← Fintech Leaders
Fintech Leaders
How Addi Reached $200M ARR with Better Margins Than Nubank, Affirm, and Klarna
00:43:15
Loading summary
Transcript42 lines
- [00:00]
A
We treat every board meeting as a dirty laundry event. So we always say this is not working. This we need help in. This is bad. We're now clocking over $1.3 billion of GMV annualized. And on revenue we're about right around the 200 million ARR mark. We have over 55% gross margins. Nubank doesn't have those margins. Affirm doesn't have those margins. Klarna doesn't have those margins. And these are fully loaded provision margins. By the time I cross about 5 or 600,000 customers, merchants that I had been calling for years start calling top 10 merchants in the country, they called us and said, I'm tired of having customers ask me about your product. When can we integrate?
- [00:40]
B
Welcome to Fintech Leaders. I'm Miguel Almasa and I'm fascinated by the company's reshaping financial services. Which is why I co founded Gilgamesh Ventures, a fund where we have backed almost 50 fintech companies worldwide. And over the last five years I've recorded conversations with over 3, 350 of the most thoughtful founders and investors in Fintech to understand how they think, what they have learned and extract insights that matter to anyone building in fintech. Today I sit down with Santiago Suarez, CEO and co founder of Addie, Colombia's leading fintech and commerce platform serving two and a half million customers and over 30,000 merchants. They are backed by Union Square Ventures, Monashees, Andreessen Horowitz, GIC and Graycroft Santi. Welcome to Fintech Leaders. Always a pleasure talking to you. How's it going today in Colombia?
- [01:39]
A
Thank you, Miguel. Delighted to be here. It's been wonderful here in Pereira and thrilled to be back. I think I was here a few years ago and then we just have been under a rock for the last three or so years. So I'm thrilled to be back talking to you.
- [01:56]
B
Yeah, and that's exactly what we need to talk about. Maybe we need to lift the rock a little bit and hear about what ADI has been up to. I want to start with a specific question though. I know some of your board members and I talked to some of them in preparation for this call and something that they consistently said was you are very good at extracting value from your board and thus your investors. Maybe share your approach of your relationship with your board and your investors.
- [02:32]
A
We see our board as two things. Number one, they are incredible network connectors. We're lucky enough to be venture funded. We have some of the world's best investors on our board. Andreessen Horowitz DIC Union Square so I start from the premise that these people know the best at whatever it is we need. And in fact, one of the values we have at the company is we obsess with and hone our craft. And the example I give people all the time is do you know who's best in the world at your job? Who is the best CMO in the world? Oh, you don't know? Okay, find that person. Great. Now we're going to get you access to that person who is the best head of lending in the world. Okay, great. And we got to model yourself. Not even the best in latam or the best in the Omega. Who is best in the world? Sometimes best in the world is in India, in Kazakhstan, Indonesia, sometimes in the America. So the first thing is we, we, we, we have that mindset and then we go to the board, we do that. The second thing, very tactical thing we did is we actually increased the frequency of our board meetings. So in 2022, when kind of like the whole money went away and I was like, oh God, we're going to be in for a world of pain. I reached out to the board and I said, guys, I want to meet monthly. And the reason I want to meet monthly is that I want to make sure every month, 30 minutes. It's optional. So if you can't join one month, don't worry, you'll get the email. We give you the metrics. We give you the metrics, give you the metrics. It's just a two pager, maybe three pager. It's the same thing. And a good discipline for us, for my co founder and I, because we got to sit down and say, what happened the last 30 days takes about two hours to put together. And we meet and we say, hey, this is what happened last 30 days. But that means that when we meet on a quarterly basis, we don't have to catch you up on the business. You already know the business. So we can go deep on a certain topic. So it could be customer acquisition, it could be product, it could be capital structure. And then the third thing we've done is we've started to specialize folks in different things. So we brought a phenomenal independent. So we have Ian Forrester, who's a senior executive at Capital One, he's our independent, kind of informally chairs our risk committee. We then start to ask different people different things. So we got John at Union Square, he's from the board of Upgrade, was on the board of Lending Club when they went public so the guy is phenomenal. Cap markets, lending.
- [04:52]
B
Question.
