Microsoft on AI Startups: Why Speed, Taste, and Trust Will Define Winners | Decoding AI

Speaker A: You've got to have taste.

Speaker B: Absolutely.

Speaker A: You know, you've got to have a great app. I think you can't outsource that to an agent. I don't know that ever before. We've had the technology and the capability to improve the margin by lowering costs at the same time, invent new products and increase margin.

Speaker B: Adoption of AI in India is a much faster pace than quite a lot of the world. So, you know, you are seeing AI enthusiastic buyers starting to happen in India so that you can build something for first test here.

Speaker A: Ignorance is the unlock. I look at like my kids sitting down and vibe coding something because, like, they have no prior. And when you have no prior, you think totally different.

Speaker B: Hello, everyone. I am Shekhar Kirani, partner at Accel. Welcome to this Seed to Scale podcast. We have a great guest today, Jason Graffy from Microsoft. Jason, thanks for coming to Accel and, uh, you know, in your busy schedule in India, spending a little bit of time to our audience of almost, I would say, 10,000 startups in this network. Wow. And quite a lot of them from India. And I'm hopeful that they significantly benefit from this conversation.

Speaker A: Well, wow, that's a lot of pressure, but thank you. It's great to be here. It's great to be in India. I always get energy from when I come to India. Just the people, the culture, the founders. It's really energizing for me. So I thank you so much. And Shekhar, thanks for having me. And the partnership we have is great. And so it's nice to get to sit with you and learn from you today too.

Speaker B: Wonderful, wonderful. Jason, you said, uh, this is your sixth trip to India. You know about that? Yeah, yeah. And then. But you haven't come in the recent years. Maybe it's just to kind of break the ice. What is the delta you are seeing from your last trip? Oh, wow.

Speaker A: Uh, it's a great question. You know, last trip was pre Covid.

Speaker B: Yes.

Speaker A: So this was my last trip. Last time I was here was the last trip I took before COVID hit.

Speaker B: Wonderful.

Speaker A: And so I was here in January, and next thing you knew, the world was shut down. So. So the delta between last time I was here and now is astounding. Right? It's a massive difference. Um, most notably, you could easily say how much the pandemic changed everything. And just the growth in a city like Bangalore is astounding to me. And so it's exciting to the new airport and things like that is very exciting. But then there's this whole thing of AI that's evolved since I was here last and so that's even more exciting. And as I travel the world, I can honestly say there are very few markets where you see such sophisticated founders, such impressive product ideas, such what I would call complete thoughts in what they're doing, uh, as you do here in Bangalore. It's a hot market.

Speaker B: I feel you are in a very envious position because you pretty much know exactly what is happening in some of the super fast scaling companies across the globe. You are responsible for working with them. In addition, you are also inside one of the super fast growing company Microsoft, where you see today, tomorrow and next five years planning and efforts that's going on and the enthusiasm, this is a uh, golden opportunity to get whatever possible from you. For our young founders who are just starting their journey in India, building for global or building for India, how AI can enable them to kind of successfully execute their visions and roadmap. So thank you for uh, uh, this conversation. Uh, maybe let's start with this overall ecosystem. You briefly mentioning saying that these ecosystems are forming across the globe. Obviously Valley, Bay Area, everybody's aware of it. But you were saying about that starting to see in different parts of the world. Maybe a quick update from you how each of the ecosystems are doing. Then we can say what is ecosystem in India can contribute.

Speaker A: No, it's a great point. Uh, there's definitely a concentration of what I would call energy or development occurring in a few cities around the world. You could even start at a country level and say the United States is a clear leader in terms of the LLMs and a lot of the work that is coming out of the west coast of the United States. You've also got this incredible amount of energy and development happening in China. Yes, one of the anomalies on the planet to me continues to be Israel and Tel Aviv. Despite everything happening in the geopolitical landscape, they continue to crank out startups and value and ideas that is unlike anyone else. And then the next one, I would say the fourth one would be India, where you just have this incredible ecosystem. But as we talked earlier, you've got this community that is developing where founders, I like the term you said about paying it forward. You've got some very great success stories that happened maybe in the cloud era and now people have spun off and start to do their own thing. And I noticed there is an ecosystem within ecosystem. So if you start at the macro level, there is an AI ecosystem that is starting around the world. And it's even things like some of the founders as we're seeing from, you know, OpenAI have gone to start Anthropic and other labs. You've got DeepMind had an incredible impact on the world where when I go to London I meet so many founders that came out of DeepMind. Here you were listing a ton of examples.

Speaker B: Flipkart, Freshworks and many of these companies now Swiggy Urban Company, many of the people who have built the scale companies, n minus 1, n minus 2 leaders, teams are coming out and saying that okay, I can also build, I have new idea and so on. So that before it was much harder to imagine what is possible. Now there are enough examples so people are looking for a great idea if they land on it, there is a possibility and examples there is enough money in the market, well that's it too right.

Speaker A: There's capital. Uh, so as we see this macro ecosystem evolving, I'm seeing this high concentration in these countries then down to a city level. Right. So if you go from the country to the city, it's San Francisco, New York, Tel Aviv, Shenzhen because of the devices that are hardware, Bangalore here. And now I see London starting to really come rise up. Singapore, I think Australia might be one of the next. But where you see concentrations of capital you start to see founders flow and ideas flow. And then of course over time they just start to spin more out of each other. So, so the ecosystem is evolving, but it's evolving much faster than I think we expected.

