know the source of the funds as well. So we basically had two different bank partners that had to do a handoff to transfer, let's say 1000 pesos and get it to show up for $50 in the US now what happens is in Mexico you can send us 1000 pesos. We collect it on the ground in Mexico. As soon as we collect it, we release USDC and that settles instantly. And then we pay out. On the US side, the longest leg is actually the US payout because in Mexico that settles instantly. And the ACH again, still takes one day, but it's much, much, much faster. And now for the end user, we don't sell the stablecoin. We sell accounts payable. We sell. Your invoices are reconciled. We sell, you know, put in your tax id, we'll download your invoices and send international payments. That's what we sell. It just so happens that the infrastructure that it runs on is now a lot better because we can do that in one hour instead of one day or two days. And so that's the way we kind of look at it. We give the end user the choice. If they want to use our Swift Rails and get the MT103 kind of confirmation, we can do that. We can do that for you. Just it'll take longer, but that's your choice. If you want to use a faster version, which you call Jeeves Instant Pay, you can do that in the product. You just have to click a dropdown and select that. So we give you that choice. But a lot of the way we kind of sell it is it's part of the product.
- [19:26]
Angela Strange
Yeah. Would you like your money in a day or five days?
- [19:30]
Dilip Tasman
You'd be surprised how many. Pick one day.
- [19:33]
Angela Strange
Yeah, exactly. Not surprised at all. Amazing. And then coming back to Argentina, which I think anyone who's been there, lived there can get a sense of, your money at the end of the year is worth nothing close to what it was at the start of the year. You're about to launch a stablecoin card. Who is that for and how is that going to look?
- [19:51]
Dilip Tasman
And I think taking a step back, and you probably remember this when we first chatted, the goal with Jeeves is really to build an enterprise version of Revolut, so a global business bank that can function in multiple countries. So to be able to do that, you have to launch multiple countries. And so, so previously launching a country is a big undertaking because you have to first get local entities. Again, in this example, local issuing licenses. You need to be able to collect funds. So it would usually take about eight months, and that's very fast. We were probably the fastest to be able to do that in eight months. And that would cost maybe 500, 600, $700,000 to just get spun up now with stablecoin. And Argentina is our first new country launch in three years. We're super, super excited about this. You can do that whole thing much, much faster and much cheaper. And so in Argentina we have our own pay in Rails, so in on Argentinian Peso, as soon as it comes in, it converts to usdc. So you instantly get currency volatility protection. Which in Argentina, where currency devalues, is a huge, huge, huge selling point. As soon as it converts to usdc, we can issue corporate cards on it. And now if you think about our core product, it's not necessarily just the stablecoin or the payments. We sell the layer on top, we sell the operating system. So now we can sell budgets, we can sell approval flows, we can sell teams, we can sell card controls, we can sell fraud controls. All of that can now work. And what's really, really cool about this setup and very unique compared to a lot of other stablecoin cards in the market is if you swipe that card in Argentina, it does not have an FX fee. And the reason that's important is the customers we go after are mid market and enterprise customers. They have 50 employees. You're not going to get a card for 50 employees. If you're going to get charged 2% when they buy a Coke, that's just not going to work. And so it's a really, really cool version of a product that we've used a lot of different technologies to piece together, but now allows us to sell the full product suite in Argentina. And tying this back to kind of what Argentina is, what it's going to become is it's going to become the blueprint for other countries. And so take Peru for instance. Nobody's going to launch an expense management suite for Peru. It's too small of a tank. Now we can. Now I basically have to staff two salespeople in Peru. They'll sell this product out there. I can onboard 10 companies and it can still be a profitable market. So the cost is completely collapsed in a way that, you know, previously like no one's going to launch in Peru. Now we can launch in Peru because my cost is three sales headcounts. It's amazing.
- [22:09]
Angela Strange
So I want to shift gears a little bit to AI, which I think has been the other. Like, oh, thank God. Like for instance, I think one of the things we all underestimated was just KYB in for instance Mexico and just the volume of documents you need to collect, what you need to link languages. Languages, exactly. So that is an example. But maybe talk through where are you using it in Jeeves? And then what is the impact that it's had on the business?
