Razorpay replaced their entire onboarding with AI Agents | Harshil Mathur | Intelligent Indians

And we had a choice: do we pay out our largest customers, like the top 10 largest customers, or do we pay out the rest thousand small customers? These large guys can manage without the money. The small guys were going to die tomorrow. And it was very simple. I didn't have to think five minutes. Let's just pay out the small guys. Right. And the business teams were very unhappy because the large guys are the ones that make money. Right. There is one learning for all founders: they have a DNA and they stick to that, their values, and they say if small businesses matter, the decision is very easy.

Yeah. Vikram knows the entire story, like he's been with us from day one. So yeah, I think the early story is fairly simple. I was born and raised in Jaipur. Average middle-class family. Went schooling in Jaipur, went to IIT, and was always interested in coding. I was coding since, I don't know, sixth standard. But unfortunately didn't make the JEE rank for CS, so got into mechanical. So I used to spend my weekends and evenings coding even through college. I had no interest in mechanical. So there I met Shashank. We started doing stuff on the side while going for lectures and mostly sleeping through lectures.

And Shashank had a better rank. He's in computer science. Because a lot of times people ask us, you know, should I choose a co-founder? Found out: choose one with a higher rank.

And he was a senior to me in college.

Yeah. So the way it started is both me and Shashank met each other. A bunch of people like me came together and we started chatting that hey, we want to code in our evenings and weekends. We don't want to do anything else. But we're not getting it. Even Shashank was in CS, most of the lectures are very theoretical. So where do you really do coding, right? And you don't learn coding by just reading lectures. So we got together a bunch of us and we started a college group called SDS Labs where we would build projects together. So we built, for example, a project where we would stream the cricket matches inside college. This was before Hotstar, before all of that. We would pick up the cricket matches from this TV and stream it inside college so people would watch it.

It's inside college. Like it's their TV rooms. We were legal. So you said TV rooms are legal. Like you have one TV room, everyone watches.

Yeah.

I mean, you just build stuff because it's cool and inside college who cares, right? Like just build it and deploy. So we were doing...

Yeah, inspired by that. But I think it was just simple: hey, we have to watch a World Cup. It was coming up and we said, hey, we have to go down to our TV room and watch the World Cup. How do we watch on our laptop? The internet speed wasn't good enough to stream anything online. But local speed is decent enough. So then we said, hey, we can just hook this TV to a computer, stream it in the campus, and everyone can watch it. And that's how that started. So we created this SDS Labs. The group survives today and it continues to bring in young...

The new generation is not like that. We cared about money after it. But during college, it's a cool thing to do. You're building on weekends. Let's just build it. We built a bunch of stuff. We built a software to download software easily and all of that. And that's how we got to work together with a bunch of folks like Shashank and a lot of us started working together. So when we were graduating from college, we said, okay, now this is ending. Now what next? And Shashank was graduating and he got placed in Microsoft in the US. And we spoke to each other and he was like, hey, we should keep chatting on the side. Maybe there's something we can do, something else.

Yeah, both of us, coincidentally, both of our dads work in State Bank of India. Used to work in State Bank of India.

I don't think I've ever thought of entrepreneurship or doing my own business. I mean, I come from an average background where people like my dad expected me to get a job. That's the mindset: the best thing would be to get a job in a great tech company. But I was not in CS, so I thought even that is going to be hard. But I think business was never on the cards.

Six months. Yeah. So for mechanical, the highest paying companies that come on day zero was Schlumberger. It was a first day company. Sat in it, got placed in it, excited and celebrated. Of course my dad celebrated and everyone congratulated and all of that. Went to the job on day one and this is not what I really want to do.

Yeah. And it was a life-changing amount of money, not just for me but for the entire family, right? So I got into Abu Dhabi, right? So it's offshore. There's great pay packages, great everything. And essentially the company takes care of your life. There's no money you have to spend. They give you five-star accommodation, they give you food, they give you travel, they give you everything. But the actual job, I did it and nothing against it, but it was not for me. I am a guy who wants to sit and code. I don't want to step on the field. And being in an oil field is its own kind of job and there are people who like it. And I was very clear this is not for me. So I told Shashank, spoke to me that we should try something of our own. But I thought, he's a guy going to the US. He's saying it once. Once he lands in the US, I'm sure he's going to forget all about me. But he kept in touch. So we kept chatting. And the good thing about oil field jobs is you have a lot of free time. You're offshore for a week, then you're free for two weeks, and then you're offshore for a week. So I would have a lot of free time. Most of my friends would go out and party doing that. I was not one who enjoyed that a lot. So I'd again speak to Shashank, build stuff on the side, code stuff. And that's when we ran into this problem. We were building some side project, similar to SDS Labs. We were building a social crowdfunding platform and we wanted to accept payments. And we thought, payments, this should be the easiest part of it. It turned out to be the hardest part of building it. So we reached out to all the largest banks and the largest payment companies and we saw it was very, very painful to get started with payments. We went to Facebook groups like Bangalore Startups, Pune Startups and asked how does everyone solve it? And realized every startup founder had the same challenge. We were not even doing a startup, we were doing a side project, but everyone had the same challenge. And everyone said, just start accepting cash. It's easier to accept cash in India than to accept digital payments. And that urged my tech side. The point of technology has generally been democratization. Like the internet just allows everyone to start something. Social media gives everyone a voice. E-commerce gives anyone the ability to sell something without setting up an office. How could digital payments be less accessible than physical payments? It didn't make any sense. So in our standard techy mindset, you're like, how hard could this be? Why is nobody solving it?