- [04:52]
A
We got Chris at gic, largest shareholder in a firm for outside Max Lovechin. He knows everything about bnpl. So we're able to start going to each of them and being a lot more targeted in our queries. Those are the things we've done. The other thing we've done, it's more common in the US than a lot of times is we treat the board as true insiders. So one of the pieces of advice I got when I started this company from a very well known Latin American entrepreneur was treat every board meeting as a fundraising event. So go give your story. And obviously when you're fundraising, you're not lying, far from it. But you're telling the best story possible. It's like every board meeting is fundraising event. We treat every board meeting as a dirty laundry event. So we go in and we say, this is not working, this is not working, this is not working. This, we need help in. This is bad, this is that. And sometimes we don't even celebrate. It's like, you know, we turn profitable for now, four quarters. We probably discussed it for five minutes at the border level. We're like, oh, okay, great, it's profitable. We knew that because we've been talking to you every month. Thanks. What's not working? So we treat them as true insiders. And that has two benefits. One, you get better advice, but secondly, much higher trust because. And this feedback I got from our board members, this person said, I don't go to your board meetings wondering, what is he not saying? I go in being like, okay, this is what's happening. And that changes the dynamic in the meeting.
- [06:24]
B
I'm going to send that clip to a bunch of our founders. Sati. Another thing that I understand you do very well is the power of saying no, right? I mean you, you've said no to markets. You, you, you say no to things every day. Talk about your philosophy of, of saying no.
- [06:45]
A
If you ask half my management team, they might say he still says yes too much. So I think that's the, I would say the core philosophy of say, you know, is probably a handful. The first thing is three years ago we made the decision to move away from objectives to North Star. Actually, four years ago we were always talking about OKRs, you know, start company. You have OKRs, the objectives, objectives for the quarter. We now have a single North Star. For two years, the North Star was profitability from August 2022 until August of 2024. That was profitability, making the company profitable. That was a single North Star. The second thing you want to think about when we think about this is you want to have an extremely contentious and I say that with love management team because the things we do with our management team that allows us to focus is we have some extremely high power folks. So you know our head of lending, Jayesh, 25 year veteran of the industry, 10 years up cap one, he actually built Paytm's BNPL business from 0 to 25 million accounts in India. So he's been around the block, worked at Dynamics for a while. Manish who just joined. Manish just joined from Amazon. He spent 15 years at Amazon, led payments globally for Amazon, four years at block. These guys, I joke, I will say guys, you always tell me no, all I do is lobby you. But that's actually what you want in a management team. You want the founder to be setting the vision, pushing and you want really grounded, high execution folks. You want people to actually get stuff done. But we're always pruning and always pruning, always pruning. And then the last thing around the power of saying no is but we get together twice a year and we say what are we doing, what are we not doing? And we plan on half year cycles. So that also helps because you're not replanning that often. But that being said, I don't think there's such thing as too much focus. Even though we're more focused than I would say most, I would even say, boy, we could even prune here, we could even prune there. But again, Northstar management team, high trust, high integrity dynamics, not only in the boardroom as we discussed, but also within the management team.
- [08:55]
B
How do you convince these world class professionals to join a company in Colombia?
- [09:03]
A
Well, it's different because when Jaish joined, boy, I mean, and I hope Jaish listens to this. I don't know. I mean with Jaish, we convinced him in a very different time, right? It was 2022. We were just going into the nuclear winter of Latam Fintech. But what Yaesh would tell you that convinced him was that we had incredible customer adoption over 70 NPS, very good traction, very good customer adoption. And he was like, the economics may be wrong but if you have customers, you have great adoption. I can fix the economics and then we're going to go to the moon. And in many ways what Daesh has done is addressing some of the economics challenges we had back in 22 with Manish and the people we have brought in over the last year. It's a Different story because then you can say, hey, I've built an incredible platform. This is a financial and commerce platform with probably the best technology stack in the region. Best margins, we have over 55% gross margins. Nubank doesn't have those margins. Affirm doesn't have those margins. Klarna doesn't have those margins. And these are fully loaded provision margins. Is that accounting? Gross margins of over 55% for a lender are unheard of. Five years of 100 plus percent year over year growth, an increasing product suite growth in customers of over 50% in merchants over 100%. So if you are these guys, in many ways they have seen the movie before. So Manish has seen the movie at Amazon, Fer saw the movie at ebay. He took GMV in emerging markets from literally $0 to billions of dollars. So they look at it, they're like you're just getting started. And not only that, they can see value in what they bring to the table. So why are we excited about having someone like Manisha at the table? Because we're launching new products every month. So now we have a world class product. Why are we excited about having Farah at the table? Because we have historically invested very little in marketing. Our payback periods are two months. Our customers pay back in two months. Now I used to say that I'd be like oh look at this, this is the best business ever. Now I say that I'm like I'm under spending in marketing. My payback should be six months, maybe 12 months. But that requires the right leader. So you go to fair and you say here's the raw material, here's the check. So I got the money. We're going to go build and scale this. So I think that's what gets them excited. And then obviously, I mean our culture, it's high integrity. They know that I'm not going to as as we were just talking about. They know that I trust them. They know that I want them to tell me where I'm wrong. So I also think we built a very unique culture with a very editorial way of working. And if you like that way of working, you're going to be extremely happ here.