- [22:36]
Dilip Tasman
So it's one of those things where just like a lot of companies we have tried to figure out how to incorporate it across the board. I would say some functions we're seeing a lot more adoption than in other functions. If you think about anything tied to document ingesting, we've seen a lot of success there. So I'll just give you an example. I was kind of mentioning again, our product, core product is like a 30 day charge card, like a ramp or Brexit. That still requires us to cover the customer spend for 30 days. So again, it's not a credit card. You can't keep it open. You have to pay us on day 31. But for 30 days we are covering your spend. That requires underwriting. Our underwriting team today is four people and that team is doing 2. 3 billion in TPV. That just wouldn't have been possible two and a half years ago. You need 15 people just to get that off the ground. That's what's changed. The models are better. It's self learning. One person that's really smart can make this work in multiple different regions. So that's one example. But we're seeing that across the board. The only real headcounts we're adding are some components of engineering and product and then go to market. And go to market is one area where we haven't had as much success using AI. And we're trying different components for enriching data, et cetera. And so we do hire humans there. But on almost all the other functions, we're seeing a significant, significant uplift using AI. And I think the biggest number is that our headcount today is 140 people. And two years ago it was 200 people. And our revenue has significantly increased. And that's why you see that margin expansion. Basically revenue's going up and costs are the same, are going down.
- [24:08]
Angela Strange
Yep. Well, customer service, I think is another sort of notoriously challenging in multiple different languages, multiple different countries.
- [24:15]
Dilip Tasman
Customer service is one where we do have our own internal kind of chatbot. We call it Lenora. And yeah, it handles most of kind of what people bring up. I think what's interesting is, and it's taken us a little while to learn this is just the cultural nuances. So everybody's got ocr. But it took us a while to figure out that in Brazil they use the commas and the dots interchangeably. And so when you're doing underwriting and you see the commas, the commas are actually the dots and vice versa. And so that training is what needs specialization and that's again, specialized to just jeeves and what we do. So, yeah, it is something where we're seeing it across the Board customer service, for sure. Kyb. I think document ingestion is another area that we've been using that as well.
- [24:59]
Angela Strange
Yep, great. And so now I want to fast forward a little bit to the future. And you gave us a little bit of a preview about this. Right. Like, it's getting easier to launch more countries. Things are coming more global. As you think of, you know, Global Business bank two years from now, five years from now, like, what does that look like? What might be a little less intuitive to anyone?
- [25:17]
Dilip Tasman
Yeah. So I actually think a lot more is going to go on chain. And so there's two, three things that we are thinking about. So, like, if you think about a business like ours, we have to still maintain capital pools in different markets, right? So we have to maintain pesos, brl, Mexican pesos, Colombian pesos, et cetera. We maintain a lot. What would be very interesting, which is what we're starting to dabble into, is actually just keeping a single source of USDC or USDT and minting when we need it. So that way the cost of capital significantly drops and we don't have to maintain these pools. So when I think about where this goes from an infrastructure perspective, I actually see a lot of the existing stack moving on chain. So the second part of this is going back to kind of that whole 1, 2, 3 billion TPV that we do. Is there a world where we can start collateralizing that on chain? So we basically take the receivables, put it on chain, get 60, 70% against that on USDC. It doesn't remove our existing capital sources, but it helps supplement that with on chain usdc. So when I think about where the infrastructure is going, I see the entire company stacked, moving on to stablecoin in the next two, three years. And I think it's actually going to make it a much more efficient company that we can operate in. And then the last one, which we've been dabbling with right now is basically thinking of programmable money. Right. So you have obviously the agents that are going to be doing purchasing from mcps, clis, et cetera. But then how do you make that something that's very usable for these different companies that we service? And so today we do have two agents that we sell. They have a lot more, but those are the ones that resonate the most. One is a reconciliation agent. So it basically looks at the purchases you have and then helps you kind of match them one to one. And then the other one is a GL coding agent. So if you think about enterprises, a Big, big function is literally someone's job is to look at every transaction and put a code on it so your ERP can read it. And we trained an agent to be able to do that with 99% accuracy or better. One of the huge advantages we have is the companies that we sell to look like Jeeps. We are operating in 12 countries. We have 60 plus bank accounts. We have a sophisticated finance. We operate in latam. So if our finance team can use it and they find that it's valuable, I'm pretty confident I can sell that product. Right. So that's a really huge advantage that we have. And as I look at the next two to three, four, five years, I see this convergence of stablecoin, I see the convergence of agents. Money is going to become programmable, and hopefully we're at that nexus of all of this happening.