Speaker B: Unbelievable. So what I see was our first generation of companies were global ideas. Mhm. How do I build better, faster, cheaper in India? That's where the Flipkart was. E Commerce. E commerce was happening somewhere else. It has to happen in India.

Speaker A: I will do it with an interesting spin to it. Right.

Speaker B: It has to be adapted to India. It does, it has to change. So for example when Flipkart started they have to do all the way logistics to digitization of their supply chain to getting into work on mobile and desktop payment infrastructure, uh, delivery infrastructure, everything, warehouse, everything they have to do. Therefore it took a lot more money, lot longer to build it. And we as ah, like a first early stage investor there, we saw how hard it was. But the next set of companies that started like Swiggy or urban company scaled much faster because the infra was built for sure. Right. But now the new age companies that are coming in are running much faster because the infrastructure of everything is there. 500 million plus phones in India in terms of payment infrastructure exists, in terms of logistics infrastructure exists. If you have a better idea, you can Build something. Now, with AI, which we talk about, you can actually leapfrog. You can build something extraordinary today.

Speaker A: Well, I mean, think about the old days, right? To start a company, you would have to actually go buy servers, put them somewhere in your basement or garage was the old story. Right. And then write your code to start your idea. Then the cloud came along and it was like, now you just needed a credit card, and it actually. Now you don't even know how to code. You can just. You need a credit card and you need a great coding interface, a vibe coding or some sort of generative coding, uh, ide that all of a sudden allows you to go from idea to code to product even faster. So I think it's interesting, when I think about India, I remember when Flipkart was so innovative and what they were doing was so incredible and fast from a US Standard. Now I look at Swiggy.

Speaker B: Absolutely.

Speaker A: And I think, how much faster can it get?

Speaker B: Right.

Speaker A: Um, but it is really impressive to see how they build upon each other and the ecosystems and ideas build. And I think that's what's exciting. As I see companies around the world. We've talked about it before. Uh, you know, some people are saying, is the market made? Where are we in? You know, some will say we're in middle innings. I don't even think we've reached middle innings. I think we're still in the early innings, and I think we're, you know, there's so much more to come and so many more ideas that are going to evolve. And I think it's that rapid innovation we're seeing. You know, some ideas stick, but then model evolutions will cause some to, you know, lose their idea or have to go out of business and reinvent themselves. But it's happening quickly.

Speaker B: Yeah. So question. That's the fundamental question a lot of founders are contemplating with the AI, where the cost of writing product code and making it work looks like drastically coming down super fast. Is the market taken already or market is wide open? Yeah, that's like the. Where most of the people are worried about, uh, maybe a quick how you think about it.

Speaker A: Yeah, yeah, easily. I mean, first off, if you look at the progress that has been made and the evolution in such a short amount of time, it's almost impossible to say any market has been made or anything has been defined. One thing that is apparent is that it costs a lot of money to build models. And so to have a model, you know, you definitely have to have the capital. And that is something where I can kind of safely say that is a market that is being made by a few and it probably will just do the cost of capital and things like that. But all the different applications of it and how it's used not only in consumers lives and in enterprise, I think is wide open still. And I think it's that phase where you're going to see rapid innovation, you need to see failures, you need to see people crack the code. I think about almost like an Uber moment where we had cloud and we had mobile. But it wasn't until Uber came along

Speaker B: and saw that how to stitch together completely different.

Speaker A: Right. And invented this new paradigm. I don't think we've had that yet. There's no chatgpt you can say is an amazing consumer application. But I don't think that one that we apply to our life in the same way as Uber has changed our life has come yet. So I think there's a ton of opportunity there. I think in the enterprise, um, you know, pilots are easy, implementation is hard and I think we can talk about that a lot more too. Yeah, but no, I would go back to saying I think we're in early innings, I think we're learning a lot. If you just think about going from chatbots to um, now agents and where that's going, I think it's too early to say something, anything is done.

Speaker B: Maybe um, let's start with that. If you look at traditional industries, say automotive or uh, travel, hospitality, healthcare, where are your food processing end to end construction, these industries, how AI penetrated are they and is like where are they in that journey across the globe when you meet these.

Speaker A: Uh, yeah, well, I mean I'd love your thoughts here too. I don't think they're penetrating. You know, I mean think about food service, think about the opportunity now of what you'd be able to do with just food cost controls with computer vision and being able to inspect and so on, not only inspect but control the portions, uh, automate your ordering systems and things. I mean the way you could really help improve uh. There is an interesting thing I've observed where we're kind of in this golden age of margin to kind of abstract it from your question, a little bit of industries where it's been very, I don't know that ever before we've had the technology and the capability to improve the margin by lowering costs, by dramatically lowering costs and so many different horizontal and places in the business at the same time invent new products and increase margin. And so you've got this unbelievable opportunity to increase Margin by reducing costs and then introduce new products at a lower cost that I don't know that we've ever been in that before.

Speaker B: You uh, usually would get one. You get one like either you get increase the revenues and you have to kind of invest in that forward invest and get it or you'll decrease the cost. How when things world move from on prem to cloud we were able to reduce the cost. Got the scale advantage. Uh but here is the first time as you are saying is you are not only able to reduce the unnecessary what I call as friction in the system totally and you can take that out using AI so you can get the margin but you can reimagine how you serve your customers as well with new products so that the people who can absorb AI and deploy for their customers can see uh, like a phenomenal double, you know, reinforcing loop to kind of scale rapidly.

