Inside the Venture Studio Model

Amit: You know there are companies like MongoDB—it's a $26 billion company today. Uh, there's Snowflake, which is a $60 billion company. There are companies like Hints and Hurst which came out of Atomic. There are companies like Dollar Shave Club, Food Panda, TalkIQ, Yelp, AirCall, Medium, Moderna—you know, the list goes on and on and on. And yet, very few people know that all of these companies actually came out of venture builders.

(Start of Interview)

Amit: Hello and welcome everyone. Um, we are back with another episode of Build AI Podcast. Um, now usually we uh talk to uh people who are building AI companies, um AI builders. Uh today we have um you know folks who are building multiple AI companies, and it's a very special one because um you know Vijay uh is deeply involved with us at gAI Ventures. Uh, so it's going to be very interesting when we talk about the various aspects of venture studios, venture builders, and how we are kind of looking at things, but also what is the general perception in the market. There's lots and lots of questions that investors, founders, and people have. Um, with this episode, we are also trying to, you know, educate um and you know share our learnings uh from the last one year that we have been doing this. So, Vijay, welcome uh to yet another episode of Build AI. This is awesome.

Vijay: Thank you, Amit. I'm really excited for this uh episode. Uh I've enjoyed all the really thoughtful conversations that we get to listen to on Build AI and that you host. Uh and so hoping to uh, as you said, you know, kind of peel the curtain back on what's happening in the world of venture studios and uh you know kind of uh shine a light on that for everybody.

Amit: That's great. Uh so I think it may uh be a good idea to start with a little bit of intro. Uh you know uh myself, Amit, I have built and sold a company in the fintech space, Medi, and uh basically have been investing in companies. I've done about 30 uh investments; some of them have done really well. And since last year, along with my co-founder and CTO, we started building gAI Ventures. Vijay came in as an adviser first last year and then, you know, he has come in full-time and is basically leading on. So, very excited to talk to Vijay. And Vijay, if you want to just give a very quick intro about yourself.

Vijay: Oh thanks. It's really again a treat to be here. Uh so, you know, I love working on the problems that we get to think about and with the technology and with the products that we have uh and looking at you know potential markets with exceptional founders. Uh and that is a big theme throughout my career. So I've spent the last 25 years being fascinated by new business models and new technologies and things that people uh can build and have been of service to entrepreneurial leaders when I haven't been a founder myself. Uh so despite a couple startups in fintech and e-commerce, I uh found my way to venture, spent 5 years at 500 Startups, became an executive coach focusing on founders uh as well as an angel investor. Um and then now, I'm really excited to come back to uh venture studios and venture building, uh which, as I can uh describe better during conversation, you know, has changed a lot since you know I started uh getting into that space about 10 years ago.

Amit: Great. Thank you so much, Vijay. Um, I I would like to actually jump in uh and you know ask my first question, which is, you know, I realize that a lot of people in the audience may or may not know everything about venture studios and venture builders. Um so you know I think a good place to start will be that uh you know there are companies like MongoDB, it's a $26 billion company today. Uh there's Snowflake, which is uh $60 billion company. Um there are companies like Hints and Hurst which came out of Atomic. Um there are companies like Dollar Shave Club, Food Panda, TalkIQ, uh Yelp, AirCall, Medium, Moderna, and and you know, the list goes on and on and on. And yet very few people know that all of these companies actually came out of venture builders or venture studios. And you know I think a good place to start, Vijay, will be if you can define what exactly is this structure and what is this model, and how is it different from VCs and accelerators and incubators.

Vijay: That's a great question. There are so many terms flying around and people use them very casually. It's very hard to distinguish between an incubator and accelerator and then what is a venture studio on top of that. So uh a venture studio is still one of these uh best-kept secrets because it has produced a number of companies like you described, uh but it works in a way that is unique and different from all the other things that you mentioned. Particularly because a venture studio really starts at the inception stage. It starts at the idea stage based on either a thesis or the talents of an individual founder. And there is a phase of incubation where an idea can be tested, where product can be built, and where uh the market can be validated in order to then start scaling and supporting that business, and then on top of that providing capital that allows it to go. If you just do the first few things, then you incubate, but you don't really [scale]. And if you come in later to the business and help it after someone else has already created it, then you're an accelerator. And if you only write checks and don't actually play a hands-on role in that business, then you are an investor. There's nothing wrong with being any of those things. I've worked at an accelerator. I've uh spent time in the corporate venture uh environment as well. Like, everyone has an important role to play in the ecosystem, but the venture studio is a unique type of entity that can do many things that catalyze and create uh just innovation and new types of businesses very, very quickly.