- [11:38]
B
Sa I think that's a good place to talk about your vision because a few years ago you were planning to build a, at a minimum a regional business and you had a big operation in Brazil. Obviously you said no to Brazil. Now you're doubling down on Colombia. But, but the vision has, has also changed you. You've been inspired by some other Companies especially coming from, from Eurasia, I guess. And I kind of know the answer, but I want to hear it from you.
- [12:14]
A
Yeah. What we realized, and this was also 20, 22, early 23, so you have to understand, 22, no one knew Caspi. Caspi is this Kazakh based public list company. They were not listed on nasdaq. And I have one of my favorite investors. These guys are based out in Eurasia and they're always, they're like two or three years ahead of the track. They've been ahead on quick commerce, they've been ahead on things. And they called me up, I remember they said, this is it. I think we figured it out. I said, okay, great. What is it? Here you go. This, this company, blah, blah, blah, like, awesome. We, we learned a little bit and we realized we had the same raw materials. But to the vision. What is our vision? We want to become the financial and commerce platform for Colombians, for everyday commerce, and eventually for other countries. For the first time, we're starting to think about where would we take this ecosystem? So it's no longer about being a thin layer of I'm just a BNPL provider. No, it's about, I have some consumers, I've got some merchants. How do I get them to come together every single day for all their needs? Because guess what, they have a lot of similar needs. You got to find each other, you got to have the right credit solutions, you got to have the right offer, you got to have the right spend. So we wake up every day thinking, how do we make commerce and financial services happen more seamlessly? So we have evolved from just having a BNPL solution to something that has credit, instantaneous payments, working capital solutions for our merchants, a full fledged marketplace that allows them to sell to the customers directly. On our app, we're about to launch our savings account once we obtain final regulatory approval. So it really becomes about using that ecosystem to deliver better service, a lower cost, with a better experience to both our consumers and our merchants.
- [14:03]
B
Yeah, my episode with the CEO of Caspi, Michael, is still the number one episode I have recorded by far.
- [14:11]
A
So you know that, you know what happened. So last year in 24. And I think I, I tried to get to him in many ways. I couldn't. So I LinkedIn him. And if Misha is hearing this, he'll. He'll attest. So I LinkedIn him like four times. Like cold LinkedIn messages being like, I will be in Almaty. I'd love to meet with you. I'll be in Almaty. I'd love to meet with you. And after the fourth message I got an email from someone be like okay fine, you can come tomorrow at 6pm and we're already in Uzbekistan because we had done. We were doing a central Asia tour because in Uzbekistan there's a lot of fintech innovation. So we were supposed to be in on monthly. Two days after that we moved the flight forward and he was incredibly generous and he sat down with us for about 90 minutes and he just said what questions do you have? I was with my co founder and we just went through some questions. He was very generous sharing about his experiences. It was extremely inspiring.
- [15:04]
B
How do you build the dominant platform not just for financial services but e commerce maybe you know, other services beyond payments financing and, and commerce. Right. How do you build it in a country that is. You're not really leapfrogging from. From nothing. Right. Colombia is a, is a very, very advanced market in many ways. You have E commerce players, you have more FinTechs than just Addi. So we what's the key to achieve that dominance in a market that is not a desert?