Very good. Nobody has connected it this well. Lived experience is important so that you can empathize and probably stay resilient through the ups and downs also.

Yeah. And I think it's still important in decision making today, right? A lot of times I have to, me and Shashank have to push on that, hey, we can't take this decision because it'll hurt the small guy. Because again, an average small business is generally not the highest revenue driver for the company. Right. So a lot of times we have to, hey, we can't do this. I remember in one of those Yes Bank moratoriums when Yes Bank's funds were frozen and a lot of money got stuck for us as well. And we had a choice: do we pay out our largest customers, like the top 10 largest customers, or do we pay out the rest thousand small customers? And essentially I took that call that these large guys can manage without the money for a week, two weeks, whatever it takes for Yes Bank to clear up. The small guys are going to die tomorrow if they don't get their money. So you have to, unless you have that empathy, it's very hard for you to understand. Everyone will say, I just pay this guy, his 2 crores is stuck, and this guy's 10 lakh is stuck. But this guy is going to die without his 10 lakh. He can manage.

Yeah. And I think I can tell you, for example, the hardest part in building a payments business is you need to get a lot of bank approvals and certifications. And so while the product part honestly was easiest because we imagined that, hey, if I was starting, I'm building an online payment platform and I need payments, how should it look like? And we built it from that. But the bank part was hardest. So what I did was, I had a holiday. I went to a bank branch next to my home and just ran into it and said, 'Hey, I want to start a payment gateway. How do I go about it?' And the guy looked at me and laughed. Nobody comes to a bank branch in Jaipur and says, 'I want to start a payment gateway.' He said, 'Do you want a payment gateway?' I said, 'No, I want to start my own. How do I get your APIs?' And the guy had no idea. So he said, 'Yeah, I can connect you to somebody in the Delhi office.' And I said, 'Sure.' He connected me to somebody. I spoke to them. They understood what I wanted to do and then they rejected us. Right. So then I went to the next bank, did the same thing, and the next bank. Shashank did the same thing on his side. We started figuring out people on LinkedIn, somebody who heads the payment gateway business at this bank and that bank. Most of those folks are based in Mumbai. So we got connected with those people, sent a lot of requests. Some of them just rejected us. But the rejections were good because every time we spoke to some of these guys, they gave us something more. They told us something more. And we pushed them like, why are you rejecting? What is missing? And they said, you don't understand the business that well. You don't understand this part of the business very well. You're not understanding the risk of this and all of that.

We would push them, hey, is there somebody? And he said, hey, maybe you can reach out to the senior in my bank and maybe he can help. And that's how we kept pushing. So I think there's no simple answer to that. Just keep pushing.

Yeah, that's how we kept moving. And I would have met, in that three to six months period, I would have met and mostly been rejected by almost 100-ish bankers in some form.

Yeah, fortunately or unfortunately, I think I had no idea which are there. It is ignorance.

But what I did know is I was in these communities of Bangalore Startups, Pune Startups, and nobody in these communities was happy with what was available.

So that was my easiest signal that maybe there were 100 gateways, but the ecosystem was not happy with what was available. The ecosystem was looking for something else. And that was my signal. I didn't care who else existed. The ecosystem that really matters was not being serviced by them. And what we discovered later was that yes, there were a lot of payment gateways. Everybody was solving for large companies.

Was Stripe already there?

We looked up at Stripe and honestly one of the questions was that why is something like this not in India? And the choice was let's wait for something like that to come to India or we build it our own.

Yeah.

Honestly, it actually worked well for us. And I'll give you some examples, right? Like most payment gateways at that point of time in India looked very similar because what most payment gateways did was they went to the bank, took their API, built their wrapper on top of it, and then sold to customers. And because bank APIs were the same, the payment gateways looked the same. We didn't get access to banks on day one, right? So we were building a payment gateway without knowing what the bank said. So we assumed the bank set is a black box. It looks in some shape. We have no idea how it will look, but we'll build from a merchant perspective: how would a merchant want to use it? And when we built our product from that perspective, we got banks maybe six months after we had built a product. And then we plugged the bank and did a lot of work to build a scaffolding because the bank looked completely different than what we had put out to our merchants. But it worked well because we were building merchant first and not bank first. Which is the right way to build it. Why should you build it based on your supplier? You should build it based on your customer. So I think being an outsider helped shape the product direction in the right way because we didn't have access to those. And I think a lot of decisions that we took, we were able to take them because we had no idea how the industry works. We had a great idea of how startups work. So we were chatting with these guys. We were in co-working spaces. There were a lot of startups around us. We would sit with them, set up the payment gateway for them, ask them what they like, what they dislike, and we'll keep modifying it. And I think the most important part is you have to be in the ecosystem that you're going to serve, rather than the ecosystem that is going to help you, sir. And I think it worked well for us. Of course, it had its own challenges. The bank approval process alone took us so long. So we started building this properly full-time. We left our jobs in 2014. It took us a year to take it live. 2015 is when we went live. Most tech spaces don't take this long because you're dependent on approvals and all of that. So that was the hard part of being an outsider.

