Sequoia

Mad Street Den: Enterprise AI for the World, Built Out of Chennai

The dawn of AI is upon us and Enterprise AI is finally getting its time in the sun. With more people experimenting with the likes of Chat GPT, Dall-E, and Midjourney, companies are also moving swiftly to implement AI tools to help them drive productivity and efficiency. One person who’s been building AI tools for over a decade is Ashwini Asokan. In 2013, she left her job as a designer at Intel and teamed up with her husband, Anand Chandrasekaran, a neuroscientist, to launch Mad Street Den, a computer vision and artificial intelligence company. On this episode of Moonshot, we chat with Ashwini about the current state of Enterprise AI, how Mad Street Den is setting itself apart from its competitors, and what it takes to build a global artificial intelligence company out of India.

Watch

Listen

Show Notes

  • Introduction [00:30]
  • Consumer AI vs Enterprise AI [03:26]
  • Different industries have the same data problems [09:55]
  • AI automation: Freeing you up to do “the job you should have been doing in the first place.” [13:58]
  • Entering the era of Enterprise AI [22:44]
  • Do you need to raise a ton of capital to get into Artificial Intelligence? [26:59]
  • Raising the next generation to be AI native [31:03]

Transcript

Introduction

Dewi Fabbri: The Artificial Intelligence space is evolving week-to-week as more people and companies adopt AI across different aspects of their life and work. One person who’s been at it for over a decade is Ashwini Asokan, co-founder and CEO of Mad Street Den.

In 2016, Ashwini and her husband, Anand Chandrasekaran, a neuroscientist, left their life in Silicon Valley to return to India and launch the company. Their first offering was Vue.ai – a product focused on the retail sector. Since then, they’ve built products for many other verticals including finance, healthcare, and logistics.

On this episode of Moonshot, Peak XV’s Anandamoy Roychowdhary, or Roy as we call him, chats with Ashwini about the evolution of Mad Street Den into a global AI leader, the state of adoption across enterprise AI, and how she’s raising her kids to coexist with Artificial Intelligence.

Anandamoy Roychowdhary: So Ashwini, I recently met a journalist and she said, “Hey, tell me about what exciting AI companies you guys are working with”. And I was like, “Well, let me tell you about MSD (Mad Street Den). And she was like, “Well, MSD can’t possibly be doing AI. They’ve been at it for too long”. And I was like, “Well, there may be something to that. I did meet Ashwini almost 10 years ago”. Well, how do you respond to that, Ashwini?

Ashwini Asokan: We’d just come back from the US. We didn’t know anyone, that’s literally true, other than family. We’d never worked in India. And we were trying to set up a base here and kind of go back and forth to the Valley, and we just had this kind of glimmer of an idea. We wanted to build a very specific type of AI platform. And we saw kind of, you know, enterprises building on top of this AI platform. We wanted to build this company; and at that point, it was literally the beginning of that whole deep learning wave in the Valley when we decided we wanted to do this, and our philosophy of how we wanted to build a platform in so many ways was very contrarian to what we were seeing happening in the Valley at that point. And so that was one… 

We didn’t want to be in a space that was very different from where our heads were at with reference to how we wanted to build AI. And secondly, we actually wanted to come back home. We wanted to… Like, Chennai is a deep tech university town of the country here, right? You’ve got so many universities and we wanted to dig into that kind of talent and build this enterprise out very closely at home, with the talent we believed that we wanted to be a part of this story. And at some level, you know what they say, right? “The revolution will be televised”, and we wanted it to be televised from Chennai.  And I remember when we came and told these stories, we met one guy who was like, “Yes!” – One. That one guy was you, Roy. And I think the rest is history.

Consumer AI vs Enterprise AI

Anandamoy Roychowdhary: No, I remember that actually… That meeting is pretty fresh in my memory. We met in the context of the hackathon. You know, for me, obviously a lot of the excitement was around the fact that building a platform is an extremely difficult endeavor. And so, the ability to sort of think about that, to have the backgrounds not just in AI, but also in design that allow you to sort of think about how to put something like that together was not something we easily intersected with, you know, in the context of companies you’re meeting. And so, that to a very large degree drove a lot of the excitement at my end.