Amit: Right. I was actually fascinated by this model when I came across it just like one and a half years back; it's not been long. And I played a uh you know I basically played both the roles of a builder and an investor for quite a while and I felt like um just being an investor was not enough. There's much more value creation I wanted to do. And when I came across this model, it was just totally fascinating, right? So same question to you, Vijay. Like, what attracted you to join a venture builder and and you know start sort of building it?

Vijay: Yeah, I think that ideation and being a thought partner to a founder who is the best person to go and build this business, but who would have to sometimes struggle and stumble on their own or with a co-founder uh for a while before they found exactly where to start and then maybe even later where to get product-market fit—like playing that uh thought partner role is really important. Uh, because it takes uh a certain set of skills in order to unlock what uh that person uh has and where their particular insights can take them, but also provide uh feedback and context to see new things. But it's not enough to just like think really hard. You have to do things. Uh and so that's where I think that hands-on role uh helps define the product and later produces the pipeline necessary for that company to uh achieve momentum and grow and ultimately become a business and not just a science experiment.

Amit: Right. Speed and momentum, I think, is very, very important, and a venture builder provides that. Uh Vijay, you and I hosted this uh very interesting session uh you know in San Francisco where we had a whole bunch of founders from different walks of life. Uh some of them were actually—most of them were domain experts in their respective areas, you know, from fashion to commerce to financial services to accounting. And there is—they already knew a little bit. Uh and there are more and more founders that we are talking to who want to know like, what are the advantages of a venture builder for a founder? Like, why should they? They have many options, right? They can build on their own, they can follow the regular path of raising money from VCs and uh you know uh go through that path. Why should they even think about going and building with a venture builder?

Vijay: Yeah, I think what you described is very telling. Even in a place where there is a high degree of talent density like in Silicon Valley, you might have great mentors and you might find, you know, passionate angel investors and you might have a network of colleagues, peers, you know, who could help you and even be part of your business and join you. But um, even if you have some of those things, the combination of all those things in a venture builder uh is really attractive. Because what we have now is this really unique moment where, for the first time, non-technical leaders who have domain expertise and know something about any of those spaces—whether it's accounting or fashion or what have you—can build AI companies. And they can be the CEOs of technology businesses. Uh because what we've done is you know created a process and, you know, others have shown it's possible to not just buy code but build real product uh in a short time that is ready for market and is accessible uh to uh you know to your customers and can be enterprise-grade very, very early. Uh so I think collapsing the cost to build means everybody is going to move earlier in the venture ecosystem. Uh but in particular, it creates this unique opportunity to do things quickly with exceptional people. Uh and that's something a studio or venture builder is extraordinarily well positioned to do.

Amit: Right. That is so true because um I I think what we realized last year when we started building our first AI product, just Tella, that it takes significantly less amount of engineering resources and significantly less cost to, you know, build a company. Um but then, you know, in my investing experience I have seen there are a lot of people with 15 years, 20 years experience in banking, financial services who would struggle finding a CTO or, you know, like a great technology team. It's so difficult to hire, right? Even in countries like uh or or regions like, you know, the Silicon Valley and Bangalore. And um I guess um they have great ideas, they understand the workflows, and this was one of the missing elements. And you know, at gAI Ventures we uh you know we are offering that. And it seems like um if you actually today understand uh industry really, really well, especially from uh you know those complex workflows and where the data resides, what integrations are there, and you have a lot of relationships and, uh you know, uh basically you know people in the industry, and if you were able to then build a solution to solve some of the problems that you have lived with uh yourself for a while—it it seems like uh you know it's a pretty good um you know way to build companies.