Yeah, success rate was the most. And again, from a merchant perspective, what do they care about? 65% success rate.

85, 90%.

90% plus. 92, 93, yeah.

And what can we do more to improve? At that point, we moved by two, three percentage points. Today, now we move by 0.1 point, 0.2 percentage points. It's still the same effort and trigger that goes in.

And I think it was simple. We were speaking to merchants. What merchants care about in a payment gateway? I tell my team, you are like a utility company. Nobody cares when your electricity is coming. Nobody says, hey, what electricity is coming? But the five minutes that you're down, you're going to get all the abuses in the world, right? So your job is to avoid the five minutes that you can ever go down, right? And that mindset is something that we still drill down in our company. Unlike a consumer company, we don't optimize for time on our platform. We have to actually minimize the time on our platform. We have to actually minimize the time it takes. The quicker we are, the less time people spend on our platform, the better it is for us. And those things are like we have to change a lot of people who come from consumer backgrounds. We have to change this DNA that they don't increase the time on our checkout. You can't add animations on our checkout which take away consumers' time. You're doing a disservice. So we have to do a lot of that mindset change. And it's still the primary metric: how do we improve success rate?

Yeah. I think YC happened completely unexpectedly, honestly. Because we were building this. We had not yet got the bank approval. This is 2014 mid and late. We were building the product. We were speaking to customers, but we were waiting for the bank approval. And the YC application form came up. We were of course always reading Hacker News. So we went and applied with literally zero expectations because not a lot of Indian companies had gone into YC. We thought we are building such an India-focused thing. And we filled the form. But I think we did a lot of diligence in filling the form, right? Because the good part about the YC form is that it asks you the questions that most founders should ask themselves when they are starting a company, but they typically don't. So we spent our time. It asks you like, why would somebody use you over something else? And what are you doing differently? And all that. Spent a lot of time figuring that out. Applied to YC. Got called for the interview in the US. And so we thought we might not make it back to the US again because I don't know if we can afford a ticket. So we said, let's pack this schedule with meetings. Let's use the best of this week because the YC interview is a 10-minute interview. How do we use the entire schedule? So I did, we did like seven meetings a day for the five days that we were there. And then we did the YC interview. It is a 10-minute interview. So it went in a jiffy. And then in the evening, we got a call that we got into YC. So it was honestly completely unexpected. But YC did impact the course of the company significantly. And we had a choice: we want to move completely to the US for these three months, we want to stay one guy here, one guy there, we want to make the best of YC. So we both moved there for three months. I think it was the best exposure we could have to global standards of building, global standards of product, what other teams are doing and how teams are solving their problems, meeting some of the best folks that we could meet in terms of mentors. It was a completely amazing experience. And honestly, it set a lot of... I don't think it was anything specific. It's not like YC told us, hey, build a product this. In fact, they don't even go into execution details. But just the exposure alone was enough to change the course of the company.

Yeah. We pitched. We started pitching. We were in Jaipur. We started pitching to some VCs and stuff. And everybody just said the same thing: there are so many payment gateways. It's a utility product. Why are you guys getting into this? And we just walked out. So after a few of those meetings, we realized it's not going to happen. We went into YC. And while we were going through YC, suddenly we got a ping from somebody on LinkedIn. And he said, hey, I heard you guys are building something in payments. This was Rajat.

No idea. No, nobody knew you were in...

Yeah, we said we don't want to do anything till the YC demo day because it's going to be painful. So we said, but then we saw, hey, Matrix Partners and all of that. And we had read about Matrix Partners. We thought it's a massive VC firm. It seems like we can't miss it. So we set up a meeting just with zero hopes.

Yeah.

Oh.

But I can tell you, the first meeting at least from our perspective didn't go very well. I don't think because we were not prepared at all. So we said something. I don't know what it landed to you, but I felt like we did a very unimpressive job. I thought this guy is never going to invest in us.

But for us, it was essentially just this. They said mobile, they said failure rates. And I had just spoken to others who had invested in payments and they said the only thing that mattered in PayPal was frictionless merchant onboarding. And somehow everything that I had learned was exactly what you guys were saying. And so for us, it was a fantastic meeting.

We broke all the rules. I remember meeting you on Skype in those days. And having built it in Bazi myself, I'm like, but here's the problem and here's the problem. And I think it's a very instructive thing. You guys had gone so deep in the banking architecture. You had an answer for everything. You had thought about it. So...

So the best million dollars we have ever invested. And so whenever you know you start your next one, please consider it. But anyway, so taking it forward from there, you guys get started. I know that there are multiple inflection points. Talk about the journey. From the outside, it looks like a raging success. What happened between? What are the big milestones in the company's history? We'll come to the next question as part of that because I know one of which was COVID. But before that, was demonetization? Talk us through and what were the big heart attacks along the way?