Ashwini Asokan: Yeah, I want to start off by saying that what we’re seeing in the market today and all the hype that you’re seeing today is what I call the consumerization of AI, alright? You have, big tech out there, your Googles and your Apples and your Facebooks of the world, that are building fantastic consumer-centric AI platforms that allow people to do a lot of things that are ‘good enough’. You can write ‘good enough’ emails, ‘good enough’ marketing material, ‘good enough’ a lot of things. This is not who MSD has ever been. And this is not what we aim to be. This is what gets the cycles of news out there in the market.

What MSD does is enterprise AI. Enterprise AI is the exact opposite of ‘good enough”, right? And there has never been a more exciting time for enterprise AI and, you know, it’s taken its while… Cloud adoption, data adoption, AI adoption, takes time. But enterprise AI is nothing like what you see out there in the market today. It’s the wild, wild West. And, we believe we’re the absolute forerunners there as part of that whole enterprise AI race.

Anandamoy Roychowdhary: Okay, so that’s interesting. And I think the enterprise AI angle is again, sort of an interesting sort of way of describing what we do at MSD. But again, a question for you there… Again, the perception here is that, you know, most people when they talk about enterprise AI, they talk about sticking an enterprise grade LLM (large language model), which is basically their own sort of model on a sandbox and saying, “Hey, look at us, we have enterprise AI.”

Ashwini Asokan: Yeah.

Anandamoy Roychowdhary: Is that what you mean or is there something else going on here?

Ashwini Asokan: Enterprise [AI] runs on three things: efficiency, efficacy, accuracy. Three things that enterprise runs on, right? And one of the most fundamental tenets of making AI work in enterprise, is that your platform has to be extensible across workflows. You can’t just come in and wave a magic wand and be like, “I’m doing the single application. Here’s a model, go do whatever you want to do”, right? Let’s take the case of someone like this $20 billion healthcare staffing marketplace out of the US. Healthcare is absolutely crumbling in the US. Everybody knows that. Especially post-pandemic, you know, surgeons, nurses, basically are being flown across states to provide care on demand, right?

So now you have systems that basically have to come alive to enable this kind of transaction and this flying and this whole experience of healthcare. We’re working with this company, literally one of the largest in its category today, to do not just… So people are uploading… These doctors, these surgeons, these nurses are uploading hundreds of paperwork into this system every day, right? And you can’t go… And they’ve tried – this company had already tried out all kinds of RPA (robotic process automation) tools, all kinds of language models to basically automate the… You know, this could be social security cards, this could be vaccine records… This could be any kind of document that basically comes through their system. What MSD does here with this platform is automatically recognize what’s going on inside all of these documents. It matches, it reconciles across all the different IDs, makes sure there’s no fraud, all of this is reconciled, everything is matching and validated and extracted out of all of this content. And it basically… The story doesn’t stop there.

If you actually think about it, we now know more about the user or the consumer that’s coming on this platform, better than anyone else. So it makes most sense for us to be the people that are personalizing the job recruitment, right? Which means, instead of just shoving, “Here are 150 jobs for you that are open, that are relevant to you, you’re a neurosurgeon.”; now it’s a question of, “Is this state relevant to you? Is this the kind of hospital that you work at? Is this the kind of timing you work? Are these weekends working for you?” There’s a whole profile and aspect to this person now that you can personalize, making sure that they’re getting the job that they want the most, right? And it’s a pretty stressful thing to imagine someone just flying across the globe providing healthcare services; and to be able to create that fantastic customer experience to make sure you’re giving them exactly what they want, right, at any given point of time is ridiculously important, right? And so all of a sudden we’d gone from being perceived as this, “Oh, you can do document processing”, to “Oh, you can match and validate all of these”, to “Oh, you’re a personalization provider”, to “Oh, now we’re moving into optimization and a whole lot of other use cases”. The point here is that enterprise AI is not a one-shot wonder. It is not a magic wand. You just don’t go and drop a bunch of models and say, “See you!”.