Vijay: Yeah. If you look at what you know accelerators value, it's finding somebody who uh has a unique insight, who knows something that other people don't. Uh who maybe have actually experienced the problem themselves because that's the only way in which that person is going to um have the ability to solve a problem and address it in a unique and uh important way. Now if you take that experience that, you know, maybe somebody most likely on the younger side has experienced early in their career and then you multiply that by decades. You either have people who accept the world the way it is or others who are highly entrepreneurially minded who are, you know, still want to a crack at making things better, or maybe this is their second or third startup and now they're going after this opportunity with like just compounded knowledge and skills uh developed over time. Um and you know you and I have talked about how there is this uh disconnect between the data that tells us that the average unicorn founder is 38 and the pattern matching that Silicon Valley is attracted to, which is looking for you know the proverbial uh 20-something founder in a hoodie who today, you know, uh it might be ex-Stripe or Palantir—exceptional companies to have worked for. Obviously, a few years ago that was Uber and Airbnb, and a few years ago it was someone who had like clocked in a few years at Google or Meta, right? Like, and this will continue and in three years we'll be seeing you know alumni of another type of company, maybe OpenAI, right? Like, get uh and that's that's great and that's a particular way to invest. But what we know is, with a different configuration of capital and resources, we can help that person who is statistically more likely to succeed in many ways to be successful. Uh and that's something a studio or a venture builder is uniquely able to do.

Amit: Yeah, that is so right. I mean um at least like one of the insights that I drew from my own investing experience of uh you know 30 portfolio companies that I have. A bunch of those people are actually people who worked for 10, 15 years in Visa or a bank like Barclays or something. And um there are areas in financial services, especially because it's a regulated industry, there's compliance, there's so many things that uh once they were able to uh build a decent technology team and they were able to build a decent technology solution, there was no stopping them. And some of those companies have outperformed everything else. And it's funny that when I—you know, I've actually invested in about—my portfolio has about nine YC companies. Six of them I invested before YC. And one of the things I saw is that almost all of those cases, they want to pick up somebody who is in their early 20s or mid-20s. Um and uh we know this very well from our investing experience that those are not the only companies which will become successful, right? There are people who somehow have not been given that opportunity or chance, that you know they are basically in their 30s or 40s. Um they know one or two particular problems really well because they've spent like 10 years. Uh so for example, doing accrual accounting as an example, or uh you know basically looking at construction management industry and going very deep into that. And so they know the problem, they have the relationships, they know the complex workflows, and they know what has to be done to solve it. It's just that they never got to a point where they had a CTO or engineering team which could help them build that. And why should they not have that chance, right? It's like that question that we keep asking ourselves all the time.

Vijay: Yeah, exactly. Like, if that's all that's missing, what if we could do that? And what if we do that in a way that is, if not scalable, at least spreadable and repeatable across many different businesses? And that's what you get in a studio context. Uh and I too have invested in companies uh that were later accepted to YC or have gone through 500 Startups or a program like that. Um and they're great. They're led by wonderful people. Um and the ones that seem to be uh chugging along with the most momentum are ones where um there was some industry experience and some knowledge of how to keep growing and scaling and uh be able to succeed. Because if you do come with many years of that industry experience, you have not just the ability to call the first four or five people on day one, but also you enter these high-stakes environments with trust, with people knowing that you know what you're talking about and that the risk of taking a bet on a startup, particularly in this B2B context, is actually much less than if you were dealing with uh folks that you know were learning as they were going.

Amit: Yeah, I would like to share an example from one of our own portfolio companies, you know, Fast Tracker AI. Most of the times when we are going to um and talking to RIAs [Registered Investment Advisors], it's not really about the technology that they have a lot of questions. It's not even about um you know like which LLM you are using or uh you know like how many engineers do you have. The questions are actually around like how much—how much do you know about financial services and regulations and compliance? Do you have SOC 2 Type II, right? And how do you sort of manage the data in transit and at rest, right? So those are the questions which actually have helped us to convert um a lot of sales. Uh so as an example, in Fast Tracker AI, our founder has spent like 14 years um in banks and financial services, you know, seven in the US, seven in uh Europe and Asia. And so when he's talking to these RIAs, which are really financial services guys, right? Uh and talking to them about what is our solution, why we have built it like that, and then it goes into questions around regulation and data and SOC 2 Type II, the answers are very trustworthy because they know that there's another guy who actually came from financial services. He understands the importance of managing customer data really well and you know uh uh about end-to-end encryption and about security of all of this uh data, right? So that is something that we found very unique: that there are industries like financial services, healthcare, um you know we have um—I think it's not just about the solution, but it's also about who is selling it, how they are selling it, and how it is being adopted.