- [15:43]
A
It's understanding that while it's not a desert, it's also not this cornucopia of offerings. And let me be very precise if you think about Colombia. Colombia first of all for those of you who don't know, it's extremely big as a market. 400 billion plus GDP, over 50 million people. So it's a huge market. Third largest country in Latam by population now it has more smartphones per capita than Mexico and Brazil. So if you think about smartphones per capita we are winning, we are leading. But over 65% of transactions are cash and less than 20% of people have a credit card. So it's actually this paradoxical thing where you have all the technology layers that you need to be able to reach with financial services. But financial services or fintech still extremely antiquated. Again 66% of all payments are cash. Less than 20% of all people have a credit card. There are reasons for this oligopoly. 70% of all assets in banks. 3 top 3 largest banks. The use recap is miserable. If you want to launch a credit card in Colombia it is likely you will fail. There hasn't been a single successful credit card launch outside of mainstream bank in the last what, 15, 20 years because the user cap is 25% a year. A year as much as you'll charge in Mexico or Brazil a month for some of your segments you have to settle the next day. Interchange is Very long. So we, we took a very different approach. We said, okay, you got a technologically sophisticated population. We cannot do a credit card. We don't want to do a credit card. We're actually double down in learning about the merchants and the consumers. First of all, because we have the whole network, you have to remember we're both the acquirer, the network, the processor, even the marketplace, and obviously the issuer. We have a lot of data. So that allows you to approve a lot of customers that other people can't approve. 40% of our customers don't have a credit card. So if they want to spend like you would spend on your credit card, they spend on our platform. So that right there was a leapfrog because we can risk base price for our premium customers, of which we have about half a mil. We can give them 0%. No credit card does that. In Colombia, there's no such thing as parcel house and judos or true zero percent. We, we are the zero percent without the ances, without any. So then they come to us as well. And then you have a very nice flywheel because the more consumers you have that can only spend on you, the more merchants want to join your network, the more merchants join your network, the more consumers want to shop on your network. You now have better merchants. We have over 60,000 physical stores countrywide. Columbia has a 1100 cities and we have presence in 1045 of them. We have physical Addy stores in the Amazon, on the islands, in the Caribbean, on the Pacific coast. Over a thousand cities have an Addy place. And by the way, they're all branded. So you can see our little brand. So they're all branded. So that also leads to the customer say, oh, I can shop with Addy here we have customers say, oh, if it doesn't have Addy, I'll go somewhere else. So then that's what allows you to then say, let me layer more products. Why did we start a marketplace? Because we had a search for in our app and the search for was for you to find a merchant or how to pay my next installment. And we looked at the search queries. 90% of them were SKU queries. Nike shoe size 35 in blue, black Adidas T shirt. So then we took that data to our merchants and we said, hey, I've got all these people who want to buy your stuff. They have money to spend because I have given them money to spend. Wouldn't you like to just offer your things on our app at a phenomenal cost for you, Mr. Merchant? Or Ms. Merchant because we can give you much better pricing than any other marketplace. And that's the other thing. Because we're the only game in town. We can distribute the savings to the consumers, distribute the savings to the merchants. So we're also usually the best cost solution for both sides.
- [19:43]
B
I think you mentioned Colombia, an economy of gdp of about 400 billion. Kazakhstan is 200 billion and it's a.
- [19:51]
A
$20 billion public listed company.
- [19:53]
B
And I think last year net revenue, net profitability was like 2 billion. Right.
- [19:59]
A
Of income. Of net income. Net income.
- [20:03]
B
Net income in cash. Right. And so clearly you can build a massive in Colombia, maybe share a little bit about your scale. You mentioned the number of cities you're in. But maybe whatever you can share in terms of customers, anything like that.
- [20:19]
A
Yeah. So customer wise we're now running around 2.5 million active customers. We have over 30,000 merchant partners by the way, we're onboarding 3,000 merchants a month. It's extremely, extremely powerful. As I said, self reinforcing flywheel GNV were now clocking over $1.3 billion of GMV annualized. And on revenue we're about right around the 200 million-ARR mark. And as we mentioned with phenomenal margins, over 50% gross margins, we've been profitable for a year and we've been GAAP profitable or in our case IFRS profitable for all of 2025. And that's by the way the number of the auto assignments. Right. So that includes stock and da da da and all these things. So there's still a lot of work. But it's a big business and it's a business that has consistently grown over 100% since 2020. We will grow at over 100% this year. Again being very disciplined with marketing investment, being very disciplined with headcount. If you look at our non call center headcount, we made the decision now to outsource the call center for a number of cultural reasons. We can talk about them but if you just look at the people who are not call center operators, it's about 260 of them. So if you do the math like circa 200 mil of revenue run rate, 260 folks. That's the company level revenue productivity per FTE. So that's, that's a little of the scale. I think we discussed the beginning. We've been under a rock for the last three years and I don't think people realize that with this Caspi model or Ozum in Uzbekistan, TBC in Georgia you can Just go extremely deep and massive businesses and then you take that eco system somewhere else. You know, CASB is just doing that in Turkey and they did it the right way. Bank marketplace like there's a certain natural affinity to these things that we think we'll be able to do at some point, but self funded. So that's the other thing. To our, to our point, our board, I said guys, it's not about if, it's when, but it has to be when. When we fund it with free cash flow, we're not going to dilute ourselves to expand internationally. We will fund it with free cash. I think we're getting to a point where we will generate enough free cash. So that is now a reality.