If you look at the McKinsey reports, the Bain reports, from the last, say, five years, two stats stand out. Only 25% of models ever created in the last five years have been deployed in production. 25%, right? That is an alarming number if you actually think about it. And even that 25% that has basically been in production, you’re having a large army of data scientists and ML (machine learning) folks who just keep constantly feeding it data because it’s breaking in the wild – seeing a lot of new data and constantly breaking. And on the other hand, there’s this other massive stat that keeps coming up no matter where you go. $50 billion flushed down the toilet in the last five years in AI investments, because they haven’t necessarily produced the kind of efficiency and efficacy and trust and, you know, numbers that these companies [need]. And so for us, the story really comes back down to: Are you able to save costs? Are you able to save time? Are you able to grow revenue and provide fantastic customer experiences? And that’s the place that we play in as a company.

Different industries have the same data problems

Anandamoy Roychowdhary: Got it. Ashwini, that sounds like something that might be very hard for a single LLM to do. How widespread is this? Like, is this like a single company that’s doing this? Do we have more such examples? What’s the general state of adoption for enterprise AI?

Ashwini Asokan: Yeah, I think… So, our journey started off with retail. You know that. I mean, that was one of the things that back in 2016, 2017, when we decided to take the platform to market and we said, “Oh, we have this whole platform. You can build any application you want on top of it. It is a generalizable AI engine that can detect all kinds of data.” And people went, “What?” Right? Obviously that story is very different today because you’ve got something like OpenAI that people are like, “Oh, I see what this is. I understand what this is.” So I think [it’s] fair to say [that] some of the story of the platform was seven years too early, right? But at the same time, we were happy enough to say, “Okay, tone it down. Let’s take one product – everything around our image recognition piece – go to the retail industry and be like, ‘Guess what? I’m going to use computer vision to deliver the best set of consumer experiences that you’ve ever seen before’”, right?
And if you’re someone that is going to click on a pink polka-dotted address, and I click on a pink polka-dotted dress. Your second click is a polka-dotted dress. Your third click is a bandana. The algorithms are looking and saying, “Roy is looking for retro 70s clothing”, right? But Ashwini is looking for pink dresses. Completely different. And retailers just went up, e-commerce guys just went up and went, “Wow. You mean you’re talking about an intent-driven AI engine that understands real time what each person is doing; and in the process, you can clean my data, manage my inventory, and you can do all of this other stuff? Fantastic!” So the story really started off for us with retail. Fast forward four years, five years, we basically said, “You know what? The premise is the same. It doesn’t matter which industry you go to”. Whether you go to retail, whether you go to finance, whether you go to insurance, whether you go to healthcare, every large enterprise has the same four problems. And this is a pretty… How do I say this – this can be a little alarming to say, “What do you mean every large enterprise can be broken?” It is true, right?

Everybody has data problems. They need to clean their data, have the best quality data, so that any AI you can do on top of it is actually meaningful, right? That’s one of our core offerings. Second, once the data is clean, I can now actually start providing fantastic customer experiences. Exactly the example that I just told you, whether it’s in the case of the healthcare staffing company, or this $60 billion e-commerce marketplace that we power for customer experiences and inventory cleanup and planning. Same story. It doesn’t matter. Clean data, create fantastic customer experiences, automate the processes along the way, and then be able to optimize in an efficient manner. Show me one enterprise that doesn’t have these four problems, right? And so for us, whether it’s retail, insurance – we work with multi-billion dollar insurance companies, multi-billion dollar finance companies – exact same story.

A person comes in, loan application, uploads an insane amount of documents, right, to apply for the loan. Now the person on the other side has to actually sit and see each and every document, right? And review it… All of that is now essentially automated. You’ve got reconciliation, matching, fraud detection, pulling in all of those documents. And then guess what? The consumer piece of the puzzle kind of platform kicks in to say, “What should we actually give this person? What are the different offers we have? What can we as an organization give this person for the application that has come in?”, right? What makes the most sense? What is the risk? How can you optimize? And then the optimization story really becomes, “Should we settle? We have like three years of collections pending with this person. Are they really gonna pay? Look up their history and their profile. Let’s figure out how much we can actually collect”. All of a sudden, it doesn’t matter whether you’re talking about pharma, healthcare, insurance, finance, retail, it’s the same four problems no matter where you go. And that is essentially where we are as an organization today.

AI automation: Freeing you up to do “the job you should have been doing in the first place.”