Vijay: That's right. So you have to be an industry native in many cases to even have the right conversations with people, and and I think that's a good example. Yeah.

Amit: Yeah. Now uh you know, as we always do, we want to be very honest uh to the topic and uh also question everything, uh right, um and provide an opportunity for people to listen to and get some of the answers to questions that they may have. Right. For example, one of the questions that founders would have and um you know even uh sort of later-stage investors would have is: why should a founder part ways with 20, 25% equity? Um in our case, I mean there are venture builders, venture studios, um you know which it all started with, you know, like 20 years back with, you know, studios was like um Rocket Internet and others.

Vijay: Idea. Yeah. Rocket Internet.

Amit: Yeah. Yeah. Idea and and then um even till late, there are many venture builders and venture studios who would take 50, 70% equity. Uh our model is slightly different. We are trying to make sure that the founder has at least 50% equity at Series A level. Um so we are basically taking 20, 25% equity. But even then, right, I think it's a fair question to ask that why should you part ways with that much equity?

Vijay: Yeah. If you're a founder, you don't want to be diluted uh and it's very um you know natural to compare different options. So, you know, as you uh and and many of our uh you know, listeners will be familiar, like if you are going through an accelerator, you're probably giving up 7%, sometimes up to 10% uh equity uh in an accelerator program in exchange for some check of some amount between 100,000 up to $250,000, right? Uh and so the question is, what else is in that? Uh and so people should read these provisions very carefully, understand MFN clauses, understand you know what happens if they don't raise another round, what are the types of investors they will need to go to downstream. Um all of these things uh matter a great deal. And the key thing is to ask ourselves, you know, what do we want the future outcome to be so that we are default investable and that we are default alive and, even dare I say, uh thriving and profitable and growing, right? So uh what are the conditions to derisk the business and push you so much further? Right? And that's something that studios are good at, right? Uh similarly, what is it that you uh you want the cap table to look like at Series A? Uh and because that's a real inflection point for the company, because from there you're truly scaling and not just um you know developing in the earlier stages. So if, as you pointed out, you're still owning half the company at Series A, you're essentially getting to the same place but just with a different approach and maybe in less time, and maybe that uh you know allows you to win because you uh are ahead of others in the market and have built it uh more quickly. So I think that's a very good question and every uh founder should have this conversation with a venture studio, with an accelerator, with an angel investor, uh to understand you know what are the things that are at the surface level and what are you know some of the um the provisions underneath, and to you know ask themselves the question, what does this start to look like um you know in a couple [of years]?

Amit: Right. It does not matter where you start, it matters where you end. Right. So um yeah, I think that's uh yeah. And I think it might be a good idea to also sort of discuss what is our founder and idea selection process at gAI Ventures. Um because you know I think it's all of this is very different and some of it is also very new, and we also figured out our own uh ways as well. Yeah. But one thing I've seen is most of the successful venture builders have a system, and I know that you have talked about it uh a lot with us internally that everything needs to work as a system because you don't want people uh to sort of you know take decisions outside of that structure. So it's kind of important that we follow a system. So maybe we start there and then we sort of you know bring it to what is our um process to select the right ideas and and founders.

Vijay: Yeah, it's—I'm glad you say that because you know intuition is good, and when our gut feelings are informed by experience, that's really powerful, right? There are people who invest in a company because they just know, they believe they have conviction. But where does conviction come from? Right. Conviction comes from uh either like some overwhelming probabilistic uh analysis that you know this is going—that it's clear an outcome is going to occur, or in the absence of that, uh a great set of belief around uh a team, around a market and a product that you believe they're going to be exceptional and they're going to be successful. So if you look at it that way, then what we need to do in our jobs is to talk to basically uh a couple hundred founders a year. uh we will talk to a couple hundred individuals, uh all of whom are talented enough to uh be in conversation with us. Uh and what we want to do is like understand, you know, are they um people who have that unique insight and knowledge of a space? Are they ready to be founders? Uh do we have an idea about why they would be good for us to work with? Uh and are they also, you know, like ready to take this plunge? Um and so we may filter that down from, you know, a couple hundred to working with ultimately a handful of people a year. Uh and then the key is to really try it out. What does it look like to uh to learn quickly, to explore a space, to build something together? Um and that's where we have a four-week sprint process to start with an opportunity area, uh to look at how to analyze the market, uh to build some version of product, show it, uh get feedback, learn, figure out why something is a bad idea—like that's really important, knowing not what not to do. Uh and then at the end of that, uh ideally having a design partnership uh or even a POC [Proof of Concept] that would allow you to then say, okay, there's a first market here. It may not be the biggest market, may not be the final market, but there's a first market and there is a path to that larger market that has, you know, a 10 billion or more uh total addressable market behind it. Uh and then from there we want to make the decision to uh incorporate the business, uh invest uh at that stage, and subsequently um you know support the buildout of the product and eventually a tech transfer, serving ultimately as a technical co-founder uh to that startup founder.