- [22:35]
B
Whenever I am talking to the leader of a company who's already reaching meaningful scale like yours, I'm interested in the structure that you've designed around your team. Right. How many direct reports do you have? What lessons have you learned about the orgs that are most important? Maybe share a little bit of that philosophy.
- [22:59]
A
Yeah. So there are a few tried and tested things that have worked for us and there are a few things that we kind of mix and match. So the things that have worked for us, first and foremost, writing. We are a writing culture in part because we're remote first but we've been writing since the beginning of time. So we're a writing culture. We send memos on absolutely everything. So it's memos and Google Sheets that's the currency of the realm here. Members and Google Sheets or for particularly fancy thing you can paste the databricks link or whatever. But generally speaking I may want a Google sheet will do the job just fine. We used to pay hundreds of thousands of dollars for BI tools. The best VI tool in many cases is Google Sheets and it's Linked or Data Warehouse and all of that. But it's just like easy to pull. So that's one thing. So memos and Google Sheets, it's a constant reading in advance. Everyone at ADY reads and comments in advance. And then the other tried and tested thing is obsession over epistemology. And this has been the biggest lesson because I joke that at ADI you will come in and you're like a world expert on anything and your first statement will be, I believe tomorrow the sun will rise from the east. And what you will hear from all of us is why show your work? And you're probably right, that sun will rise from the east. But we want you to show your work. We want you to be able to articulate the why? What are your core assumptions? How do they move? And why do they matter? So those were what I'm going to say, the tried and tested pieces. And there's a very important impact when you have someone, when you have a writing culture, a culture that's obsessed over epistemology. No one has any problems at poking holes and generating constructive conflict to get to the right answer. So decisions that we've made, like exit in Brazil, at some point we would give everyone zero percent. We gave everyone zero percent because that was the afterpay thing to do or the whatever. At some point, like one of our guys was like, that's a dumb idea. So we kill the sacred cow, we kill the golden goose. What are children? We just. This was an untouchable principle. It's where BNPL company everyone needs to get 0%. One of our guys was like, no, why would you do that? We went ahead and did that. Now more tactical aspects that have changed over the years. I've played with more direct reports, fewer direct reports. I probably have more now as we try to do delayer again. It's a small org, but I have what, 11 direct reports. So we don't have one on ones. We do a lot of one on ones when you are being on boarded. So when you're on boarded, we go extremely deep and do daily one on ones. But once you're on boarded, it's like, you got my WhatsApp, you have my slack handle, find me. And what that means, by the way, is you end up in more natural rhythms. So there will be times where I'll be speaking to one of my directs twice a day and we'll be like, boom, boom, boom. And there will be times where I'll be speaking to them, I don't know, once every 15 days. And maybe the last thing is, we've finally figured out a fairly effective and robust way of tracking the execution of the company. So we have a weekly business review. It's awesome. It works really well. It's on Tuesdays. It sets the cadence of the company. People show up. It's a whole ritual, like, you know, Monday, our chief of staff ahead of FP and a choose what topics they want to cover. You get, you get told if you're one of the topics, you got to send a write up. It's only 200 words. We go down to that. It's 200 words or fewer. So you don't send tons of things. It's a very specific format. You send the video. So you record yourself, explain this on the loom. We all watch the loom in advance and many times we're like, yep, Miguel, that makes sense. Thanks for the loom. We don't even need to cover it. So it's, it's extremely effective. 90 minutes. Most often we don't even cover the 90 minutes by minute 60. We're like, okay, this was great. Make it happen. So those are some of the things that, that have worked for us. And of course the North Star. One North Star.
- [26:48]
B
Why'd you move customer service externally? Because I've heard other companies actually say, no, I want this to be internal.