Anandamoy Roychowdhary: And Ashwini, I think firstly, you know, my love for polka-dot dresses is not a secret… Thank you for sharing that. But secondly, to the point you were making about all these enterprises having the same set of problems… Like, a fairly standard and somewhat tired clichés that, “Oh, you know, you deploy AI, and people are going to lose their jobs”. And what you are describing is essentially that, right? Because, you know, what you’re saying is [that] you can do a lot of the sort of error handling and special casing that you would traditionally have said is the domain of human intelligence. What’s the… How do you think about that? And how do we sort of, you know… What do we say to people who believe this?

Ashwini Asokan: I mean, I’ve been on this whole tirade for what, like six, seven years since I spoke about this whole “Brains, bots, and bullshit”. I don’t think my story or stance on this topic has changed. Which is, people’s jobs have always been changing from time immemorial. The job just keeps changing as the market changes, as the technology adoption changes. And one of the things that I’ve been speaking about, especially recently is, if you think about SaaS, you have people behind [a] UI (user interface) entering data, and storing data, and doing a bunch of workflows. We went from there to replacing it with model-building software. And instead of anybody, we said, “It’s now data scientists and ML folks”. It’s broadly the same thing. Instead of sitting behind SaaS software, you now have people sitting behind model-building software, right? 

And now where we’re basically headed is saying, “Oh wait, if all of that can be automated, then what are we people gonna do?” I think the question… And I’m seeing this with stylists, merchandisers in the e-commerce space. I’m seeing this with teachers in the EdTech space. I’m seeing this with people who are processing these loans and people who are recruiters. I’m seeing this across all of these different jobs across these different organizations. Companies are not downsizing teams, at least in the spaces that we are working in and the enterprises we’re working with. And some of the CIOs and CDTOs, right, the Chief Digital Transformation Officers, so much of their work and their vision is about what is the role of these people now, right? And the answer is very simple. Imagine a really… You know, all of these surgeons and all of these nurses who are calling in and who are working to kind of be placed across the country in such stressful times. The job of the recruiter is to actually make sure that the person is taken care of, not to send 150 irrelevant jobs to them, and instead spend time on something that AI is never going to work on, which is making the person on the other side have a fantastic customer experience.

And so all of a sudden what you’re seeing is that software is, instead of doing eight hours of number keying and key typing in there, it’s basically one hour of that; because the other seven hours has been automated, and you’re spending seven hours instead providing a fantastic customer experience. You’re on the phone, you’re talking to the people on the other side, you’re doing the job that you should have been doing in the first place.

Anandamoy Roychowdhary: Got it. No, and so, you know, in some ways, this is sort of a tried and tested theory of the evolution of mankind, right? Which is that every time we’ve built a new tool or had sort of a tooling revolution, what we’ve ended up with is humans essentially ascending the cognitive chain to higher value activities, and tools make us more effective. And this is a…

Ashwini Asokan: That’s right. And also can you imagine the fatigue, right? Of someone sitting behind a laptop, just keying in stuff for like 10 hours. Like, all of a sudden I think there is emphasis on things like that, right? Like fatigue jobs versus jobs where you should have been doing that to begin with. And that should have been your job to begin with.

Anandamoy Roychowdhary: No, for sure, I think you know, at least in the VC world, it’s not a role that is often associated with fatigue, but we’ve also used machine intelligence quite effectively, just to keep an eye on just a larger and larger number of companies, right? So, you know, we never really reduced our need for hiring the folks we hired and all of that stuff. But what it’s done is it has sharpened our spear and given us a chance to sort of, you know, look at companies more carefully and not miss things. Again, because of the fatigue you talk about, which is, you know, you look at a company 10 times in a row, you can miss things because the narrative is set in your head and things of that nature. Ashwini, if you had one stat to share on how real the enterprise AI revolution is, what would that be? Let’s take an MSD lens on this. Is this stuff real or is this a 2024 roadmap item?

Ashwini Asokan: Now, let me… I’ll give you three stats. The multi-million dollar cohort for MSD has grown 6x in the last year…

Anandamoy Roychowdhary: Ashwini, is that one to six?

