Inside Oolka: Money, Trust & AI for Bharat | ft. Utkrishta Kumar

0:00 We all know how powerful AI is, but its relevance in Indian [music] context and its ability to have a very huge impact on masses is [music] unfolding in real time right now. When it comes to consumer AI in a few years, [music] India will be world's biggest, if not one of the biggest markets. >> Fintech is a crowded space, you know, and it's easy for one to assume everything is done. How did you decide to [music] get into financial services or fintech?

0:29 >> There was a very fundamental trust gap that a lot of users had because, you know, it was our fintech products were not very easy for an average user to understand. >> I think it's very easy to assume that [music] the creditworthiness or credit scoring or knowing what impacts your credit score is a knowledge problem. You know, you and Ulka essentially realized that this is not just about information. This is about actioning. >> We are doing the heavy lifting on [music] closing the loop. We are not just telling the user uh what to do. We are doing it [music] on their behalf.

1:04 >> What should a consumer tech founder think about if they're building [music] using AI when it comes to defensibility? Hello everyone. Welcome to Lunch Break with Lightspeed. And uh today is a very special episode. We are talking to UK uh the founder of Ulka uh one of the leading uh consumer tech AI companies in India. Uh very excited to uh to ask some deep questions to UK and and share his learnings with everybody. Hi UK.

1:46 >> Hi. Hi Harsha. Thanks for having me. >> Very good to have you here. Before we kick this off, I have a like a googly of a question for you. Uh what do you feel um will be the biggest takeaway for your listeners today? >> So, uh, we all know how powerful AI is. And, uh, generally, you know, we believe that uh, the power of intelligence that comes with AI is going to unlock a lot. But, its relevance in Indian context and its ability to have a very huge impact on masses is what uh, is un folding in real time right now. And, uh, its relevance to India to actually solve very difficult problems at scale which was not feasible in pre-AI era because of the cost it would take to solve those is where we are going to see a massive massive movement. And, I I do believe that when it comes to consumer AI in a few years, India will be world's biggest, if not one of the biggest markets.

2:56 >> I like what you said. I think this idea that that so far, given the price point where these models are a lot of the maybe like the top 20 30% of the world is using AI. >> Yes. >> And, that in the next few years we will see this unfold into what the masses need at the price point that the masses need. And, India will probably lead that uh, journey. I fully agree with that. Thank you for sharing that. You can tell me like you were a chief business officer at Meesho uh, which IPO'd. And, congratulations on that. I know you're a rich guy, definitely richer than me.

3:29 But, uh, how did you take the decision to move on from that? It was a very interesting job with very very very smart interesting people. Uh, you decided to move on and start up in fintech >> Yeah. >> uh, and in AI. Like, help us understand why. >> So, my stint at Meesho was the learning was like drinking from a firehose, right? So, uh the problems that Meesho solved at scale for mass India uh was phenomenal. And uh I had this feeling that if some of similar design principles could be applied uh in a very different sector to solve for challenges uh which are different but are very meaningful to uh a user's life is where I could see the opening of a new J-curve. Of course, on a personal level, uh you know, I really wanted to build something of my own. Uh uh so, that was uh definitely one of the triggers. But, more importantly, it was uh building for Bharat at scale, uh understanding the tradeoffs between funnel and flywheel, and uh how, uh you know, you can build a a very strong consumer journey, a loop uh that gets stronger with every new outcome achieved in a sector like fintech. And then, definitely would love to expand more on it in in in the discussion. Uh is is what excited me and and led to >> Yeah.

5:01 >> you know, me moving and starting Ulka. >> Well, let's expand on it, actually, because fintech is a crowded space, you know, and it's easy for one to assume everything is done. There's a player for every everything in fintech. How how did you decide to get into uh financial services or fintech? >> So, there were some very strong tailwinds that were visible. On one hand, the TAM of credit was expanding non-linearly. More and more users were becoming credit active. They were taking loans. Uh they were getting new credit cards. Also, uh the broader creditworthiness of an Indian [snorts] consumer was not keeping pace with how, uh you know, the credit penetration was increasing. And some data points about, uh you know, only 7% of Indian users are super prime while in US that number is as high as 24%. So, while penetration [clears throat] was increasing, uh education was not keeping pace with it.

5:59 Uh there was a very fundamental trust gap that a lot of users had because, you know, it was uh our fintech products were not very easy for an average user to understand. So, they'll not understand the difference between an APR and an interest rate, how processing fee plays into the equation. Uh they were not aware uh that some of the loans that they had taken, uh if they were a co-guarantor, I know it could impact their credit score, too.