- [27:42]
Angela Strange
Amazing. All right, I want to wrap maybe with a couple more personal questions about you that could be helpful. I think there's going to be more global businesses built. I think huge opportunities there are very challenging. You've talked pretty publicly about how you need to almost reinvent your superpower every six to eight months. What do you think your superpower was when you started, and how is that evolving?
- [28:03]
Dilip Tasman
Yeah, I think a lot of founders try to do a lot of different things when they start and when you start. The way I see it is your job is to focus on one or two maximum metrics that matter. And then you need to be like a monkey with a stick, which is like, how do you move that metric? And everything else doesn't matter. And then once those metrics are moving, then you figure out other things to do. So I'll give you an example. When we started, our number one metric is spend. It's not revenue. It's not necessarily, does it have margin? It's, can I get someone to use this card in Latin America? I don't know. I did not grow up in Latin America. Actually, when I first flew to Mexico City, it was kind of dawning on me. I'm like, I'm going to launch this product Mexico. And I've. That was the second time I've been.
- [28:43]
Angela Strange
Spanish is a little rusty,
- [28:47]
Dilip Tasman
but that was the focus. It was just focused on, can we get people to use the cards? Started there. We were clearly getting acceleration on the spend. And then we figured out all the other things. Revenue, we figured out margin, et cetera. And I think a lot of times founders try to do everything in the beginning. And in the beginning, you need to just focus on what matters to get you from point A to point B. And then the second part, which is where I think, think most founders need to really focus on, is you have to be 70% good at almost every function, and you just have to rotate. And that's your job. Today your job is product. Today I spend a lot of time on product engineering. It's just so, so important. With stablecoin, two years ago was finance, because we were navigating 23 and interest rates had shot up 600% and that, like every day I was on these calls, and that was a job. And now we have a amazing cfo, amazing finance team. So now I'm like, I don't need to spend that much time, but as a CEO, as a founder, your job is to do whatever job is needed for the company to get to the next stage. And I think a lot of times founders are like, oh, I can't do sales. It's like, nobody can do sales. You just, like, you learn what you need to do and then you hire someone better than you. That's also a skill. But you have to be dangerous enough to know what you're looking at. Right. And I think that's now even more easier with AI. And so if people can't do that now, I don't know what to say.
- [29:59]
Angela Strange
Yep. Even if you don't have an AI thesis, you should be able to leave.
- [30:02]
Dilip Tasman
It's amazing. Yeah.
- [30:04]
Angela Strange
And the media. I think the other biggest challenge, building a company anywhere is talent and how you think about talent. What is Jeeves philosophy, especially being in the number of countries that you're in.
- [30:15]
Dilip Tasman
Yeah. So we need local talent to succeed. And that's like the first hire we do in any geo. Actually in Argentina, we just hired this person is what we call kind of a local gm. They have to be local, they have to know the market, they have to live there, because I'm not the one selling on the ground. So that's always the first hire. What we look for, we have this internal, you know, kind of like bar, which is. We call it relentlessness. Like, we. There's no real playbook for what we're doing. Like, there's different types of playbooks, but no one's launching this in 25 countries with multiple product stacks. And everyone's like, are you guys doing too much? And are you guys doing too little? But that's what we do, that's what we offer. And so you need to have this just very, very high motor that is constantly just going. The good thing with a company like geese is you'll know within probably a week if this is the right fit because it's either going to be like, holy shit, I'm learning so much, or it's going to be like, like, I don't know what's happening. I feel like I'm drowning every dent. And a lot of people, you know, like, cycle out fairly quickly, which I think is fine. Like, it might not be the right fit. So we really push that engine part of, like, you have to just have this high, high bar and keep, keep going. And then the second part here is one thing we've been able to do is bring in good seasoned leaders for like the second cycle of Juice. Right. So the first cycle is a zero to one cycle. You know, this year we should be crossing hopefully 100 million in top line. ARR. And we need a second set of leaders. And I think we've done a very good job at least so far of getting those seasoned leaders. We brought in a CRO from Brex, we brought in a risk officer from Capital One that joined us last year. Our CFO took payoneer public. Getting those people to join has really upped the talent bar in the company as well.