Ashwini Asokan: I’m not gonna tell you that. I’m definitely not telling you that. And also, have you been reading the updates? Because you should know. So that’s one. Two, our 100K to 300K cohort has grown 123% in the last year. And this is entry [level]. This is not average, right? This is entry ACV (annual contract value). Our partnerships, and we definitely see a partner first kind of approach because our audience is an enterprise audience, right? Large BPOs (business process outsourcing) and IT services are already in these organizations. And they’re now working with us very closely to integrate our platform into their systems so that systems can work much faster, much better, delivering a lot more value. Our partnership side of the story has grown 217% in the last one year, and is currently forecasted to grow 465% in the next nine months. And so you’re talking about… It’s the age of enterprise AI.

Anandamoy Roychowdhary: Alright, we’ll figure that out, but thank you for sharing, Ashwini. I think that’s also sort of an important data point, right? Like jokes aside, the fact is this stuff is serious, right? It’s not, you know… My worry sometimes, especially with folks who play with ChatGPT, which at least is fantastic, but it can feel a little bit like a toy application, but this is tough… These aren’t toy apps, right? Like moving nurses and surgeons across the country in the US is a non-trivial activity. And that stuff will take some time, right? You gotta make sure it works. You gotta make sure that it delivers the value that it delivers. And so, you know, I think great work on that.

Ashwini Asokan: A little bit of an anecdote on this – last night, I had a journalist reach out to me about a story where a very large big tech company is putting out a virtual try-on kind of a studio. And someone reached out to me saying, “Are you guys worried? Like you guys were one of the first to do the generative AI on the whole model photography automation. Are you guys worried?” And I was like, “Big tech doing virtual try-on? Tell me more. Like, no one’s gonna make money out of something like that. Like, it’s probably something that came out of the labs”. And the journalist was like, “Oh, they’re publishing… There’s a whole piece that’s going to come out on how they’re able to generate tops.” – which is basically blouses.
And I was like, “What about the pants?” [They said,] “Oh, they’re not there yet.” And it’s for the end consumer, right? And it’s the exact same point because ChatGPT, Stable Diffusion, DALL-E – they can hallucinate beautiful things. They can create these illusions of spaces that exist in ‘good enough’ space. But when all of a sudden you say, “Generate a human model for me in this body type, this screen type with this piece of clothing”, you get three eyes, four fingers, right? Like, you know, one sleeve, right? Like, it doesn’t work. Try going to, you know, Valentino or Gucci or Armani or anyone and saying, I’m going to use this software. Let’s go. It’s not going to work. You’re talking about 4K resolution, generative AI images that we do for these companies. And then you talk about this. So on one hand, I think what you’re saying about this whole ChatGPT thing, it’s ultimately that democratization. Everybody can use this for ‘good enough’ stuff, [it’s] very different from enterprise AI, completely different.

Entering the era of Enterprise AI

Anandamoy Roychowdhary: You know, my third standard math teacher had told me this once, which is, “Precision is the enemy of creativity”. And, it’s a little bit of that here, right? Which is just that there are two sort of different types of use cases that require you to think about these models very differently. Ashwini, switching gears a little bit. Now as people remind us, you’ve been at it for a while. We met in 2014 when I was sort of very new to this whole VC thing myself. And I just finished my 10-year anniversary very recently. But quick question on that – you’ve been doing this for a while, it’s not been the easiest of journeys, right? We’ve had to wait a little bit for the platform to show up, for AI to have its moment; and some of the moments we thought were moments, weren’t really moments, flattered or deceived, all of that. Does this moment feel real to you?

Ashwini Asokan: A resounding “yes”, and I say that in a much more calmer, pulled back manner than an exciting “yes”, because I’m at peace with the fact that it’s here. It’s here, it’s now. And I see this because of three things, right? Three things really that I’m spotting right now. Yes, there’s the generative AI madness… Like, I’m just gonna set that aside. But one good thing that’s come out of that, is that today when I go and tell people, “Oh, you know what? The fundamental IP behind this platform is vector search engines, which is now a ‘category’.” And all the keywords from seven years ago are now actual categories, right? You’ve got Gartner producing… Giving them names, they’re things, and so the market has come along, right? And today when you see these stories, people actually understand on the other side, which is a very different place to be, so that’s one, right? And I think a lot of goodness has come from, like the OpenAIs of the world putting stuff out that help people understand, “Okay, this is there”, right? So that’s one. Two, I think this is a very important stat that’s often overlooked, which is, do you know the percentage of cloud adoption in enterprise? I was shocked.