6:26 They did not understand how credit score worked while the awareness of credit score was through the roof. And it was becoming difficult for them to understand what the best product for them was, right? So, while uh everyone could get a loan, they were not uh there was not enough insight whether this loan is right for you. >> Mhm. >> And there we could see a very strong opening. Uh and in addition, I also felt that the existing uh you know, mega fintech distributors had started to act more like manufacturers.

7:00 So, basically, uh from a user lens, users were confused whether this platform is actually helping me or is this platform selling me a product. And that is where we could feel that uh true user first fintech product could be reimagined, uh which is not transaction first, but which is trying to help users understand uh what their current credit state is, how can they become more credit worthy, and in the process help them choose what is the best financial instrument for them.

7:32 >> Very insightful. When you said uh that there are products that are selling things to the user versus helping the user. >> And I think it's interesting because that's a shift that you can now do with AI. >> Yes. >> Right? Where >> Yes. >> uh traditionally when you built apps, all you could do [snorts] is build a discovery and a transaction layer. >> Yes. >> Right? >> Yes. >> Uh but now because of AI and conversations, >> Yeah.

7:56 >> you're now able to also build a a helpfulness >> Yes. >> layer. >> Yes. >> So, very, very interesting. Thank you for sharing that. What like as you as you started building uh Ulka, I'm curious to know what do you feel is the aha moment for a user who comes on the product? Is there a vision where you would like for it to go? >> Since, you know, we have a lot of context on the user and uh uh you know, we have fine-tuned our modules to help tailor uh not just advice to the user, but also actions on for the user in a very hyper-personalized, a very contextual way. The aha moment happens much sooner. So, if I were to were to give you an example, let's say a user uh comes on the app, puts in his PAN number, uh and you know, we with their permission, we fetch a lot of structured data that we have on the user, and we are able to see information that is not visible to, you know, I mean, a lot of users or even lenders for that matter.

8:56 So, we can figure out that for this particular user, there is some money that is lying in one of their credit cards which they have already closed. Or we can find out that for their particular home loan, uh their interest rates have not been revised for last 2 years while the repo rates have come down. So, on one hand, we have we understand this specific user pain point or, you know, user helpful information. We then explain this to the user. Our agent, uh named Dhruva, will start this discussion as soon as they are onboarded on how we can help them on on this uh frontier. Uh but the aha moment is, you know, when this app goes beyond a companion mode, right? So, it's not limited to advice, but our ability to then just take an action with user's consent, where these journeys start to get initiated. And in a matter of seconds, user who was not even aware of some of the challenges that was hampering his creditworthiness or his, uh, you know, financial prosperity have not only been diagnosed, but they have also been actioned upon. And, uh, then there is a complete feedback loop depending on the timelines, these issues start to get resolved. And that leads to, uh, a massive trust that unfolds and users then keep coming back and keep sharing more, keep asking more questions. And, and that's the aha moment of today. When we look at the vision in the long run, uh, what we believe that, uh, what AI today is enabling is a hyper-personalized workflow where, you know, there is an advice which is like a relationship manager who understands your context and helps you with a specific advice and a specific actionable that is very relevant for you.

10:45 Uh, but end of the day, when it comes to choosing a product, it's still not hyper-personalized to a user's needs. So, we believe that our ability to influence and shape the ecosystem to build financial products for the user, uh, which are tailor-made for them. So, I, you know, I mean, if you have a loan that could be serviced depending on your needs, you don't need to have a monthly EMI option, you can, you know, choose to pay a lot of it up front. Uh, when it comes to insurance, you know, you can have an insurance policy that is totally tailored to your needs. And we believe that this was not feasible in a non-AI world, but our ability to remove all sorts [snorts] of, uh, you know I mean bundled pieces from a user's financial product journey and make it completely unbundled so that every user has access to a very unique product is what we see the future we want to work towards.

11:45 >> Very exciting. So not just hyper-personalized experience but also hyper-personalized financial products. >> Yes. >> Very interesting. Um you're you are at several million users now and you're scaling well. I'm curious, what is the thing that has surprised you the most um about what your users [snorts] want from from Milka what you didn't expect they would be wanting. Um and on the flip side, right? What did you think they would want but they don't? Right? So how has your how have your users surprised you essentially?

12:15 >> Yeah, no great question and I think the answer to both the questions are kind of interlinked, right? So what we definitely underestimated was the intensity of uh you know the challenges that a user today faces, right? So if I were to like just talk about uh you know a user who is a bus operator in a North Indian tier two city, uh he was finding it tough to get a loan uh for his daughter uh she he he wanted a laptop for his daughter and he was finding it very tough to get a loan and his score was not uh you know I mean great and he just could not understand why this is a such a huge challenge and you know I mean why he's not even getting a a laptop on EMI because his score being uh an issue.