Yeah, I mean, the first two, three years were extremely hard. Extremely hard. Not from a perspective we were still very motivated, but a B2B business in payments takes its own inflection curve. You can't fast track it unlike consumer and similar spaces. There's no way to even pour money at the problem. We had raised a significantly large amount of money because of post-YC and 2015. There was a lot of excitement about India. So we got funded from you guys. We got funded from Tiger. We raised a large round for the size of company we were. And then there was so much pressure that, hey, we've raised like $1 million, you need to scale really rapidly. But the unfortunate part, or fortunate part, of a B2B business is there's no way to use that money to grow faster. It's not like you can discount your way through it. You can't cashback your way through it. When business folks buy payments, they don't buy it because you're offering... Price is a factor, but it's not a primary factor. And it's still very less understood. The B2B side is comparatively much lesser price sensitive than the consumer side because the guy cares about utility and value so much more. If I deliver a poor success rate by one percentage point, it doesn't matter what price I'm giving. It doesn't matter at all. The price is not even a discussion in those meetings. So we had to prove our product. We had to prove that we can serve large customers. And honestly, for the first two, three years, we had zero large customers. We were only very, very small startups who were using us, not even midsize customers. I remember a conversation I had with Vikram where we were trying to attack some of these large guys and we were not getting any headroom. And Vikram said, you can... There's a hunting example he gave: you can hunt elephants, you can hunt deer, or you can hunt rabbits. And you can think that you'll hunt one elephant and you'll be chill for your life. It's never going to happen. So generally, you have to hunt rabbits and then maybe try to get one deer. Don't try to go for elephants early. And that landed very well for us because we were trying to do everything. We're trying to support startups. We're trying to do mid companies. We're trying to do elephants. And there is this greed of sort that if we get this one large guy, our life will be sorted. We have to take so much effort to onboard these thousand small guys. And we realized that there's no fast way to fast track it. So the first large customer that went live with Razorpay was Goibibo. And it went live in, I think, 2017, two years after we had started. For one reason, their core payment partner failed for a week and they wanted to try something else. So that was the first win. And we said we need to do this really well. So we had an entire team dedicated to ensuring that that merchant sees the best of Razorpay. We built a lot of features just for them. And by the way, it didn't go so well. They went live with us, they used us, and then after a couple of months, they paused us completely because we were not meeting the standards. We put all our energy to ensure because that was going to be our benchmark merchant. That's going to tell every other merchant that, hey, we can serve enterprises now. It took a lot of effort, almost six to eight months, to stabilize that. And that became our gateway to tell every other merchant, yes, Goibibo uses us. Call them. And B2B takes a lot of time.

I think it was just hustle. We essentially just sat down and said, no matter what happens, we have to give them the best experience they can have. And not just experience in terms of product, but servicing, sales, support. So you can, as a startup, probably go deeper with a customer than a large incumbent.

So the large incumbent had hundreds of customers. They were one of them, right? And they were getting attention. It's not like they were not, but not the attention that we could get. But honestly, none of that mattered if our product couldn't deliver. Even with all the efforts and all the support, we got paused after a month or two because we weren't meeting the standard. So product is first. Everything else is scaffolding on top. And you can use that scaffolding to fight a lot of things. You can't fight a bad product. So having the product right and then having all of that allowed us to really compete and ensure that they get stabilized and they continue to give us the majority of the volumes. Once we had that, then it became our proving point. It was our foot in the door to large customers.

Yeah. No, and I remember that we were building and we had a lot of money in the bank, I think. But we had no way to forcefully spend it because our spend was hiring. And you can't just hire. We were very, very selective in our quality of hire. And I think that was one of the good decisions because there was a lot of pressure at times that, hey, if we hire like 20 more...

Yeah. And there's a lot of pressure that if I hire 50 more people, I'll actually put it on Shashank. Because at one point, I even told Shashank, why don't you just reduce our standards? Get these 50 people, get the things we want to get out of the door, and maybe you can raise our standards again. And that doesn't work, right? Once you reduce your standards, once you have a B team, you have a B team. You can't change it. Once you have a B team, you have a B team. You are building foundation and reputation even when the metrics may not be showing up. So I think that relentless grind in that period on the right things is very important.

Yeah, I think first of all, on the COVID piece, right? Like COVID hit...

Maybe it was a Yes Bank, I don't remember which one.

Yeah, I think there's no fighting, and this came from day one. From day one, I would speak to our customers. As I said, when we were building Razorpay, we were working with banks, trying to do all of that. But the one thing that kept us going when we were facing all these rejections is that I was constantly speaking to our customers, constantly speaking to our potential customers at that point of time, on how they do things. And that was my real source of energy. We'd go to a bank, get rejected, and then I would go and speak to the customer, and they'd say, 'Please solve this for me. If you solve this for me, you're going to solve something for my life.' And I get that energy to go to the next bank and pitch the same thing. And I think that continued as the company was built on. So today also, I speak to at least one customer a week. And the advantage of that is, it gives me a real view of what we are doing well, what we are not doing well. But also, it's my real source of energy. Like, are we making a difference in their life? Do they care if we go away for a day? And I think that is hard to match. So this is you are our customer, this is our real source of energy.