Anandamoy Roychowdhary: It is actually quite small compared to what we think it is.

Ashwini Asokan: That’s right, right? It’s in the early 20s, right. So you sit up and go, “Wait, what? Like it’s 2023 and you’re telling me we’re somewhere around 20%?”

Anandamoy Roychowdhary: Just to interrupt you, but you know, it’s a story that makes me very popular at parties. I keep asking people, “Do you think the mainframe market is growing or shrinking?”, and everyone thinks it’s shrinking.

Ashwini Asokan: It’s growing!

Anandamoy Roychowdhary: It keeps growing.

Ashwini Asokan: That’s right!

Anandamoy Roychowdhary: So nothing in tech truly ever dies, right?

Ashwini Asokan: That’s right. That’s right… And so all of a sudden… And cloud is a prerequisite for AI, right? Like, it doesn’t work otherwise. And it’s taken time, right? So much of the last decade has, in the name of digital transformation, has really been about companies going to the cloud, right? And even then you’re seeing, you’re talking about 20%, right? As opposed to how much more the market has to go. So there is such a thing called, these steps have to fall in place, before you’re able to get to the other side of it. And more importantly for us, for us today, the number one thing that we’re hearing in the market and enterprise, “What do you mean? You’re telling me I can go live in 30 days?” And we’re like, “Yeah”. And [they say], “You mean I can build the next set of use cases in 60 days?”, “Yeah.” [They ask], “You mean I can build any of these use cases across?”, “Yeah.” Because all along, you’ve needed a large BPO or an IT services company, a McKinsey, a Bain, an Accenture, or another four consulting companies, and to come in, sit in a room for two years and tell you, “Here’s Snowflake, here’s Databricks, here’s this, here’s this”, put together a story for you. And then a two-year timeframe to go implement all of that and then start buying AI. And if you buy AI or build AI before this, what you get is no models in production, essentially, right?. And so this change, I think, is, and people just sitting up and going, “I can go live in 30 days? Really?” And saying, “Okay, let’s go, let’s go, let’s go.” Right? That friction is just in a very different place today than it was seven years ago.

Do you need to raise a ton of capital to get into Artificial Intelligence?

Anandamoy Roychowdhary: That’s true. And I think we see that, which is… You know, the modern data stack got a little bit of flack for this, but getting set up, getting most folks to the point where they have it up and running has helped us get to this point. There’s been a little bit of Twitter twizzy around like Indian startups and India’s right to win and, “Oh, we don’t have enough money”. Which, you know, now that we’re almost the third-largest economy, always makes me wonder a little bit. But what’s your take on that? Is that really an issue? Like, why do people get so hung up on like, “Hey, [you] can’t do this without a ton of capital.” I know MSD has been extremely frugal through its journey. How are you thinking about this stuff? And what’s your advice to people who are like, “Hey, you’ve got to raise a hundred million dollars if you want to even start an LLM project”, what do you say to folks like that?

Ashwini Asokan: Yeah, I think there’s some truth to it, but there’s also the other side of it because we’ve lived that other side of the story. Today in the market, without raising a billion dollars to feed our server costs, we routinely find ourselves competing with Palantir, with UiPath, with C3.ai, with Adobe. This is broadly who we encounter in the market for us when we are out there competing, right? One could argue that makes zero sense. Like, how do you guys even get to the table along with those names, when you haven’t necessarily had that kind of capital? So I say that it’s not the easiest path to get there without a ton of capital. If you’re trying to do some foundational work… And it is true, Sam Altman’s words in one sense is very true that if you are looking… What has he done? They’ve gone and churned up the internet, right?

Anandamoy Roychowdhary: And I think to be fair to Sam, I think he was answering a very specific question, which was like, “Hey, if you want to build what I’ve built with that money…”, like he couldn’t get it done. And he’s not sitting there saying, “Hey, I got this done with $10 million”.

Ashwini Asokan: Yeah, yeah, exactly.

Anandamoy Roychowdhary: He raised a billion dollars, they did what they had to do. And, you know, and at least the limitation in AI is real, right? Like, “Hey, if you don’t have GPUs, what are you going to do?” Right?