- [26:54]
A
Oh no, that's what I mean. We always have it internal. We have an internal and we have a little bit external for peaks. But we've all. But when we talk about our full headcount, it's like 590. But that includes a lot of customer service folks. But if you look at any company of our scale, revenue wise, they tend to have at least half of their customer service operation outside. We've always kept it in house. We give them equity. I think we're one of the few companies that give equity to our customer service agents. We give them incredible benefits. We have two full time English teachers on staff. The company runs in English. So we want to make sure that our customer service agents kind of like progress and move to other parts of the company. You don't have to be fluent in English to work in customer service, but if you want to move anywhere else, you have to be fluent in English. So we have two full time teachers. They move. And then of course, I mean this has become a little more academic now that we have been able to deploy some pretty powerful AI tools. So you know, we now have our AI powered agent called ADRI. ADRE processes 70% of all customer interactions with a 78, give or take CSAT. So we've been able to basically keep our customer service headcount flat. Even though our volumes have doubled, our customers have increased by about 50% and our merchants have close to tripled in that period.
- [28:12]
B
Yeah, it's actually going to be my next question. Right. Because customer service is one of the first places where LLMs have made an impact. Now I understand that at Addie the impact is being felt across the company. Right. And you know, we talked a little bit this early. It ties back to your architecture. Right. How you're able to deploy these tools. Let's talk about that because that might be one of the most interesting things that you're doing.
- [28:44]
A
Yeah. So when we started the company this is back in 2018. We hired Carlos, our CTO. He was our second hire. He remains our CTO. He's awesome. And Carlos basically said, we're going to do two things. We're going to have a mono repo, which means we only have a single code repository for the entire company. Google works that way. For example, Anthropic works that way. A lot of companies don't work that way. They go full microservices. So they have different code repositories. That gains speed, and it does, but you lose standards and critically, you lose the ability to have everything in the same place, which turns out to be extremely valuable in the age of AI. So first decision we made monorepo. Now there are ways you can make the monorepo work. I mean, Google's a fairly efficient company in some ways. So it's not like everyone said, oh, you do a monolith. It's like, no, it's a monorepo. There are a number of things. The second decision we made was an event based architecture. What that means is every time something happens in this platform, we log an event. And an event can be something as basic as Miguel applied. But because of the business model I have, I have so much data. Because if you're an issuer, if you're a credit card issuer, you have data for your credit history. You may have data on your income, but you have no data on where you are when you physically buy things. As you're just swiping, you have no data on what device you're using. You obviously have none of the merchant data because that's all the acquirer's job. And you obviously don't have ecom data. I have the whole thing. So I've got all the data that an issuer has, I've got all the network data. I know when the network moves, I know what are fraudulent signals, I know what are good credit signals. For about 30% of my transactions, I even get SKU level data. So not only do I know where you're buying, I know that you're buying a pair of white shoes, size 37. It's pretty awesome. And then I also get commerce data. Your neighbors, what are they buying? What are you buying? So all of that gets logged and has been logged since the beginning of time on these events. We have over 10 billion events to date. And then three plus years ago we said, okay, events are great, but they're in this thing called an event log. They're impossible to query. And we started to see some Issues. We're like, we have this data, we can't quite use it. And we couldn't use it even in basic ways like what was today's dmv? And we're like, I don't know, wait until tomorrow. No, no, no, no. So Carlos embarked upon this at the end of 21 pre LLMs, pre ChatGPT. I said, we're going to fix this for once and for all. So we built a pretty powerful data ingestion layer, takes all these events and processes them into queryable tables and data. That's close to 50 terabytes of data. And not only that, he built the infrastructure. And if he were here, he would talk to you about data contracts and why they are so powerful. But for the purpose of this, he built an infrastructure so that anytime we have a new product, by default that goes into that data. So what we have today is close to 50 terabytes of data that no one has in the country. Data from merchants from E. Com, from network, from devices, from the tax authority, all there, and we can query. And that powers over 50 AI agents. Agents that do everything. They do everything from figure out what's your next purchase and show that on your marketplace. Agents that respond to customer complaints. So by Colombian law, certain type of customer complaints, if you say pqr, it has to be responded in a legal way. We got a legal agent, we've got an agent that takes all this data to create features for our credit and fraud models. We have five models involved in every single transaction. Every time a customer does a transaction, we have over five models that are looking at the transaction multiple ways turning out. Is it fraudulent? Is it not? Do we approve it? Do we not? One of my favorite stories, by the way of that is at a job candidate and she applied with her Android device and we turned her down and she applied with her iPhone and we approved her. And I was like, of course, because you're like a fancy lady living in the fancy part of Bogota. And for reasons related to her job, she had two devices. A very cheap Android phone and a very fancy iPhone. And our model was like, this is funny, you shouldn't have a cheap Android device. That's just really weird. But then when she applied with her iPhone, we're like, oh, of course. The whole thing makes sense. So we approved her. So you have all these data, extremely clearable, and everyone has access to it. Our data analysts, our data scientists, our machine learning specialists, but critically, our AI agents. So when you load an AI agent that is like an LLM Powered AI, truly autonomous AI agent. They have access to all the data they need everything. And it's so easy to train them, it's so easy to do reinforcement learning, it's so easy to make them accurate. I say easy and of course my team is probably thinking yeah, that's what he thinks. But all things considered and you know, for a company that has go 590 people, we estimate we have saved over 50 heads so far this year and increasing by day thanks to our AI implementations.