13:05 Uh our ability to diagnose the challenge being a very small amount in one of his earlier uh loans that he had taken which was reflecting as an overdue, our ability to settle it and then uh you know within 30 days his scores improved and he could go back and and get this laptop for his for his daughter >> Wow. >> It >> amazing because the kind of feedback that we started to see from him, the ability I mean, the impact it had on his life, right? I mean, from a very high stress situation where this meant a lot to a scenario where he could access it because he could not buy a laptop from his monthly salary. He needed an EMI option, which he was not getting. The extent of trust that it ended up building and the impact it had on his life was something that we totally under indexed.

13:58 And similarly, if I were to say it on the flip side, you know, earlier we were trying to show a lot of information. We were trying to be information dense to tell a user what all needs to be fixed. Our ability to move from, you know, being very intelligent and information dense to a lot more useful for the user and then drive these outcomes for them is what was, you know, slightly counter intuitive because we felt that it was a problem of information asymmetry where these users were not not aware, but we realized that we don't need another utilitarian app which just tells you from an information lens what all you need to do, but we needed something that is that's less DIY and more it done for the user and and that is where we saw, you know, a very very heavy pull. Yeah.

14:53 >> I think it's very easy to assume that the creditworthiness or credit scoring or knowing what impacts your credit score is a knowledge problem. >> Yeah. >> But what I find fascinating is that, you know, you and Ulka essentially realized that this is not just about information. >> Yes. >> This is about actioning on that information. We always apply this thesis to education, right? Like where education as a market without outcomes. >> Yeah. >> You rarely see large companies getting built, right? You need to tie the delivery of knowledge with an outcome.

15:28 >> Yes. >> Uh and it's interesting to apply the same paradigm to FinTech. >> Yeah. >> That it's not just about giving information, it's about actioning it. Uh and then the trust that that builds. I I think it's a in the story about the father who wanted to buy a laptop. Oh, I'm moved. >> Yeah. >> Thank you for sharing. >> And yes, and and India in that lens is a very unique market, right? Because generally, uh you know, we have had an ecosystem where DIY has not really worked to the fullest extent.

15:58 Uh so, uh we have generally seen that the socioeconomic way things have progressed in the country is where things have been done for them, right? So, there is also some sort of an expectation in users' heads on And second, there is some inertia, and then it's also not easy, you know I mean, the infra is not conducive enough for a lot of these actioning to happen in a very smooth manner, right? So, someone needs to do the heavy plumbing and ensure that the actioning is is happening. And once the actioning starts to happen, then outcomes start, for sure. Yeah.

16:32 >> I think that's a like extremely insightful. I remember discussing this internally actually at Lightspeed. We spoke about how in India, >> Yeah. >> uh which is unlike the West, right? If [snorts] you go shopping, >> Yeah. >> uh in the western markets, you are left to yourself. >> Yes. >> You're looking for your own clothes, and you're looking for your own shoes, and you know, you're going through materials. [snorts] In fact, in shoe shops, you're actually checking if the stock is there in your size and your and you're picking out your own size shoe, and but in India, you know, there's a salesman, and you tell him, "I want it in my size." He comes and measures your size, and he goes to his warehouse, gets you your size, and you try. It's such a you know, do it for me market.

17:12 >> Market, yes. >> We even that even when it comes to shopping, somebody's doing it for us. >> Again, powerful. I I love how how you apply it such a something that, you know, seems so simple uh to something so complex. >> complex >> Uh and you're seeing the outcome of it, right? Incredible. So, let me ask you a very easy question. Uh and I'm I'm sure it's a question that, you know, every investor is asking themselves, especially when it comes to consumer tech, right? What what makes the product defensible?

17:39 Right? Why would um someone else not else not build a companion that asks the right questions and solve the problem for a user? Why would horizontal LLMs not end up doing this themselves? Right? What makes a product uh in consumer tech, in AI, uh defensible? >> Yeah, that's that's a a tough one, but I'll I'll try to, uh you know, answer uh this question. So, the way we are building Ulka, uh we are trying to build uh not just features, but we are building an end-to-end consumer journey.