- [31:51]
Angela Strange
Very focused on stablecoins. You've also been chief AI officer of Jeeves. I think it is a huge advantage if you started pre AI, you have the distribution and then you can use that to really drive it forward. I think the challenge though is getting the company to be AI native.
- [32:10]
Dilip Tasman
Yeah.
- [32:11]
Angela Strange
So maybe talk through, how do you make that happen and how does AI benefit?
- [32:14]
Dilip Tasman
It's a good question. So today we are at 140 people. We were at 200 people two years ago and our revenue has grown 10x, our volume has grown 8x. So clearly it's accelerating and it's accelerating with less people. This just wouldn't be possible without AI. So I'll walk through some concrete examples where we use it. So one which I think is really important is we use it on the underwriting side. We do pretty serious underwriting for any company that comes to us because just like a Ramp or Brax, we have to give them 30 days of spend and then at the end of 30 days, they have to pay us back. So we need to know that they're good to pay us back on day 31. And today we have a team of four people that does this underwriting across all these regions. And that just simply wouldn't have been possible even two years ago. You'd need 15 people just to get the underwriting Going better models, self learning models, self training models. We have a lot of data and now the models are actually keeping up with the data. So that's a huge unlock. Same thing when you think about customer service, we again have maybe three people on the customer service team and we have to handle different languages, we have to handle people that are obviously talking about payments, but then also cards and also ap. We just are starting to test an AR product and it's a very small team, would not be possible without training what we have on the customer service side to be able to handle that. So I think that's another area. And then if you think about KYB and document ingestion, that's an area that we had to build our own version of an OCR machine because again, even if you can do Spanish and Portuguese, there are little nuances. Like in Brazil, the commas and the dots are intertwined. So if you look at something and it says $2,000, it's actually $200,000 because the comma and the dot is switched. And it took us a while to train our models to be able to do that. But just like your question on the infrastructure, we want to own this core infrastructure ourselves. I think without that, we will not be able to succeed. And that to me falls on the founder and the CEO. If you are not AI pilled, you're not going to make it. And I think part of what I've been pushing internally is just, look, we are pre AI and we are trying to make that jump. If you think you're fast at 3x, there are native AI companies that are moving at 10x, so you're still slow and you need to be 20x just to keep up with the 10x. And so just making that pace. But if you don't have that pace yourself and you don't set that clock, it's not going to happen. You can't outsource this. You can't hire a chief AI person. Like, you can hire that. And there are many very smart people. But that energy, that pace has to come from the CEO. Thankfully, I have a little bit of a technical background and so I'm one of those maybe annoyingly dangerous CEOs where I'm like, oh, I'm good enough to tell you what to do, but then you can go figure out what to do. But I feel so unblocked as a X engineer now with AI because I feel like I can actually start building. I mean, another example, as soon as openclaw came out, like I built the first Openclaw for Jeeves. Not because I'm like some genius, but if I don't do it, everyone's gonna say, hey, what's the security? Et cetera. And the real security? Things you need to figure out. But if you don't show that you can build an openclaw and someone can tag it in a WhatsApp channel and it's gonna give you jokes or whatever you wanna do, then people just don't feel like they can do it themselves. And so it's just so important. And everybody says the same thing. You don't have enough time, et cetera. You have to make the time. Like, if it's important for the company to succeed, you need AI. And if it's important to do AI, you need to do it. If you need to do it, you need to figure out the time. And so that's how I've kind of thought about it. And you know, candidly it helps that it's actually very exciting building an AI. It doesn't feel like a chore. Like, I wish I had more time because there's just so much I can do. And every week it feels like Anthropic is releasing something else and I'm like, give me one day so I can like catch up to this.
- [35:39]
Angela Strange
Amazing. Well, thank you very much for joining us. I'm sure the great enterprises of Peru will be very happy, very happy to hear you can now economically launch there and appreciate you coming on the plot.
- [35:50]
Dilip Tasman
Thanks for having me. This is fantastic. And yeah, if there are companies out there that, you know, are looking to build that stablecoin and AI, we'd love for them to try us and give us feedback. Send an email directly to me and you know, I'll respond probably pretty quickly.
- [36:02]
Podcast Host / Narrator
Thanks for listening to this episode of the A6TZ 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 go to YouTube, Apple Podcasts and Spotify. Follow us on X16Z and subscribe to our substack@A16Z substack.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 disclosures.