Speaker A: Well and the ones who are doing this and taking advantage of it will win.

Speaker B: Right.

Speaker A: I mean Shaker, you've seen this. You know we did a Microsoft for start for um. Actually M12 our investment fund did a study and it showed that more startups are getting to profitability faster with less employees.

Speaker B: Yes.

Speaker A: Than ever in the history. Because they don't need all these different in house functions. Right. They've got AI to really get them to where they need at this early stage. So the ones that take advantage of that will leapfrog the others. And so it's interesting I'm sure you're seeing where early to later stage I guess. Later early stage startups are now trying to reinvent themselves.

Speaker B: Correct.

Speaker A: And compete with the ones who are starting fresh.

Speaker B: Yes.

Speaker A: So interesting.

Speaker B: It's a complex dance that's going on. We come back to that part on the market.

Speaker A: Mhm.

Speaker B: Small companies, medium sized companies, large companies if they absorb AI do you see? World continues to be like this way where small companies still exist, medium sized companies still exist. Only all the market goes to large companies and traditional industries I'm talking about. Or it's equilibrium.

Speaker A: It's tricky because you have to think about what are we defining as traditional industries. So like take car industry for example. I think you know we're seeing some incredible innovation there but it still takes a lot of capital to produce cars.

Speaker B: Yes.

Speaker A: And service them and create and the ecosystem that goes, you know, you don't buy a car and get rid of it two years later you've got this longevity of the parts and everything. So I think we'll get efficiencies like we've Never seen before. I think you'll get advancements now the trick will be, you know, think uh, about cars five years ago and cars now. How you know, buying a new car for us used to just be you got tired of looking out the dashboard and so you. But now the technology is moving so fast you want to keep up with that. So I think the buying cycle will reduce student innovation. But I don't see, you know, this idea of having you know, automotive startups out of your garage. I don't see that happening because of the capital it takes and the thing. But um, I do think you'll have companies that will operate and be larger at scale on a global basis, but be smaller in their footprint.

Speaker B: Yes.

Speaker A: And I think we'll see that. We're already seeing that. Think about how many young people who are in let's say high school or college are already operating businesses that have seven figure revenue numbers that just because they vibe coded something into a marketplace and they've got a just in time inventory and an idea going. That's amazing.

Speaker B: Yes. Just add this part at least internally. The way we think about is if you're moving atoms, it is not as easy as wipe code. And I can disrupt.

Speaker A: Oh my gosh, totally different. Wait, Sadi and you can disrupt it?

Speaker B: No, it can't disrupt so easily. But if I'm moving bits and bytes it can run faster. But if I'm moving atoms, whether it's a logistics company, yes, I can get the benefit. But it is still there is enough things on the world to make sure it works because that's what the real world everybody is dependent on. Right. Your AC has to work, your construction has to finish your healthcare problem, you go to the hospital, you have to see a doctor, you have to go through the medicine, all of them. I can't do it on ChatGPT and say I'm done. So that world there is a benefit, it moves that opportunity for us at least I see is it is like still very large market.

Speaker A: Oh, couldn't agree more.

Speaker B: For anybody else to go and grab those markets before we think like oh, it's over like, oh, so many startups in Bangalore, so many startups in the Valley. Are uh, there room for me versus there is so much of opportunity.

Speaker A: Yeah, yeah, I agree. I think we're gonna see that early disruption in the bits and bytes. But I definitely, I had an incredible opportunity to spend time with Fred Smith before he passed away at FedEx and he, he always, he taught me that. And Raj now doing an amazing job There. But, um, that was one of the things he reinforced to me was like, moving atoms is hard and it's very different and it takes a whole new. Now you can optimize it and you can reduce cost, which is the hardest part about it. But there's certain things you'll never change. But certain aspects of healthcare, I do think the physical therapy and the emotional connection and the personal connection with will always be there. And I hope we never disrupt that or try to replace that with AI. Um, but there's so many other things that I think are ripe for that optimization, I would say. And so to your original question, do I think there's still opportunity? Tons, yeah, tons. Like we are literally just getting started now. It is going faster than we expected. I think what happened in 15 to 20 years in the cloud will go in faster, but I wouldn't count out human creativity and ingenuity.

Speaker B: So if that market exists, let's switch a gear a little bit about how the startups can take advantage of it. So, um, one aspect is, hey, I quickly write something on LLMs foundation models, horizontal and I can make that work better. And I have market today. But differentiation, durability of these ideas versus, say slightly vertical in the sense, uh, how you think about when we meet so many customers, so many startups, horizontals are moving faster, but vertical takes a little bit time. But there is durability. More aspects of that. What have you seen?

Speaker A: Well, you nailed it. So it's the durability, right? You can succeed quickly at high risk, meaning as a model evolves, it puts your business at risk. Uh, or you can take your time and define your solution that I see. The best startups I see have a very thoughtful, what I'd call a complete thought on how they're going to change something in the enterprise. But, uh, how their solution is not limited by the model, but it gets better as the model gets better. And it is an adjacency to the workflow that's specific to the enterprise. And so each enterprise, it's easy to say, I'm going to solve hr. Well, every business, every company has their own version of hr.

Speaker B: Correct.