Amit: Right. Very well said. And uh you know some of the numbers are fresh in my hand. Just to add to what you said, uh you know I would just like to share some numbers that we recently uh published investment thesis in a couple of different areas. And then uh we also used various tools like LinkedIn, jobs, etc. to uh you know uh reach out and get interest from potential founders. We got more than 150 uh plus applications, um shortlisted that to you know about 25 plus, started having discussions with them, and then uh brought it down to a meaningful number. We actually met some of them um in in San Francisco uh last month. And then um we have started a 4-week uh customer development program or the initial program, as we call it internally, uh with about two of them um and and that's currently going. So that's how you know like our funnel uh works. And as I said, like because there was an investment thesis, so we we have these ideas or use cases defined in particular areas like AI in accounting, uh AI in lending, AI in various spaces. And then we would—we would have a founder who got interested and attracted to this because they also have a similar idea or a uh you know or a problem-solution set that they have been thinking about in that particular area. And while our use cases are slightly, you know, like 30,000 ft level and 10,000 ft level and 5,000 ft level, but this person, because of his domain expertise, comes in with a very, very niche uh problem uh that they want to solve. And their understanding of the landscape is at the grassroots level.

Vijay: Yes. Yes. Exactly. And it's very interesting, like how secondary research and talking to people versus actually spending 10 years in a particular niche—the difference that it makes, the product-market fit that it can bring to the table, right?

Amit: So absolutely.

Vijay: You know, I've spent years thinking about product-market fit and how you know it's about trying a whole bunch of things and then you just kind of get it. I've heard famous VCs say, "Oh, how you get to product-market fit, that's a kind of black magic," you know? Like, and we know that, you know, you're crossing the chasm, you're doing different things um to to learn. But it turns out some people just come into a space and they're ready to win. And it's really exciting when you see that, right? And that goes into that unique development of conviction. Because um you know, to devil's advocate here, if you're trying to attract people to build a company together, you might attract some people who are you know not the ideal uh folks to build with. Uh but you try and see with this thesis approach, saying like we care about this space, and you will find the person who is maniacally and unreasonably equally interested and or even obsessed about that uh space. And that person is going to do something extraordinary and move really fast. They're going to send out hundreds of emails in a week uh to try and get uh someone to uh take a call and demo a product. So like those types of outcomes are really exciting. And that's also not what we're trying to necessarily test for, but where we know these extraordinarily talented, high-energy folks are, and we want to give them as many tools as possible to succeed.

Amit: Great. Vijay, you uh you just said uh devil's advocate. I would also like to play devil's advocate for a moment. Um I have heard this argument being made by a couple of VCs, you know, I was talking to that, "Oh, you know what? Like building startups is a very irrational behavior. Um you know, fires are burning all the, odds are against you." Yeah. So you need people who are not very rational, and usually these people are in their 20s. And I was like, hold on a second. Is this like again pattern matching that they have to be in their 20s to be like that? And while even myself, right, this is my third time I have started a new thing and um I don't feel like the person has to be in their 20s to be having that kind of passion and fire in the belly to start something. Um and so how do you sort of explain that? Because YC and a lot of people believe that people have to be in their 20s to to be mad and to be so passionate? Does passion restrict to some age and is fire in the belly restricted to some age?