- [33:38]
B
So when you, I mean you're talking about the importance of data and tag and being able to tag your data correctly, the importance of your infrastructure. Right. What is your reaction when you read of this study? I think it was by MIT that 95% of AI projects inside companies have failed. To me that sounds more like a problem of the data of these companies on the infrastructure. But what's your reaction?
- [34:09]
A
My reaction is a bit nuanced in part because my wife until very recently ran a massive AI initiative and one of her big things was always you can deploy AI extremely quickly even without having to clean your data. And that is true. So I would say in some cases true, but in some other cases, when you look at, let's not name names, but like these big AI vendors that are emerging that are sitting between you and anthropic Gemini and OpenAI, what they do is good old fashioned systems integration. Build your rag pipelines, build your cajole the right data sources, ensure they're being updated so that your LLM hopefully is not an idiot when it speaks to your customers. So I actually think that process of grabbing the right data, exposing it and using it to do reinforcement learning is very important. And by the way, we're very deliberate on this because we started this big project even before oms. We just knew we needed what I'm going to call a resilient, scalable, durable data infrastructure that we could always use. I mean data is so critical to our business. We use it to recommend, to underwrite, to approve merchants. We underwrite both merchants and consumers, the whole thing. But when LLM started, I remember I said okay guys, what are we going to do? And you know, you had all these people putting out press releases and eventually I'm like, guys, it seems like everyone's putting these press releases. And I remember Mauro, my head of data said we're going to start with legal. I'm like, why would we start with legal? That feels so low hanging like. I'm like, no, that's not the issue. Why do we start with customer service. And he said, no, no, you need to build the right. Need to build the right pipelines. And turns out that the legal use case was fantastic because the legal use case was responding to what we call an escalated complaint, which is when someone is basically saying, I'm going to go to the Consumer Protection Authority or I may sue you or whatever. For a company that processes millions of transactions a month, you get a lot of those. But for you to respond to those things accurately, you need to have phenomenal data. You need to know exactly where the customer file is. What was the situation? When did the approval happen? What loan are we talking about? What merchant? What is the chat history with our customer agents? What is the history with the merchant? So it turns out responding to an escalated complaint is a fantastic way of building the right pipelines to then have all your data easily accessible. And once we did that, we saved a few lawyers. So, awesome. Like, you know, otherwise we would have had to linearly scale the number of lawyers responding to complaints. But then we were able to say, now we can turn this around for customer service. And we went from like, 20 to, again, close to 70% of customer queries handled by Adri from January to August of this year.
- [36:55]
B
Have you cut the plug on certain SaaS applications to build in house, or is that not a thing?
- [37:02]
A
Yes, but we can't talk about it.
- [37:05]
B
Fair enough, fair enough.
- [37:07]
A
But the cool thing is, and this is one of the boldest decisions I think Carlos has led, we're doing a very principal way. So what I mean by that is we kind of sat back and we said, what is the customer experience of the future? And then what is it that you need to be able to power it? And what can we do in house in a better way? And that led to some surprising conclusions, let's just say. So I'm pretty excited to see where. Where that goes.
- [37:31]
B
Fundraising. There was a period where I would see you in the news every. Every couple months, and every time it was a bigger and bigger round thing. You had three rounds in nine months. Was that a mistake or was that the right call?