Speaker A: HR is such a manifestation of the culture and so you have to understand and tap into that difference. So if you can come up with a solution. And again, I get the opportunity to see a lot of B2B solutions as well as B2C, but mostly B2B. Um, but the ones that I'm seeing do these amazing things and take their time, but I look at them and think they're really onto something big. Like a global scale business is a unique idea to help drive an outcome for a business and rewire the business for an AI world but really understand and tap into that specific workflow of the enterprise that's different for all of them. And I think they're getting a better semantic understanding of it. There's a lot of different tools and frameworks. We've got one with work IQ coming where you can really tap into that unique workflow of the individuals coupled with your solution and then enhanced by the LLMs. And so I think that's the kind of core framework I would say I'm seeing founders uh, take their time to your point, be thoughtful about it. Now it's hard. You know pilots are easy in sandbox but implementations in enterprise is very hard.

Speaker B: So if I want to add a little bit more to it.

Speaker A: Yeah.

Speaker B: See one is take the same hr. There is nuances inside the company, inside the domain. Mhm. Right. But there is maybe 80% horizontal every hr. People know how it is. I can build an 80% hr horizontally, but if I make it a little bit more adaptive to the domain, to the specific company, some sort of a personalization that adapts around it and gets and enhances the wants of the HR they wanted in that org that was not possible before then it becomes very sticky and durable because you are part of the system.

Speaker A: I agree. I think that's one of the challenges today if you look at a lot of solutions is there are solutions, but from the company standpoint they're not our solution.

Speaker B: Yes.

Speaker A: Right. And so you just said it perfectly and very eloquently of how do I provide a solution that can adapt and learn and kind of provide that company something that feels more like them. Yes. And I think that's the, that's what AI is going to allow us to do and really um, connect those solutions to the culture, the ethos of the companies in a way that we haven't seen before.

Speaker B: Yeah. Like I'll give an example. We had kind of deployed a uh, performance review for us.

Speaker A: Sure.

Speaker B: You know historically when we see there used to be 14 questions, generic questions applicable to any company, we also see it say okay, it's uh, getting some functionally complete. But recently we looked at when startups performance review. I said I'm a VC firm and uh, it came up with 14 questions based on the roles. Like if you're an EA, the questions it asked is how many startup founders meetings have you scheduled? See that so specific. I just like a glimpse of what is possible. Right now I just have to say I'm a VC firm. Maybe they are hard coded for accel to see whether I could invest or not. But I'm just saying it adapted so well to our world and that is the possibility. Now we have deployed it, there is no way to take it off because it knows our system very well. This is, um, applicable to every industry, every company.

Speaker A: I agree. It reminds me of a good friend. He said, um, I have a lot of children at home. I have four kids. And um, someone once said to me, I've read every book there is about kids. I just haven't read the book about my kids. And to me it feels just like what you described. Right. Where I can deploy. I've tried every HR solution known in the universe to my company, but I haven't found the one for my company. And I think that's what really the unlock is like. To your point, it's like, how do I now have a performance review system for my company? Right. And then even further, like, we're a vc, but we're Excel. We're Excel in India. Right. We're Excel in Bangalore. You can get very specific and I think it's just so much richer. So, yeah, I'm optimistic about what can be done. And uh, that's what I'm seeing is the companies that really tune into that capability and then of course build on this macro framework of the right models. The interchangeability of the models don't tie themselves to one model because we're going to see this model war play out for years, which is great, but I don't want to have to, uh, when the model changes and one leapfrogs the other, I don't want it to destroy my business. I want to benefit from it.

Speaker B: Absolutely. There are two classes for startups we are seeing. It'll be good to whether you say acknowledge the same or not. When they go to a customer, if customer has not been exposed to. They're aware, not deployed anything like ChatGPT or anything else. They're still quite. They're aware of.

Speaker A: They're really living under a rock.

Speaker B: Uh-huh. They've done a little bit of their personal things. Ah. But. But not for the company.

Speaker A: Got it. They don't think they are. Their employees are playing with it and using it every day. Just the company doesn't.

Speaker B: Company doesn't know company is not deployed. So when these companies approach them, they say, I will use CHAT GPT. I know AI don't need anything that is the first behavior pattern we are seeing, the second behavior pattern is people say okay, I will use chat GPT for myself, for the company, uh, they have attempted to do some things, it gets to I say 80th percentile, 90th percentile, but it cannot absorb like what we are discussing, then the conversion happens faster. So some of our companies are uh, pre qualifying question is what has been the current AI initiatives? You are doing it if the AI initiatives you have tried. But if they tried three initiatives, they failed. Fourth one they say outright reject, saying that I don't want to see one more demo, but it's a tried one but not working yet. So capturing this time and intercepting is the most important thing for some of these young startups because they don't have a lot of money to experiment. They have to get right customers, they have to absorb it, they have to get the revenue stream going and so on. So as you look at these large enterprises, you guys are the kings of enterprise market. Maybe help for the startups that are coming up. The importance of trust, importance of demo versus uh, commercial quality. How do you demonstrate that this works without hallucination? How do you see this, making sure that this has legs?

Speaker A: Yeah, um, the first sale, I wish every founder understood this. The first sale, especially to an enterprise is a trust sale.

Speaker B: Yes.