Vijay: Right. You know, again, I'm reminded of like uh this old CB Insights t-shirt I had which said, "Data is greater than opinion," right? Like, and the data is that um you know the median age of a decacorn or unicorn founder is 38. Uh the data tells me that um you need to look at people who are doing stuff and not just talking about their passion. They're living that passion. And that might have been that they were starting multiple companies. It might have been that um they were frustrated so much by what they saw in the corporate world that they decided uh they wanted to do something different. And they are so obsessed that they're willing to go out and um ideate and collaborate and build and sell uh for the sake of giving this a shot. And so, you know, uh passion is both like ideas and it can be sometimes youth, but mostly it is like the energy that you bring to doing things, right? And we see all these very passionate people all the time. Uh given that you know we are talking to so many potential co-founders. So yeah, I completely debunk that myth that people have.

Amit: In fact, another example is that I'm sure you you know about this side also is the interpretation of data, right? Like um who is talking about it and how much, uh you know? Like for example, media always talks a lot about certain things and so people believe that that is a fact or truth. For example, m uh during the World War, I think they analyzed uh I think it was Japanese planes or like uh I don't remember.

Vijay: Yes, the American bombers.

Amit: Yes, the American bombers, yes. They analyzed that, okay, which ones were surviving and where were they hit and where were they not hit. And then they thought that because the ones which were surviving, uh the bullets were hitting a certain place, and so they said like we have to make it stronger in those places.

Vijay: Absolutely.

Amit: Turns out that that data was skewed because the weakest point was in the planes which were destroyed or brought down.

Vijay: Yes. No, you're spot on. Right. So it turns out when you have bullet holes through the wings of the airplane, it can still fly. Why? Because the engine is still, you know, turning the propeller. And that's exactly what you need to reinforce: keep that engine, you know, flying the plane. Uh and so yeah, this is a great example what you're describing of survivorship bias. Uh so yes, there is a lot of survivorship bias out there. Um and it's our job to separate signal from noise, to try and uh address all of our biases, you know, of what we want to be true and what you know past uh you know evidence might be pointing us towards. But we have to ask ourselves, what can be different now?

Amit: Yeah. I think just to summarize so far, right, what we have discussed, and then there's a couple of other things I wanted to discuss, and maybe you have uh some other points that we can cover. Just to summarize so far: basically three of us—myself, you, and Kushel—we have built companies, we have sold companies, we have worked at VC firms, we have invested in you know like more than 50 companies combined. And we like—there are three, four things that we believe in, right? One is that there's no one way of building companies, right? Whatever media narrative has been set up that the only way to build companies is you go and do an angel round and then a Series A, B, C, and the VC route is the only route to build companies. I think uh that is not true. Like there are many different ways of building companies, and we gave the example of um you know a bunch of companies in the beginning. Um they're all glaring examples of that. There's enough and I think there are more than uh 70 companies which have been built by venture builders or venture studios which are really successful billion-dollar, billion-dollar-plus companies. Um so I think that that myth is totally sort of debunked. The other thing that we believe in is that this is the first time in the history of software that domain experts can actually build tech companies and especially AI companies. It is possible because, one, it is easier and cheaper to build technology. Um and um also with platforms like gAI Ventures, you have a partner which can help you ideate and build um the engineering platform and the AI engines and so on. So that is—the third thing we believe in is that sales and marketing actually has become even more important than ever before. Now I think distribution is the hardest part because you could use Cursor and Codeium and assistants and some engineering effort to build and come up with like you know products. But how do you sort of convince the customer, especially in industries like the ones that we are focusing on like financial services, commerce, enterprise productivity where you are dealing with companies who cannot take a vibe-coded product? Any production-level things which have been evaluated is going to be taken care of. Um so it's very important that um I think domain expertise plays a very important role. Like people, just as you know they say, right, people bankers buy from bankers much more easily than they'll buy from anybody else.

Vijay: Oh absolutely. Yeah, that's really true within a number of industries, and that's where you can have an unfair advantage as an industry expert, as someone who has domain expertise. Uh what else do you believe in? What else do we believe in?