I'm in a lot of startup groups, and every startup, people were like, 'Hey, I have this issue or this issue,' and I'm the first one to DM them, even on Twitter. And the idea is, it helps me get a very real view of what we did wrong. I can always pass it to somebody and take a screenshot and send to somebody, 'Hey, reach out to this guy,' but then I won't get a real view of what we actually did wrong, where did our system fail because of which this guy had to post on a group. People are, I think in the B2B world, people are generally fairly empathetic and also fairly logical. If they're posting in a group, it means they have exhausted their standard options, and I need to figure out why the standard ways didn't work. And that's how we really improve.

I think I have generally seen that we work with founders, we work with people who are building businesses. I've generally seen they understand. In fact, we have messed up. All the good ones have to, right? And we have messed up a lot of times. In the early days, I think Vikram knows this story. We did our YC launch, we went live, and I think about a month after our YC launch demo day, where we were covered in TechCrunch, the bank that we were partnered with pulled the plug, and our entire system was down for a week. There were 20-30 merchants who had gone live, who trusted us, a young startup, to power their payments, and their payments couldn't go through because Razorpay is now down. Now what do we do? And one rule I set for the team—we were a team of seven or eight people—I said, 'We'll pick up every phone, we'll talk to every customer, we'll tell them, even if they abuse us, we're going to constantly talk to them.' Because it's easy for people to say, 'This guy is going to just call me and abuse me, and let me switch off my phone, let me turn off the interfaces.' We spoke to every single customer. Some of them were empathetic, some of them were abusive—I mean proper Hindi, aggressive abuses that I heard. But I can tell you that the guy who abused is still live as a merchant on Razorpay today.

Yeah, I think it sort of happened honestly. When you're going through a crisis, I don't think it's—and in that mindset, it's really hard to think. I remember the Yes Bank situation. It was an evening, I think in March 2020, and it was at 7-8 PM. We were all heading home. I just left the office, and then this notification came that Yes Bank is going to be frozen. And we had a lot of dependence on Yes Bank. So I just went back to office.

So a lot of our core payment gateway business fortunately was not on Yes Bank, but our RazorpayX payouts, all of that was on Yes Bank. It was 100-ish plus crores of our customers' money stuck in Yes Bank.

And so I just went back to office. I realized this is going to be a night. We pinged a non-leadership group. Everyone came back to office quickly. Slowly, everyone started coming back to office, and we sat in the room and we started figuring out what all its impacts are, who all is impacted, what is our game plan, and we just started. And everyone just started rolling in. I think within an hour, around 40-50 people, the entire leadership team was there. I think by midnight, there were like 200-300 people in the office. Everyone was just working and figuring out on its own. And it's not just the people who are directly involved—our housekeeping staff came back, our admin staff came back, they were arranging food for these people, they were arranging coffee. And all of this was happening organically. These are the times when you really see the company culture come alive. And I remember after a point, I was just sitting on the side and I had nothing to do, because once I said, 'Hey, these are the five things we need to focus on,' everyone just took charge. The team just took charge, and I was just sitting waiting if there's a decision pending on me that should come to me, but execution was completely online. There's nothing for me to do in terms of execution.

I think the company culture is just built.

The grind, that's like the culture that you set over the years.

Showing up. And we were very, very transparent. We called every customer, we put on email that this is what we're going to do. And this is when we took the call that—I remember the finance team presented that this is where how much money is stuck, these are the guys I can pay out, and we dumped our own cash to say, 'Can we pay out some of these guys?' And the choice was like, pay out these 10 large guys or pay out these thousand small guys. And it was very simple. I didn't have to think five minutes. Let's just pay out the small guys. And the business team was very unhappy because the large guys are the ones that make money. So, 'Hey, this guy's going to be pissed, you're not going to get his business again,' and all of that. I said, 'It honestly doesn't matter.' We got legal notice from those large guys: 'How can you not pay us out because we know you've paid out these small guys? How can you not pay us out?' And we said, 'We'll handle this.'

I will say, in these crises, there is one learning for all founders: they have a DNA and they stick to that, their values. And they say, if small businesses matter, the decision is very easy. If product, tech, and failure rates matter, the decision is very easy. And I think that's lasted you through every crisis.

Very good. And I think you have to stay by it. And I think we spoke about company culture and all of that, it's something we spend a lot of time on. But some of these decisions come to founders. Actually, it's very hard for anyone else to take the decisions because it's a very hard call. You're essentially taking a call to hurt your business for your principles. It's going to be very hard for anyone who's not in that position to take that call. So being there in that room, my only job was to take such calls.

No, but I think it's true. I think everyone in that leadership team was watching what decisions we were taking, and it set the tone for the next few years. Because everyone saw that in that moment, the decisions we took. So it was not just about the customers, it was also for our team. If we diluted from that positioning, it would set the tone for everyone else as well. 'Hey, at the end of the day, culture and all is fine, startups and all we talk about is fine, at the end of the day it's revenue that matters.' You would set that tone, and then that is your culture. It's not what you put up on the wall. It's not a saying that 'memes matter.' It's what you do, because the team is going to carry it now for every.