Speaker A: It has to be about building trust and it has to be about uh, bringing to bear every, every tool you can to do that. So one of the ways I see a lot of founders do it, or you know, especially earlier stage companies is to come into like our marketplace or even Amazon's marketplace. But you know, for us trust and security and safety of our cloud is everything and that's been a foundational pillar of it. So when what we help founders understand is that when you're in our marketplace and come in through that, couple things happen is one, you don't have to worry about the agreements, you don't have to worry about the T's and C's. Your speed to sell is dramatically increased. And so the founders tell me, boy, when I went through Azure Marketplace, I could just transact so much faster. But most importantly it's then okay, how do you spend the time with the customer to make sure that they understand that the trust isn't there just because you're on Azure. But how are you building solutions that they can trust and, and to really build that reference case you need? And so I think that first sale, being a trust sale, is super important. Using the tools of a hyperscaler to actually give you credibility is also another important thing. I think there's another opportunity to use funding programs and things that uh, we're putting into the marketplace to accelerate the sale. But at the same time, as you said, capital is just so valuable and time is so tight for these startups. Find those programs that will help you be successful because then you can start that. Going from 0 to 1 is hard, 1 to 10, super hard. But 10 to 20, 10 to 100 can go super fast. Once you've built that trust fund, once

Speaker B: you cross that valley wall, then momentum can build for you in your favor.

Speaker A: And I think many of them, it's easy to overestimate the impact you can have in an enterprise, but underestimate how hard it is to get those first

Speaker B: 10 companies that are starting out right now. Mhm. They fear that 80th percentile, 90th percentile product, is that sufficient? Or our enterprises are more experimental in nature to be willing to work with me to build something because these opportunities won't last long enough. So as startups, when they start, most of the enterprise want this, they heard about it, but not enough people are making it work for them. If somebody can do that together, they are more benevolent in helping you work with you. As long as the trust is there, the data doesn't leak and you are high intent team trying to solve what we call it as design partners. Yes, getting the first four, five high quality design partners where you can work together and create and learn with them is becoming extremely important in this market because once you get it right for them, they will be your super endorsers, uh, to all other next 10, 15 customers. But if you cannot crack these first five, then you have no play for the next 15.

Speaker A: Yeah, I agree 100%. And I think you also hit on an interesting point of they're more willing to be a design partner now more than ever.

Speaker B: I agree.

Speaker A: One of the industries we're seeing move faster, which blows my mind, is financial services. A highly regulated industry that took forever to get to the cloud is now embracing AI as one of the fastest industries. And when I talk to JPMorgan Chase or Lazard or when I go and sit down with them, they have whole groups that are designed to identify AI startups, bring them in, but they're absolutely willing to learn and develop with you. But to your point, you have to be willing to take the feedback and make the change. The data has to be safe and secure. You have to be built on a trusted cloud. I mean I do believe that because they've worked so hard to build that infrastructure that is trusted. They can't put it at risk. And then you have to be intent on helping them and really solving their problem. But they are way more willing to help design build with you now more than ever.

Speaker B: Yeah, yeah.

Speaker A: Because I think 90% is good. Because I don't think anyone knows what 100% looks like.

Speaker B: Absolutely.

Speaker A: You know, and if it was, it changes week to week.

Speaker B: Absolutely. And your iteration faster. Uh, so one question. Everybody says, assume I get these good partners, I build it, then LLM launches the next iteration. Am I, like, irrelevant where you are seeing? Because we're discussing about a little bit of domain knowledge, uh, expertise. We spoke about personalization, but what else Companies can add secret sauce that makes them, uh, uh, be able to absorb innovation rather than afraid of innovation.

Speaker A: Yeah, it's a great question. I don't know that I have the exact answer. I do think, as we said earlier, building a solution that can get better from LLM enhancements is super important. Um, and I think that takes a good founder and someone being very creative and thoughtful. How you build your solution and part of your architecture is the uniqueness of the organization, especially in business. And then the one thing I found that's universal across business and consumer is you've got to have taste.

Speaker B: Absolutely.

Speaker A: You know, you've got to have a great app, great apps, great technology, great interfaces, whether you're generating them on the fly for users. What. There's something about taste that can't be replaced. And I think we all feel that. And when you know it, when you see it, you know it.

Speaker B: Absolutely.

Speaker A: It's hard to describe, but there are tastemakers. And I think you can't outsource that to an agent. You can't outsource that to a coding agent. There's only certain things that good founders know in terms of how to create good flows for the customer, good flows for the users. And so I would say that is a universal truth to me, where you can get all the architectural things. Of course you don't want to be something that, you know, the latest drop from anthropic puts you out of business, which we see that too. But you want to be something that is enhanced by the LLM, um, enhanced by the data of the enterprise and just has great taste.

Speaker B: Absolutely.

Speaker A: Good apps will always win.

Speaker B: So the taste one is a very important one. We are seeing it now.

Speaker A: What are you seeing?

Speaker B: What are you seeing? Yeah, so this is an excellent, uh, kind of next topic. What I'm seeing is when companies are Constructing these products, um, there is a life cycle. What I mean by that is our old software where UI built for cohort of ICPs could be small company, medium company, large company. The products from Microsoft or everyone we are used to pre AI era was built for a cohort of users. And that cohort of users, what they have to see was decided by a product manager telling the engineers how to build it. The new age ones, what we are seeing is where the user behavior, what are they doing their feedback loop in the AI where they're correcting. What I'm showing is asking the question again and again is a feedback loop. Using that feedback loop they are re engineering the flow of how the product gets rebuilt and feedback loop gets into the product and that loop. It used to be months before quarters and months because I have to hear to a lot of customers go through the tickets, understand where the challenges are bugs and so on and build it. Now as the cost of writing code is coming down, if I can intercept this taste aspects from the customers and my feedback loop is at most a week. Many of our companies are scaling rapidly, are on a daily release cycle. So what is happening is what used to take a year, they've reduced to a month. What is take months, they're reducing into an hours. So I can quickly launch something that people are struggling with, get that feedback loop. If it works for X number of customers, I expand the number of customers I can test this with. So more customers, I have more feedback loop and therefore I can make the work better and also personalize to each person.