18:14 Right? So, uh what does that mean is on one hand you have a lot of, you know, fine-tuned LLM responses which are based on structured inputs which are relevant to fintech. Uh and of course, a lot of unstructured data that comes through the user. But in addition, if you look at, uh you know, I mean, what we are trying to do, we are not trying to be a companion. We're trying to be an agent. Uh so, what does that mean is we are doing the heavy lifting on closing the loop. So, we are not just telling the user uh what to do, we are doing it on their behalf. What does that mean? That it needs us to build uh or have robust partnerships with players in the ecosystem, with NBFCs, with banks. Uh we are, uh you know, under uh we are in RIA, so we have account aggregator information on behalf of the user with user's consent. So, we we've gone very deep in building this end-to-end journey, which uh we believe that a horizontal LLM will not be keen to get into. In addition, of course, you know, I mean, we believe that this compounds So, basically, if a user comes to discuss credit score issues with you, your lend- their lendability challenges, and as we serve them well, they start to discuss other aspects of their personal finance, right? And that gives us a lot more data on the challenges that are more you know, about the user personas we are targeting and then our ability to serve those users even better, right? So, this there is a very virtuous loop that keeps on compounding, which is some sort of a flywheel, and that in a sense becomes a moat.

19:54 Of course, in addition, we do believe that end of the day, speed of shipping and the quality benchmarks you have are are going to be temporary moats, which we definitely have a very high benchmark for. Plus, you know, I mean, distribution cross memory is what is going to be very powerful, and if I were to explain with an example, we had a user who few months back talked about marriage in his family and why he needed to buy a two-wheeler for the same.

20:26 And with a knowledge memory graph, which is built for fintech, our ability to have this information handy for the user at the right juncture so that you know, there is a very strong continuity in user experience, and when he's looking for a two-wheeler loan, you know, it doesn't come as a push, but it is more contextual and user backwards in their lending journey is what an aha moment for us look like, and I we do believe that some of this is not feasible if you're not like, you know, very much verticalized in your focus, and if you're not looking to build the complete end-to-end user journey.

21:09 >> How much financial information a user would share with your platform versus a horizontal platform is likely to be massively different. >> Yes. >> And I think that that you're saying gives you a unique insight into what you offer. How you sell cross-sell. >> Yes. >> When you do it and as a result what is the experience that the user has? >> Yeah. >> Changes. Of course there is a huge element of offline partnerships and with other financial institutions. I'm curious if you abstract out you know >> [snorts] >> how you're building Ulka to consumer tech more generally, right?

21:43 Like what should a consumer tech founder think about if they're building using AI when it comes to defensibility, right? So if you abstract your learnings from Ulka, what would be some two or three takeaways that a consumer tech founder could have if they're building an AI >> Yeah. >> and they want their solution to be defensible. >> I wanted to refer to Paul Graham's recent tweet where he mentioned that a lot of founders today worry about whether you know their business will be eaten up by a horizontal LLM in future and his response was that you know I mean this is similar to saying your life will be so hectic if you become a very popular movie star, right? So it's more likely that you won't become a movie star than you know you worrying about the pain it would cause you, right? So my suggestion to anyone building in consumer AI is to not be worried about whether a horizontal platform is coming for you.

22:44 What is more important is whether you are solving a unique user problem. And if you're driving an outcome for the user, right? So if you're solving a unique user problem, if you can double down on it if your NPS for the users are very high and if your users cannot live without the problem that you're solving for them, that itself is a huge foundation to build things on. >> Yeah. >> And I'm sure uh you know, I mean uh every problem statement that you try to solve has its own nuances, has its own context. And since you're going to be all in on solving that user problem, your ability to solve it better than any other horizontal play that is trying to solve a million problems at the same time uh will be far, far more stronger.

23:25 Yeah. >> I think that people underestimate uh the power of verticalization, specialization, and depth. >> Uh and I think the deeper you are into a use case, the better your solution is. Uh it's it's constantly underestimated. >> Yes. >> So, thank you for for spelling that out. My last question, >> Yeah. >> if everything went right, what does a day in the life of a an Ulka user look like 5 years from now? >> The power of AI today is to and and at least for uh an Ulka user is to have a relationship manager that understands your needs very well. Right? That uh is giving you a personalized, not just a personalized guidance, but also a personalized solution that helps you improve your financial well-being. Few years down the line, on one hand, we believe that we'll have uh you know, behind our our agent millions of smaller agents working on different nuances uh of a user's financial life through the clock uh without them needing constant supervision or you know, constant approvals.

24:31 Uh and on the other hand, we also believe that this will lead to a world uh where users will have a different benchmark of what they expect from the ecosystem. So, a personalized advice will lead to personalized products. And as uh manufacturers in the ecosystem continue to uh imbibe and and you know, become more AI native, their ability to push solutions which are very hyper personal to the user. So, a product or an insurance product or a wealth product which is not one size fits all, but has been tailored and designed keeping in mind the needs of an ordinary citizen is what the future looks like and we are hopeful that we'll continue to work towards shaping that that future.

25:18 >> Awesome. Thank you so much for for a very insightful session. >> Thank you. Thanks. Thanks for having me over. Thank you.