Ashwini Asokan: Yes.

Anandamoy Roychowdhary: Like before GPUs…

Ashwini Asokan: That’s exactly right.

Anandamoy Roychowdhary: So it’s not a theoretical constraint. It’s a very real one today.

Ashwini Asokan: Yes… And the other end of that spectrum. But if you don’t want that to be daunting… One of the reasons why MSD genuinely succeeded with very large logos, right? Like, I mean, this was another common wisdom five years ago. Like you and I remember, right? Every time we were in the market, they were like, “You guys are spending your time between the US and India and, you know, startups never make it with large enterprises. You should not be going behind these logos, you should start small.”

And for us, right from Day Zero, it’s been the Fortune 2000, Fortune 4000 logos. That’s really been our north-star all along. We haven’t really changed our goal. You know, we’ve kind of worked with SMBs to prove some of our tech, but that’s always been our north-star, right? So, it is a harder climb, but there’s a piece of success… There’s a pot of gold waiting at the end over there, right? It might be slower than just selling shovels in the gold rush. It might be a little bit of a slower… 

But the value is real, the stickiness is real. You actually add meaning to that organization, right? You’re not easily replaceable. You’re not yet another vendor. You’re not…These are names that have stayed with us for a very long time,  through all of our ups and downs. And I think there is something to be said about the fact that [go] back to basics. Don’t just stay at that layer, because if you’re gonna stay on that application layer, you’re going to be disrupted and you’re going to be sitting around and wondering what’s going on. But if you’re building something real and the deep tech talent in India is mind blowing, right? What we need is focus, and we need a lot of clarity in terms of where are the opportunities in the stack that we can really go after, as opposed to trying to say, “I’m going to go boil the ocean and, you know, churn the internet”, right? I think there are very focused opportunities that are available for us.

Raising the next generation to be AI native

Anandamoy Roychowdhary: I mean, the reality of this is, and that is true, no matter where in the world you are. I mean, I do see, you know, [I] sometimes run into folks in India who want to build something for India. And, you know, and I think it’s a fair question, is, “Why does this need to exist for India?” Like, you know, there has to be an angle, there has to be a reason to exist, a right to win. And that’s true in the US as well. Like you can’t really go around laying claim to large pieces of real estate without having a very solid sort of right to win on that one. So yeah, okay… Ashwini again, switching gears a little bit; you know, everyone’s very worried about also the long-term impact of AI, because apparently, again, you know, fear in this and that… You have children, what are you telling them to study? How are you telling them to get ahead of this AI curve? That seems to be a question everyone’s asking…

Ashwini Asokan: Yeah, I’m also writing about it a lot. I’m writing about it as a mom, as a parent. This is always on top of my mind. By the way, my 13-year-old this morning was sitting over breakfast at seven [o’clock] and quickly sketching something. And we were like, “What’s going on? Why are you sketching?” And she said, “Oh, we have an art competition today in school”. And we were like, “Oh, what’s the topic?” She was like, “AI versus humans”. And I was like, “What?” And she was like, “Mommy, I saw a couple of pieces of drawing on the books that you have on your table”. And I was like, “Do you really wanna say the story of evil AI?” And she was like, “Yeah”. And I was like, “What?!” Like that hurt, right? I was like, “That’s not what you should be doing”.

But my point here is things are gonna change, whether we like it or not, right? We’ve all seen that meme of the dog sitting on the chair with the rest of the room on fire. It is true. It is absolutely true. This is not fear mongering. People have to figure out what scaling up looks like. And people have done it every 10 years, every 20 years. This is not new, right? All of us are running to claim the space to say, “Oh, you know, we got to protect our kids”. We do. We really do have to protect our kids. I routinely send my kids out into the playgrounds and into the fresh air and into nature because I’m like, “No screens”. You’ve got to be really careful about when you let the screen into your life, right? And when you don’t. And things like Apple Vision Pro scare the daylights out of me as a parent. But at the same time, I think it’s really important for all of us to understand how do we use all of this new technology to be relevant, right?