Speaker A: No, you nailed it. And so I think the ones, the ones that I see accelerating the fast, the ones I would call my highest potential ones, I'm looking at, we have agents that actually scour the world and look to markets, we look at social signal, we look at multiple databases. The ones that I'm seeing pop, they're built on this.

Speaker B: Yes.

Speaker A: They're not adding it. It is part of their product, it is part of their life cycle, it's

Speaker B: part of their harness. Oh completely.

Speaker A: And not only that, they have built in the ability to triage out and separate out the things that are bug fixes that just need to that could be fixed by an agent, triage quickly. They also know how to get the signal of the things that are part of the taste.

Speaker B: Right.

Speaker A: If enough users are using something a certain way, is it just this unique set of users or is it a fundamental shift to the way the product is used? And I think they do that very quickly and that is a Whole new paradigm in terms of taste making and applications.

Speaker B: The tastemaking part there is another aspect. We are seeing people who are coming from the domain with an AI expert together, they know what is real, who have been deeply thinking for decades and they're seeing the power of AI. Can I combine Is like one approach. The second approach is two young engineers think, like, this is a great market, let me build it, but learn the taste making with the initial design partners and um, build towards.

Speaker A: It's interesting because you've got this interesting scenario, the first one you described. You've got this domain expert with AI expert, and one person doesn't really know what's possible and you've got a person saying, no, no, we can do something different. Then you almost have these other two who are ignorant to what they're not bound by the limitations of prior knowledge. They're actually just thinking, well, why can't we do this? And they're thinking so radically that had a subject matter expert been involved, they would have said, well, you can't do that. We never did it that way. And so you have this interesting thing for both where you're right, they're coming up with big ideas. Right. And I do think one of the biggest limitations or slowest things that will hold us back is just people limiting themselves in their imagination.

Speaker B: Absolutely.

Speaker A: Ignorance is the unlock.

Speaker B: Absolutely.

Speaker A: It's bliss as the old adage goes, where, you know, I look at like my kids sitting down and vibe coding something because like they have no prior. And when you have no prior, you think totally differently. And now when you have a tool that lets you do things that were never, uh, capable before, it's a totally different game.

Speaker B: Yeah. So this no prior because we all have seen how hard it was to write a production quality code that to unlock in our mind is so hard. All of us were grown up as engineers and it has taken like, you know, the process launch, you know, your old world of burning the CD and launching. Sure.

Speaker A: Oh, gosh.

Speaker B: So it has taken uh, how to unlock it versus the new age. But here is one thing I would add here. It's not about the tastemaker, not about the new young kids. I see. It's a first more advantage. Hmm.

Speaker A: M interesting.

Speaker B: Uh, so if you are a first mover in an industry, you get first set of customers. Your learnability index is very high. You really listen to what customers are saying, what is possible, you reduce the feedback loop and you run fast. By the time a domain expert, an AI person comes in, 18 months to say, I know how to do it versus somebody who has done this faster. There is a higher chance of the first mover to win.

Speaker A: That's true.

Speaker B: Because they have the momentum.

Speaker A: That's true.

Speaker B: And that we are seeing industry after industry that's happening. And that's where I see some of the incumbents may miss an opportunity because a new age people are running fast compared to the incumbent saying that, okay, I need to rehash myself.

Speaker A: Yeah. I mean I think of that as speed is the moat.

Speaker B: Absolutely right.

Speaker A: You can launch a product and it doesn't even have to be amazingly great. You just have to be willing to get the feedback and increase your cycle time and rev the product quickly because there's so many products out there. Someone will try one and if they don't like it, they'll move to the next one. But if they try yours and you just constantly see the innovation coming, they'll stick with you. And so I think that speed, uh, we're doing that internally as well where we've started this process called the cohort process where we are now bringing engineering and sales so much closer together and trying to really increase that cycle time to be infinitesimal. And it's really working. And the market will start to see the fruit that this will bear in the coming months. But we feel a difference.

Speaker B: Right?

Speaker A: You just feel what used to be this when I started at the company 24 years ago was like 3 year ship cycle. We're going to send it then we're going to do customers usability studies and maybe it'll get into the next version. Now things are going so much faster that we've brought these two together in a way that I'm very inspired by. And it's been some pretty welcome changes by some new leadership at the company.

Speaker B: Unbelievable. Shifting a gear for specific to our audience. Good product people, good tastemakers, early absorbers, early adopters. India or global. I think most of our startups have nailed that part and we'll do that little bit more application oriented as we discussed. Not so core on the LLM side. Agree. Uh, not so much on the model side. From India, from this part of the world.

Speaker A: This part of the world.