- [37:47]
A
I was definitely the right call. There's no way I'd be here had we not raised all that money. There's no way we survived. And by the way, we've seen. We've seen the whole movie, right? We raised like crazy people back in the Go go years. And then we did an internal round in 2023 that was lower valuation than the previous fancy round. And we took our pain. We decided it's more insider baseball but we decided to take much less structure. Clean, as clean as possible, but a lower price. I'm like, who cares? Like we just want to make sure we do right, right by the business. So it wasn't a mistake because I don't think we'd be here had we not raised all that money. By the way, going back to your question about hey, Colombia is not quite the desert, but those venture dollars paid for a lot of customers coming in that 0% most by the way, about half of Those customers still 0% customers and they love the product. I think in hindsight what I would even say is we may turn out to be and jury's still out, but things are heading all in the right direction. We may turn out to be one of those examples where poaching a ton of dollars on lending early on actually builds a scalable venture backable business. In our case, it's because we use that to create network effects. I still remember by the time I crossed about 5 or 600,000 customers, merchants that I had been calling for years, for years with no have no impact start calling us. So top 10 merchants in the country, they called us and said, I'm tired of having customers ask me about your product. When can we integrate? So that was the battle along. What I would like to say we're extremely explicit about it. Probably not, but that was the battle along. It turned out to be as a business, as I said, great revenue, phenomenal margins and now profitable. So no, raising that amount of money wasn't a mistake. I think there was a mistaking how we spend a few things and some, I mean I can tell we made many mistakes, but raising the amount of money was definitely not a mistake. We got some phenomenal partners along the way like you know, gic, usp, Greycroft. So it also not only do you get good capital, we got great people around the table.
- [39:59]
B
You mentioned that towards the end of 23 you felt pretty burnt out. And I think a lot of founders can relate to what did you do about it? How did you address the burnout?
- [40:09]
A
I took 10 days off and went kite surfing. So I was very tactical thing I did. I think it was also very natural. You know, I had moved to Brazil with with Daniel, but also with my family to make adding Brazil a success. And we pulled out of the market. The best decision we've made, but still a very hard decision. Right. Like we went all in. We moved our young family, he's our kids, our wives, everyone. And we just couldn't make it work. So I think there was a little bit of, like, ensuring. We. We kind of had that grief. We also had that brutal round. We try to make things happen. And it was funny. Like we did in 23, we did 30mil of revenue. I think when we went to a lot of investors, they said, I think that's the ceiling in Colombia. We're going to go over 5x this year. Calendar year. Our run rate is much higher. Even so, you're sitting there, you're like, God, like, have me raise money. I had to exit the market. You start going through some growing pains. And it was just a lot. I think what really changed for me coming into 24 was A. I moved back to Colombia to. I got way closer to the product. The thing you got to understand about our product is it's awesome. You're out on field and you see it. You go on streets and you see every street because we created this thing which the Addy logo comes out kind of like perpendicular out of street. You go on screen, you see these things. You're starting engaging with it. You see it in action. For the first semester of 24, every Wednesday, I just went out on the field and it was like every Wednesday was field day. And I would take videos, record, send voice notes. People like the company started fearing this field day. So I've been like, this is not working. And you'd go into a random store and the guy would be like, oh, I haven't received the call back. And I'm like, don't worry about it. Let me call this person. He's like, who are you? I'm like, oh, I'm the CEO of the company. Good to meet you. And our field guys have, like a little jacket. So I put the jacket on and that was awesome. And then at the end of 24, we could just see the mousetrap working and saying, wow, we can actually build a great business. We've built a good business. We can build a great business. And that was also a bit of an inspiration.
- [42:22]
B
Sandy, I think there's a lot to learn from this conversation. I'm rooting for you. I always enjoy our conversations. Thanks for stopping by and thanks for being also a friend of Gilgamesh. You know, there's going to be very big things ahead. No doubt about that.
- [42:39]
A
Oh, Miguel, thank you. This was a. This was a wonderful, wonderful time to reconnect and to lift the rock a little bit and share what we've been up to over the last three years.
- [42:49]
B
All right, thank you. Thanks for tuning in. I hope you enjoyed this great episode with Santiago from Addy. If you want more interviews, make sure to subscribe, follow and leave a review you on Apple, Spotify, YouTube or wherever you get your shows. It helps and means a lot. And if you have any suggestions or thoughts about the show, just drop me a line on LinkedIn or Twitter. Signing off till next week. I'm your host Miguel Armasa.