Our jobs are changing, right? What we’re doing is changing and you can choose to say, “I don’t want to be a part of that”. That’s fine, but you will become irrelevant. So that is something to me, the way I see it, and you’ve seen Roy, MSD talk about AI in the context of AI nativity, or “Are you AI native?” We keep asking this question because I see AI nativity as a form of citizenship. You’re either going to be a citizen of this world, you’re going to be AI native, or you’re not. And that is the reality. And to me, this is a question of citizenship and it’s a delicate balance between making sure your kids get fresh air and run out and are doing stupid, silly things. And at the same time, you’re preparing them for citizenship in a world and it is equally our responsibility to do that.

Anandamoy Roychowdhary: That’s very well said. Popular culture for a very long time has demonized AI, it’s demonized robotics, it’s demonized sort of tech, right, in general. What is sort of your… Do you have like a go-to cultural artifact that you use to help people understand the positive impact of this? Anything you’d recommend?

Ashwini Asokan: Oh, wow. That is a beautiful question. A cultural artifact… I was just going to say a book… Books. It’s a weird thing to say, but I’m just going to say… Wow, that is an interesting question. I’m not sure I have an answer for that, but I will tell you this. I really hope that we will see more popular culture where humans coexist with AI. We don’t have enough of those narratives. We need more of those narratives, right? You’ve got to go back in time.

You’re going to have to go back to the Star ‘Xs’ of history to understand… Like you see Star Trek and you’re like, “Yeah, okay. You know, there are many ways for different types of beings to exist”. And, you know, even this is not just popular culture in the movies kind of a thing, but there are entire groups of people across the globe who are experimenting with machine augmented bodies, right? It is real. The Borg itself is real. It is a future that is coming. And I would love to see more stories of… In ways in which this could work together, than just this ‘us versus them’, because that’s not gonna end well for us. Like that is not gonna end well for us. So I mean, I would just urge people to read a lot of books because there’s some fantastic writing out there that can show you the power of how the ‘other’… It doesn’t have to be the ‘other’. And the sooner we figure out how to not make it the ‘other’, the sooner we’re gonna succeed.

Anandamoy Roychowdhary: But here’s a question for you. Like lots of folks are very excited. They’re just getting started. The OpenAI stuff has sparked a lot of interest and you know that’s great, we are all sort of very happy for that. What’s your advice for people starting up now? What do you think you would tell them?

Ashwini Asokan: Fast forward 5 to 7 years. This is going to look like what happened with the mobile phone and the iOS App Store. You’re going to have a whole large percentage of the world that’s churning out applications that are going to be like someone building an app for an iOS App Store. That is where the world is headed. And you have to ask yourself, are you going to be that company, or are you going to be the company that’s actually building the infrastructure to support that, right? Because one is gonna get commoditized. And that’s fine, if that’s what you wanna do, fantastic. But you’re gonna have thousands of players in every use case or application that you’re gonna be building, because that’s the way the world is headed.

As opposed to if you’re building a company that is feeding the infrastructure piece of the puzzle that enables that future… The story, that’s one opportunity. And the second opportunity that I’m seeing more and more as I get closer and deeper into enterprises, [is] vertical-centric stacks, right? Now, MSD for example, like FedEx is one of our customers that is building out a logistics-based vertical stack using our platform, right? It’s a fantastic story.

Think about a company like FedEx, right? Like a $50 [billion] to $75 billion company that’s sitting up and going, “Guess what? We’re going to build our own vertical stack, and that’s the game that we’re going to play”. And we’re going to build it on this platform, and that’s what you’re going to get. You’re going to get a lot of vertical players that are going to be able to capture value from the market and from the domain, right? So my advice is pick one of these two. This one’s going to get commoditized sooner or later. And if you are playing the app game, you probably want to think about why you’re going to win, right? But it’s a great time to start. There has never been a great time… A better time to start.

Anandamoy Roychowdhary: That’s awesome and so with that, the Borg rests. Thank you, Ashwini. Thank you for taking the time.

Ashwini Asokan: Thanks for taking the time, Roy, and thanks for going on this journey with me for as long as you have.

Dewi Fabbri: You’ve been listening to Mad Street Den’s Ashwini Asokan, in conversation with Peak XV’s Anandamoy Roychowdhury. Thank you for tuning in. I’m Dewi Fabbri and for more startup stories, visit our website PeakXV.com or follow us on your favorite podcast platform.