Speaker B: What we all more desirable is the both ambition index in the sense that hey I can build really large company. Of course there are companies in the valley us growing very rapidly, not getting worried about it. So anybody can build because we are seeing from China, from Singapore and London everywhere people are building. So it's nothing belong to valley. Everybody can build from anywhere. Agreed. That's 12 go to market. M there is something about US companies that go to market. Well have m you seen. I've got some ideas but have you seen how this wheel runs fast?

Speaker A: Sure, sure, yeah. I mean the best ones I've seen think go to market from the beginning. Yeah, you know I think it's before

Speaker B: the product is even ready.

Speaker A: Sure. Well it's part of the design.

Speaker B: Right.

Speaker A: Because you're just thinking of again it goes back to this theory of a complete thought. Yes, you can have a great product but without a complete thought of how I'm going to bring it to market. You know I, I kind of, you know in tongue in cheek make this joke of you can take, you know some people are like PhD level which they really are in terms of data science and generative AI, but like junior high or high school level in terms of go to market. And it is one thing where you've got to think about that go to market early, you've got to think about your partnerships early and how are you going to get that product into the hands of the customers and what does it mean to um, everything from your business model to your pricing and how you're going to kind of have that line logic of the product as it lands and expands. Are you going to be um, product led growth? What is your, what is your angle in that regard? But I think um, the good thing is for us, my charter, my purpose in the world, for my team, this great group I have around the world is really around how do we help startups do that? And so how does my team lean in to help your founders from seed all the way up through scale to say as you mature from. We have this construct of 01, 2, 3 and a maturity life cycle that starts at seeing and even friends and family and no product to de round and things like that along that journey. What are the things we can do to help you with go to market? Because to your point enterprise is very unique.

Speaker B: Absolutely.

Speaker A: Consumer is very unique. And understanding enterprise, understanding enterprise by industry, by region, by country, it gets very complex and I think the ones that are successful are asking those questions early. They're thinking about am I a regional provider, am I a national hero, am I thinking globally, how am I going to distribute myself? Things like localization, um, all those things matter and I think the earlier you start thinking about it or asking the questions the better I think you are in the long run.

Speaker B: Um, especially if you are saying you don't have to act on it, there should be a thought perhaps or you get going on there. It cannot be an afterthought.

Speaker A: Absolutely.

Speaker B: It has to be a pre thought before you kind of put everything there and you start experimenting against it. Because one of the things we are seeing the value companies do really well is the experimentation.

Speaker A: Absolutely. Well, it goes back to what you just talked about with the user feedback.

Speaker B: Absolutely.

Speaker A: It's the same thing. If you build that into your framework and your harness and your business if you will, you do the same thing with the go to market. How do I learn? And maybe my PLG motion is wrong, maybe I need to adjust that, maybe I need to move the paywall on certain things. And I mean all those things I think have to be looked at the same as important as your product dev cycle, your product architecture, things like that.

Speaker B: One change we are observing now, which was not there before, which is very beneficial to the startups, is if you look at valley startups or US startups, if you're starting something in accounting, AI first accounting in New York, your first set of customers are all in New York, they buy locally. If you're starting in Valley, lot of other startups and slightly mature companies are willing to try out. So there's a local ecosystem of buying your first set of customers.

Speaker A: Ecosystem or echo chamber?

Speaker B: Uh, one or both.

Speaker A: Exactly.

Speaker B: We were missing in India for a very long time. But if you look at all our modern companies which have all started recently in the last 15 years, they are equally willing to experiment because they also want this modern, well constructed, tasteful products to try it out and because they're willing to try it out as because they're more open. Your first set of feedback is very good from this part and they're equally as demanding as global customers.

Speaker A: Absolutely.

Speaker B: So if I can, if not more, because here it is, there is still price conscious customers. They want value for every dollar they pay you and it has to work and it has to adapt to their needs and all of them. If you battle test here with your design partners and first set of customers, it is a little bit easier to go global versus a while back most of our startups starting out here, day zero, they're trying to sell globally, travel, convince a customer, you have to describe where you have come from. Much harder. Uh, odds were against them. Now odds are in favor of these companies because India as a market is pretty big. In fact I did see that both a lot of AI products from Anthropic, ChatGPT, even Microsoft significant user base in India. Adoption of AI in India is a much faster pace than quite a lot of the world. So you are seeing AI enthusiastic buyers starting to happen in India so that you can build something for first test here and go global.

Speaker A: Yeah, uh, I think that's a smart model. I mean one of the things anytime I meet with a founder and they tell me they have customers is if they rattle off some of the largest multinationals. Honestly I'm skeptical and it makes me want to say great, who can I call there to ask? But the ones who say, Look, I've got 10 who are smaller that I've tested it with that are. And you can call any of them and they will sing my praises. And I've got these as I move up the stack. That means a lot more.

Speaker B: Right.

Speaker A: Having legitimate. Um, you're right for the number of. I wish I had a, uh, uh, I guess it would be a rupee for every startup that wants me to walk them into Walmart.

Speaker B: Exactly.

Speaker A: You gotta be ready to play on the main stage when you. And they're my metaphor for other incredibly large companies. You know, think of JPMorgan Chase, Walmart and it's just you've got to be ready to play that game and you've got to be ready to your point of earn the trust.

Speaker B: Absolutely.

Speaker A: Spend the time and understand that your greatest resource you have as a startup is time.

Speaker B: Yes.