Yeah.

Yeah, and honestly, that's the harder part. I think we are a pro-tech company. The only thing that differentiates us from the other players in the same industry is how good our product and tech is. At the end of the day, it's that simple. We can't say, 'Hey, our logistics is better, our operations is better, or something else.' It's only product and tech. The hard part of product and tech is it's ever-changing. It's always changing. We spoke about how we build the fastest onboarding, and everyone else will catch up with it. We build the fastest integration, everyone else will catch up with it. Because in tech, it's not very hard to replicate. Look at what is successful and replicate it. The only thing that keeps us going is how do we ensure that we are next first to the next thing that happens. First to the next thing that happens, and it's consistently a hard part. And it gets harder and harder at scale. This FTX, we launched 100-plus launches. Most companies launch one or two things. We did 100 launches. The reason we had to do it is because we have to be first at everything. And that's the culture in the company. We have to be first in everything new that happens in this ecosystem. We have to be the first at it. And a lot of times, we'll do things which don't take off, and it's fine. You fail fast, you close it, you move to the next thing. But you try and be the first at everything. So if even if there's one new launch that happens in the ecosystem that somebody else does and that we have not done, my first ping to the team is, 'Why did we not do it first?'

Yeah, you can't not be first. So somebody launches something, I see it on social media and I post it to my team. 'How did we miss it out? How were we not first?' And I think we do a DRC. 'Why did we not pick it up? Who took that call?' And so it's simple.

We love our customers, we build a culture. 'How are we not first?'

Yeah, that's also part of culture. So every single new thing that happens in the payment industry, we are generally the first one. 2016, remember, UPI came in. Everyone was saying, 'Hey, will UPI take off or not? SBI and SDC were not on UPI.' So everyone said, 'Without two large banks, who's going to do UPI?' We were the first to launch UPI. We were very small, we had like 10,000 businesses, but we went live. And what happened is that when UPI did take off in 2-3 months, suddenly UPI started taking off. We were the only payment gateway serving UPI. So a lot of large customers came to us just because they wanted UPI. The largest guys were not, and even was 6 months away. We were ready, and we got that in. So we rode that wave. And similarly, it happened. And while these are the positive ones, we built a bunch of stuff that never took off. We built an invoicing system that never took off. We built a lot of stop-start products.

Yeah, because if you're taking that call, you want to be first at everything. You can't ensure that everything is going to be successful. So you take that hard choice, but it works.

We looked at the blockchain wave very closely. We looked at crypto very closely. One part was the regulation part was very clear that you can't touch it.

Yeah. Yeah. I think being in building in the fintech space, one of the most regulated spaces, from day one it was very clear to us that you can fight everything, but if you mess up on regulation, it's going to be very hard to fight it out. And you guys are actually the opposite. You're very, very—

Yeah, mostly I think, see, I think everybody has been using AI and ChatGPT, but in late December, the newer models started rolling in. And I think I tried first time in late December or early January the newer models from the model companies, and something changed. I think it became from AI being an assistant to AI being an execution engine, or AI as a world took over. The model was smart enough that it didn't need guidance, it didn't need for you to tell it to do this and do all the tool calling. It became so amazing. And I remember I tried OpenClaw for the first time around end of January, and I set it up, and I'm like, 'Wow, this is just going to change the world.' Because not just the product itself, but just this capability now that the model has, that you can really trust the model with hardcore execution end to end. You create a plan with it, you tell it what it needs to do, and I think that intelligence on the fly, execution intelligence on the fly, was just mind-blowing. So for the first time, I started picking up and building products myself. I think after 6 years maybe, I started building stuff with it, started tinkering with it, started using it to set up an agent to do meeting review notes for me, started using it to do a diet plan for me and be a dietician for me, and all of that. I just started doing a bunch of stuff. I think after I set up OpenClaw, for three to four or five days, people in the office said, 'I don't know where you are, but you are completely distracted,' because I couldn't focus on work at all. I was just logged into this, testing its limits, testing what it can do, testing what I can do with it, and it just blew my mind. And after that, I moved to Cloud Code and all of that. But I think what fundamentally became clear to me is that this changes everything, and there's no point of saying that the world is going to be the same. So after a couple of weeks from there, we sat down in a room and Shashank said, 'If this is where it's going to head, we don't want to wait to get disrupted by something of this sort. So if I were to start Razorpay today, how would I build it knowing that this technology exists?' And we said, 'How is our onboarding going to look like? How would our support look like? How would our integration look like? How will our dashboard look like? How everything that customer interacts with look like?' And we realized that it's going to be completely different. Everything has to completely change. We said, 'Let's do that.' So we worked for the next two or three months. We just said, 'Let's completely rebuild everything.' And this FTX we launched, which was everything from onboarding to integration to support to dashboard, we completely rebuilt. And it is not easy because there's a lot of love for what we have built as well. You're proud of it.

Yeah. And that's what you do now. That's what it is now.

But I think just for the audience to know, ChatGPT brought AI to the masses. OpenClaw is probably a bigger disruption when it comes to workflow for the masses. Anybody can build their workflow.