Amit: I think we believe that, you know, there's three really interesting areas that are changing in front of us right now. And those are, you know, as you pointed out, fintech and financial services. The second is in what we could call enterprise productivity, which covers a lot of industries that have really complicated ways of, uh, or structuring data and telling people to do things where AI has a big role to play. And that can cover healthcare, cover law, professional services, a whole number of different areas. Uh but you know that is like really interesting um because it requires somebody to understand what actually goes on in your business in order to address it with AI, which will happen, uh but again it provides an unfair advantage to domain experts. And the third area is commerce, because we see it as very data-centric and really integrated to a number of other fields. So retail to payments to connecting to supply chain and logistics. And in an age where all those things are feeling more volatile than ever, um we think that more sophisticated tools are going to uh come into play.

Amit: And to that point, uh there was a report recently from MIT which said 95% of enterprise AI efforts have failed. Projects have failed. Uh what people—people just went with the headline and they didn't dig deeper. Uh the report also said that these results are coming from the fact that a lot of these were internal initiatives, and we know the innovator's dilemma. We know how hard it is for large companies to build technology and build like really good experiences. A lot of them were internal. And they also said that 67% of the times when they use a third-party AI vendor, uh AI product, it was successful. Uh so, you know, again going back to, I think there's so many sort of misconceptions because of the narratives out there and how the media projects it. I think it is our job to sort of look at it deeper. At gAI Ventures, we have this very sort of counter-narrative that companies can be built by anyone and they can be built in any number of ways. There's no pattern. Whoever said that there's a pattern of successful large companies, you know, it's like we don't believe in that.

Vijay: Yeah. I mean, we've seen this before. People saying you can't build a technology company outside Silicon Valley or saying that, you know, meaningful business models have got to be, you know, in software, not hardware, or in hardware, not software. Like, there's always a pendulum swinging in some direction and someone who wants to take the argument that it's never going to swing anywhere else. Uh but you know where we do recognize there is this opportunity is because of what um AI does. It takes these different elements of building a business—and in particular design and development—it democratizes it, collapses the cost of doing that. It makes it possible to start and build things. The hard part is distribution. Distribution is going to get even harder. Um and so what is your unfair advantage to win there? The path to moat might include data, but data is often, you know, something that belongs to the incumbents. Um so I mean, that is a question. Do you think that data can be part of the AI strategy for you know startups to find competitive advantage?

Amit: Yeah, I think um data is actually one of the most important things right now. If you were to think about like three pillars of building an AI company today, um I think you have to structure the company in such a way that you are generating either—you're generating data which is proprietary and you are basically generating data because the users are using the platform and they are sharing the data and the way that they are using it, you get all of those inputs, and therefore your platform keeps getting better and better. Or you have to go and partner with incumbents, uh people, companies who actually have the data. So they will work with you because they need your AI agents, agentic framework, autonomous agents, what have you. But then they basically control the data, and then working with them you can actually build a solution, right? So uh let me give you some examples. So let's say in the RIA industry, um there's tons of software which is already there, right? Financial planning software, CRM and all that. It's all like it's all over the place. There are different vendors. These systems don't talk to each other, and so there's a problem of um how the data flows from one system to the other, how it is fed into one system, what is the output, and then it goes to the another one and another one. One of the things that we started doing there was first we launched like couple of AI-native solutions like a meeting bot that joins the meeting and does very specific wealth management notes, uh and then it gets transferred to the financial planning software and CRM. And similarly, we built a document processing engine specific to again um you know things like brokerage statements, tax statements, K1s and so on. Um which is something that uh earlier somebody would be punching into uh an Excel sheet and then transferring the data to uh a financial planning software, CRM, and and then now what we are building there is a customer onboarding journey which basically takes care of um the data flows between the different systems, right? So so what is happening over a period of time is that we, because of all the data that we are consuming and also the data which is generated by you know the use of our platform, uh you know we are able to do reinforcement learning based on that. And so our system is actually sort of improving over a period of time. And that is like one way of uh you know sort of dealing with uh data; it's very, very important. And the other way that I have seen is um something like what Harvey AI did. They basically said like we will go and partner with law firms, and not just like small, mid-size law firms. They went after the top 100 law firms and somehow magically were able to strike deals with a lot of them, work with them, and they are basically feeding—they are basically consuming it directly from the host, right? These are like the largest law firms with thousands and tens of thousands of cases. So that is the other way. And yeah, to your point, absolutely important to figure out your data strategy.

Vijay: Yeah. I think that's really useful for anyone building uh today. Who knows, my models need to get better. I need to keep training. We need to be smarter and also we need to have ways to keep our customers, uh because data creates the stickiness uh when it comes to retention.