Speaker A: And that is your most limited one that is non renewable. And so you've got to select very carefully where you're going to spend that time to build success. And if you go too big too soon, they can be spectacular failures. It's better to kind of build yourself along the way so that when you land there you can really win.

Speaker B: Absolutely. I like your framework, Jason. On the go to market side, there is one thing we are hearing from our companies that are in the valley saying I will be very horizontal. Uh, I may have one or two engineers working with customers, but I want to be everything product, product, product, AI first product. Companies that are starting here, they are saying that's great but I do need few engineers to be with the customers to absorb my product and make it work. How do you see this engagement model of these two different styles?

Speaker A: Yeah, uh, it's a great question. I think that topic today, this concept of the forward deployed engineer, the FDE is just on everybody's mind and you're seeing teams spring up everywhere. And I think if you separate who's doing it, why they're doing it, from just the need for it, I think is where you start to say it needs to be outcome driven. We're at this Stage where over time and in the last say 20 years, so many enterprises have adopted new technology with the promise of change. And I think now riding on the back of this hype cycle around AI, everyone is very skeptical but yet they're very optimistic and they see the opportunity. And so to bring a new solution in, I do think startups need to think about how am I going to ensure the success.

Speaker B: Yes.

Speaker A: And so that's where I would say whether you're, you know, someone told me the other day, they said it's FD is founder, deployed engineer. Right. It's that idea of I'm going to get in and help make that solution successful because I need to drive the outcome.

Speaker B: Yes.

Speaker A: So I think gone are the days where you can just hand something over and assume it's going to work. I think you have to, going back to that earlier statement, you have to build the trust by really engaging with the customer to make sure the outcome is achieved. And I think the challenge is, what we're seeing a lot of companies do is they're willing to make the bet on the outcome versus saying pay me first and then I'll get you the outcome. They're saying when I get you the outcome, pay me.

Speaker B: Yeah.

Speaker A: And that I think is going to shape a lot of the market. It's going to shape a lot of P and ls.

Speaker B: Absolutely.

Speaker A: But I do think that's important, especially in a time when it's still early days. It's unknown if these solutions are really going to fundamentally change a business.

Speaker B: Don't you think that is an advantage for companies starting here?

Speaker A: Absolutely.

Speaker B: Where there are highly trained engineers, they have to get closer to the customers. And the startup starting here from India could benefit by having that thin layer of forward deployed engineers to focus on outcome, feedback loop and being closer to the customers so that the adoption cycle is faster to make things work faster compared to say a five people team or a six people team somewhere else.

Speaker A: Yeah, I think that is one of the structural advantages that India has in their labor market. You have this, you have two things. You have a low cost labor market with an extremely skilled labor market. And so I think at the same cost per basis or a lower cost basis, a Indian team will, or uh, startup will be able to deploy more talent or more individuals that have a higher degree of talent than a uh, startup in the valley that is operating off a much higher cost basis. So yes, I think that is a structural advantage Indian startups have over I would say a few uh, markets around the world.

Speaker B: Absolutely, absolutely. And as AI is being used here, the diffusion of AI tech in India is super fast. So therefore there is like reinforcement happening

Speaker A: and there is a uniqueness about the culture of learning in India.

Speaker B: Yeah.

Speaker A: I will say I was in the car driving down the road this week or yesterday and I noticed just on the side of the road on these corrugated metal panels, all alongside there were posters for AI Skilling and they were every five feet and it was call this number, take this class. And I thought you wouldn't see that in the United States.

Speaker B: Absolutely.

Speaker A: You just wouldn't see it like literally plastered along the side of the road even in San Francisco. Everyone is hungry to learn something here and I think that will benefit where the company will upskill itself with this necessary and high demand skill set that will benefit these startups in a significant way.

Speaker B: Absolutely.

Speaker A: Yeah. I think that is a unique thing about India.

Speaker B: Jason, basically we have spoken for the last hour or so on um, market is intact. It's very early and if you're thinking of starting a time could be better than now. M. Right. So that, that what we believe right now both of us and we are so early in this new cycle. So many new companies are able to be formed and like many of them will win. So that's great. Second one we spoke about, uh, a taste making product. Product has to have something high quality good works adapts into the each company into the market so that people can absorb it and fall in love with that get some good design partners and how do you can make sure that product works really well. Yeah.

Speaker A: And knows them right.

Speaker B: Knows the customer and absorbs them so that it is and AI makes it easier and you can reimagine services that you were not able to do before.

Speaker A: Right.

Speaker B: Uh, the third one is about go to market where I think your idea of be thoughtful about before even you launch anything is how I want to play the go to market equally important as important as building the product. Not as an afterthought saying that I'll build the product and then let me think about how do I scale it. So if you have some thoughts about it, you can bake the go to market ideas into your business model, into your pricing into your product and enabling it. Either it's PLG or bottom up or top down or using hyperscalers as a way to enter. Having a thought is more important than like because there's very limited time of people because we spoke about first more advantage if you move and you become a big. It doesn't matter how others come, they have to catch up to you and it takes longer and this is a phenomenal advantage. And lastly, these ecosystem pay forward things you are seeing across the globe. We are very excited about Bangalore, we are excited about India. Number of AI startups are coming. Thanks for taking time to spend today with our audience of seed to scale and with Accel. This is great pleasure. Jason, thank you.

Speaker A: Sekhar, thank you so much. Always a pleasure to see you and just learn from you and so, uh, we'll do it again.

Speaker B: Wonderful, thank you.