And you can put Cloud Code equivalent. You can do it anywhere. You can use Cloud Code, you can use Cowork, you can use API and build your own harness. It doesn't really matter. The model is the power here. The model has reached that potential that you can now trust it with long-form execution. It can—I sometimes, it's become a habit for me. Every evening, I set a whole plan with my Cloud Code and leave it jobs for the night: 'These are the five things you have to do before I wake up.' And it does it completely on its own. In the morning, it's there. Maybe there's one bug, maybe there's two bugs, maybe sometimes it's completely perfect. The point is, it's almost there. And I think that power that this represents is just amazing. How should people react?

Yeah, I can tell you, my GitHub profile, I think it was last green in 2015. And in the last 3 months, it has fully been green.

Of course, I try to fake it. I'm just saying, it has become green in the last 20-30 days since I've been fully on this, because it's just shipping. It's just fun. Half of the things I'm building have nothing to do with—

So much. All your tedious routine things that irritate you, just try them with it. It will change your life.

No, but no, he is on steroids just like you. And for you, it was a six-year hiatus. I think for him it was a 25-year—started building like any a lot of tech founders like me. You started with tech. The reason I got into it, I like doing tech. I like building. Over time, of course, you have to scale the company, and the tech part takes a lot of time, so you move away from tech. I think it's the best time for everyone to return back to tech. If you have liked coding at any point in your life, you will love it when you pick it up now, because it's just so much faster. I was speaking about SDS right now. I look at it like, if this existed, what I spent like one year building SDS, I could have built it in a day. I could have built it in 5 minutes. It's just amazing. We spent four years whatever we built in the four years of college at SDS could now be built—I hope all of you can just hear the animation and the enthusiasm in his voice. But that's sort of how you feel about tech.

Yeah, and it's—I mean, we are still a work in progress. I would say I'm happy with where we are, but there's a lot more to do. I think it starts at the top. I think as I'm pushing that, we've told every engineering manager that—because a lot of engineering managers don't get away from coding—every single person has to pick up Cloud Code, every single person has to try something. It doesn't matter what you do. The reason I had to do it and the reason I want all managers to do it is because you need to know the boundaries of the new technology. It's great at a lot of things, it's still not good for a lot of things. If you don't try it, you'll either overestimate or underestimate it. You might expect your engineers to do something in 5 minutes which they can't do anyway still, or you would expect that it's going to take months which can now be done in 10 days or maybe 5. So that's the first thing: how do you ensure every single one? We did a leadership hackathon last year. Typically, hackathons are done by companies where young engineers participate. We flipped it and said every leader built something on AI and showed it. And the young builders in the company became judges. So our last hackathon winners became judges, and every leader including me presented that, 'Hey, this is what I've built.' I came fourth, by the way. So flip your org. It has to start at the top. Unless you start at the top, you won't understand the difference between the power and the limitations. Us going down this rabbit hole has told us where we can invest. Half the time when we were investing, we were worried, 'Is this going to be in the path of the model? ChatGPT will launch something and will this go?' When you go deep, you use the word 'harness.' It's actually our investment thesis: who's building harnesses? So in every industry, going back to the question, go deep, do it yourself, and then you'll know a lot of the answers.

So I think AI makes the execution speed really fast. It makes intelligence available on the fly. But the pro and con of it is, it's going to make your weaknesses appear far faster, but it's going to make them far bigger as well. So AI doesn't solve your org weaknesses. In fact, I tried this tool called—it allows you to build AI-native companies, which is basically you can give it a task and it goes and builds the entire company. It hires agents, it creates—like you tell it, 'I want to build a B2C marketing platform,' it'll hire five agents, it will create tasks to them, and they'll start building. And I tried it. It looked very cool and all of that, and it produced garbage. And I said, 'Why is it producing garbage?' I said, 'Because there's no plan.' I just told it, 'Build a D2C marketing company.' I don't have a plan. I don't have a structure. The agent doesn't know what it's supposed to do and what it's not supposed to do. The boundary is not clear. So the agent just picked that up, gave it to five hired people, told them, 'This is what you need to do. This is your task.' And the agent started doing something. Initially, they were doing something right, and then they started drifting in random directions. They started going in some other direction. And that's the beauty of it. So when people worry about jobs, when people worry about disruption, what they should realize is there is a limitation. And this limitation, by the way, is very architectural. LLMs are not deterministic by definition. They will drift. So the business opportunity may be even larger in a much larger time. It's just shifted. But to understand where you have to play in the shift, you have to go deep. And it's the same as companies. If you start a company and nobody knows the plan, nobody knows what you're supposed to do, and you just say, 'Hey, we're going to build a payment gateway. You are a sales person, you are an engineering person, you are an ops person. Let's just do it.' Companies can't be built like that. So this just proved to me in 10 minutes what you learn in a company in a month: that you can't operate without a plan. You can't operate without a structure. And I think it was a great learning for me.