Amit: Yeah. Because you know, GPT-5 is great and all these foundational models are getting better. You know, I would not bet against them. And they have worldly knowledge. They have been fed on like, you know, everything that's out there, from books to uh you know, Wikipedia and and GitHub and Slack and Reddit. But then the biggest thing is that when you're talking about vertical AI companies, um when you're talking about let's say something in compliance, uh let's say FINRA compliance, then you need to uh basically have very specific data and you know you need to build something which can solve that very complex workflow. Um otherwise, um what we have seen—the MIT study also said that in a lot of cases, what they explained that a lot of failures were caused because you were using a generic foundational model and were expecting that it will give all the answers correctly, but the models hallucinate. And vector search, you know Google just published a paper that they had some issues. So you still need a lot of stuff before and after AI as well, right? For example, you need to understand systems engineering, you need to understand evals, you need to understand to test these systems. And just in the olden world, we used to write those test cases and do the testing. Now, uh I think evals are very important. You need to create like hundreds or thousands of cases where you say if the input is A or the question is A, then what should be the answer and how far—what is the deviation. And then, you know, now our CTO Kushel basically is implementing things like LLM as a judge, right? So it's not just important to get the output from the LLM, but you also have another model or another AI engine which is acting as a judge and is figuring out whether the output was uh you know correct or not and and sort of scoring and marking it.

Vijay: Yeah. And that's very powerful because that's how you actually build products that people trust, that they find reliable, that they refer to others, that you know creates that engine for growth.

Amit: I guess a final question is, you know, when we think about some of the exciting AI companies that are getting headlines now, which are, you know, small teams that are able to generate millions and millions, some tens of millions of revenue in the matter of like a year or maybe just months, um you know where does this end and where you know can we see bigger, faster, leaner teams? What's possible?

Vijay: Yeah, I think it all started uh with horizontal use cases like uh coding assistants and coding IDEs has been the biggest, most successful, has generated the most amount of value between Cursor, Lovable, Webplate, and so on. Um and then now it is going more into various niches within traditional industries, what we call as vertical AI companies or B2B companies. The reason for that is very simple: like if you think about coding, um there are like 10 or 20 dominant languages, the format, um uh you know and things you know for a particular language, let's say like Java or Python, there's a syntax, right? Like you know how to write that code. And a lot of this data was there on GitHub, just fed into the foundational models. And so the output, because the entire thing is so structured and there was so much of data and there are like defined syntaxes for languages, it was easier to nail that use case and the output is very, very reliable and it works most of the time, right? So so those were the first ones, right? So if you think about it, I almost feel like people have to just follow the trend of reliability and production-level AI software. As models are getting better and as we are able to build very specific solutions in niches, automating complex workflows, and a lot of data is getting generated in those areas, you will see more and more reliable systems and production-level systems to those complex problems. Harvey AI is a great example uh because they got access to all the data. One of the reasons was the co-founder—the main founder of Harvey AI was a domain expert. He actually came from a law background, and so he knew a lot of stuff right from the start. So yeah, that's what I feel like. As these uh AI systems automate complex workflows in these niche areas, the reliability improves and we have systems that can be deployed at a production level. I think that's where the biggest opportunity lies. And um it could be like you will see like uh things which are not possible before because the technology was just not available before. Now AI makes it happen. So things that we always thought were not possible—right now we will see very, very niche areas where probably a billion-dollar company can be. That will be really interesting.

Vijay: These niches that were previously thought as too small can be, uh you know, unlocked and they can grow to be very big because of what AI can do, what agents can play as a role. uh and yeah, that is a place for experts to go and explore and for studios and other types of folks who are, you know, serious about and committed to those spaces to support them.

Amit: Right. It's all very exciting and, you know, it's um yeah, it's a very interesting time to be building in the AI space. Uh thank you so much, Vijay, for your time. It was a super interesting conversation.

Vijay: Of course. Yeah. And let's do it uh another time as well.

Amit: Uh but thank you so much and um you know, looking forward.

Vijay: Yeah. Onward. Great.

Amit: Thanks, Amit.

Vijay: Cheers.

Amit: All right. Bye-bye.

Vijay: Bye.