No, I worked for two and a half weeks with our AI developer to design to redo our entire portfolio review, which we covered yesterday, and these guys are looking at it, garbage. So I said, 'Something input can be used.' Yes. Because it is so good as it produces massive amounts of garbage, and it sounds smart because it's so massive. It looks like, 'Oh wow, whatever.' And I think there's one fundamental delta. While people are saying that engineers are better at it, I think one fundamental problem that engineers have—and I have the same problem—a lot of times engineers are excited by the complexity of the build than the outcome of the build. So a lot of times you do something with this, it creates something amazing. You look at it, 'Oh wow, how it's so complex, it's so intricate.' You're proud of what you've built because it's so complex. So I tweeted this: whenever I rebuild something, I show it to my wife, and I tell her, 'Criticize it.' And when she breaks it down, then I realize, 'Oh, I was just excited about the complexity, not by the output.' And you need a fresh take of somebody who just cares about the output, who doesn't care what you've built, to just tell you that the output doesn't matter, the output is all garbage. Sometimes it can be—

No, I do want to pick up one last thread, which is on the agent studio. People have called it different things, but I would say it's the most ambitious bet. And I know you believe in the power of AI to finally make memes adopt technology. So what's that trend line?

Yeah, I think India has always jumped trends, and I think the same applies here. India never really adopted the SaaS wave honestly. The reason SaaS never took off in India is because SaaS expects a business owner to learn the way a tool works, to change their life, and it doesn't work. You expect a business owner who's been doing sales in a certain way to say, 'Hey, now you have a CRM. Whenever you do a sale, put it in this and type it in this, and then whenever you want to do a new sale, pick it up from there.' It's not a natural part of their workflow. So the US adopted SaaS in a big way. India never really adopted B2B SaaS. There are exceptions to this, for example, maybe accounting had SaaS and some of those, but most of the stuff never really took on the SaaS wave. But AI allows to jump that wave because most memes understand hiring. They understand, 'Hey, I want to do sales, I should hire a salesperson. I want to do finance, I should hire a finance person.' And I think agents allow you to build it in that way. And that's our stab with Agent Studio. When you build payments, most startups when we work with them, they take payments from us, then they have to hire a team to manage payments because as you scale, you have a lot of operations around it: payment failed, a customer has a complaint, I have to do a refund, I have a chargeback, I have a dispute. Somebody has to manage operations around it. And now we can provide agents because a lot of small guys can't hire all of those, they can't scale beyond a point.

Give us more time.

Yeah, give us a bit more time. But that's where the world is headed. When you're starting a company, how can I allow you to hire an agent for everything? So we have a cart abandonment agent, for example. It's nothing to do with payments. Honestly, somebody comes on a website, adds something to cart, goes to the payment screen, and then drops off. And it's at the last payment screen, which means the intent was extremely high. They tried a payment and then they dropped off. Now there's an agent that picks up the phone, calls this guy, tries to convince them, 'Hey, could you want to come back? Why did you fail and why did you walk away?' He might say, 'The payment failed.' 'Sure, I'll send you a payment link.' 'I thought it's slightly more expensive.' 'Sure, I'll give you 200 rupees discount, whatever.' And manages it end to end. It works beautifully. So just intelligence on the fly. And I think the models are being built by the large LLM companies, but our job is translating the model and the API into a form that a theme can consume. And I think there's a lot of delta in those two things. I think just that, and that's our attempt with Agent Studio: how do we translate the capabilities of the model to what a meme needs for his day-to-day business? And if we can do that well, I think it can be a massive opportunity.

Herself. Well, thank you. Congratulations for everything you guys have done. And it's been a privilege to partner. For people who are watching you, what is your advice for those who are in their unstarted phase and thinking about starting, or maybe have started and feel they are still unstarted?

I'll separate it. I think for people who are starting up, I think it's fairly clear. AI is making it very easy to start new things, and a lot of people are starting it because it's just easy to start. I think my recommendation to people is, as we spoke about the grind, find a problem statement that you can spend the next 5 to 10 years of your life on. It's very easy to start something because it looks cool on AI, but when you think about it, is this really what you want to do for the next five years? Can you dedicate, when things are going to be hard, when things are not going to work, can you work through the nights because you so care about the problem statement so much that you don't care what everyone is saying? You want to do this. Just think about that, because that's what it's going to take. Just because something is easy to start doesn't mean it's going to get easier to sustain. It's still going to take that much effort to build something large and meaningful. If you have started up and you're scaling and you're still figuring out, I think one thing that I have learned, maybe in some ways through failures and success, is that there's a famous Paul essay on founder mode. I think AI is essentially the time when every founder needs to get into founder mode.

Because the core thing is, nobody can care about your company more than you do.

So founder mode is—there is a micromanager mode, which is you get into everything and you can't scale. And then there's a manager mode, where you say, 'I have executives who do the job, and I just direct them,' which is how large CEOs operate. Founder mode is that you have executives, you have teams who do the things, but the things that really matter to the company, the founder has to get into themselves. They can't offload the core parts of the company. For Razorpay, that's product and tech. For Razorpay, there's going to be brand—what do we stand for—and culture. These three things I can't offload. We can hire the best of the best talent in the world, but these three things I have to get into myself, otherwise you're setting it up for failure. So for whatever you are building, there are a couple of things that are going to matter for your business. It could be different for different, but those things as a founder you can't get away from.

Thank you. Thanks, Aish. Thanks. Welcome. Superb.