After $9B IPO This Founder is Rebuilding Internet for Agents | Sudheesh Nair, Thoughtspot, Nutanix

Sales is like a mercenary profession. If your software code didn't release and slipped by a week, nobody gets fired because it is considered to be an artistic profession. Whereas sales, March 31st comes, the quarter ends and you're a hero or a zero.

A legend in the India-US corridor, core pillar of the team that built Nutanix along with Dheeraj, Manoj, and then with Ajit who rebuilt ThoughtSpot.

We heard the news that VMware is going to compete with us directly. Dheeraj made the decision that we're going to compete with VMware.

Within one and a half year of starting Tiny Fish, you have some of the largest companies in Fortune 500 as customers.

In the new world, everything should be a product problem because product problems have high leverage. If I fix it, I can get that 100,000 people immediately affected. So, you got to wire all of them in the product so that the product is learning from the interactions and getting better. Everybody believes that sales is what they're doing. The job that you are hired to do is the second job. Your first job is sales. Nothing of value is ever created until it is sold.

Thank you, Siddharth. Thanks for having me here.

The web itself is not being transformed and that is the beauty of it. So there is only going to be one web and the innate power of the internet is that it is going to be always one. In the sense that if there is a new paper that is published by IIT in Delhi, it should be immediately readable and researchable and augmentable in MIT. And a kid in Nicaragua or Nigeria should be able to pick up and work on it. That is the reason why the internet is powerful because it removed silos of knowledge. So it should never be forked. Now the problem is that today the internet is mostly built for human commerce, human understanding. And I'll be very honest with you. If you think of bot detection and Cloudflare and all of these things that they talk about, they sometimes go on a soapbox and say we are guarding the internet, we are keeping the bots away. It's all because what they're actually doing is preventing value creation for humans and focusing on dark patterns for merchants. So, let me explain what I mean by that.

Imagine you go to Amazon and try to buy something. You work there. I'm pretty sure you studied this. Amazon will, if you search for anything today, I guarantee you the first 30 to 40 seconds of your time is trying to figure out how to not fall into a trap because you search for a soap and then you're trying to avoid which one is a sponsored product. How do I not click on it? And then you see something that looks organic, but then you read the reviews and like, dang, there's a lot of reviews. This might be AI. And then you are trying to figure out what is an organically good product for me here. That showed up really nicely. Why? Because Amazon has completely optimized this thing for ads and value creation and whichever merchant is providing the most margin, right? Not about the user at all. And why is that? It is because we as a species are kind of weak. We have fear, we have greed, we have all these emotions and we have scientifically figured out how to exploit it. See, every time people say, 'I bought this.' Would you like insurance? You're going for travel. You just came to us. I'm pretty sure somebody, the ticket people probably would say, 'Would you like travel insurance? Would you like health insurance?' And you're like, 'The chances of me falling sick is pretty low.' But my god, US is crazy. I bet I buy that. And then if you buy a rental car, you get a rental car. In the rental counter, they'll say, 'Siddharth, are you sure you don't want fuel options? Because what if you forget?' And you're like, 'Oh, dang. You're right. Because if I forget or I don't have time because I didn't plan well, I will be paying a lot more for gas. Let me prepay.' People who bought this also bought that. These are all dark patterns.

Now imagine AI agents or robots interacting with the web. They don't care about the journey. They only care about the destination. And that's a fundamentally different way of operating the web, right? But they have to get to the same web. So now imagine this web as a singular entity. There are two different kind of users coming in: humans, which let's say 4.4 billion users out of 8. The potential for between robots, devices, and agents, it is going to be trillions. An order of magnitude more users are going to come to the internet. They're fundamentally different. This group is emotional. They have greed and fear that the web has figured out how to exploit. These are like that Terminator movie. You know, they watch this internet with extreme efficiency. They see ads, they see JavaScript, they see DOMs that are mutating and it will ignore everything that is noise and go to the signal. If it is trying to buy that soap, it knows exactly which soap to buy and it'll execute that task with extreme efficiency. You can't treat them like humans. If you, for example, if you're renting the car and say a robot is renting the car and if it says 'would you like fuel option', the robot is like why would I ever need that because I'll never forget. I'll know exactly what the mileage is, how much efficiency my car has, exactly how much time, what time the flight is leaving. I will figure out exactly when to do that. It's part of my plan. I will fill the gas and deliver to you. Unemotional, perfectly built, right? This is where the challenge is and what we are doing at Tiny Fish is we think there has to be an abstraction layer built for the same web. But this only has one persona to think of which is how do agents, robots and devices will interact with it. How do they need an operational internet that is outcome driven? They won't watch cat videos because they're bored. They won't pre-buy or extra-buy soap because they feel like they're going to Costco and will buy a large bundle. They are going to be extremely efficient. So it's all starting by the way. There is no idea how this is all going to play out. But one thing is for sure the web will be used by trillions of users like these and they will not want the same abstraction layer. It has to be something new and we are trying to figure it out.

You know, customers don't have agents is true but it is also not true. Let me explain what I mean by that. We are focused on two kinds of users. The enterprise business users. So an insurance company claims management specialist potentially or a supply chain specialist in a manufacturing company. Two years ago an enterprise buyer like that if they want something, you know what happens, they will call Accenture or McKinsey and they will do a consulting session. And then today what has changed is that this buyer either will tell their data science people or a machine learning team or their BI team, 'Hey we have this unique problem we need fresh data and can you help us.' And they don't pick up the phone and call a BI company or a large consultant company, what they usually do now is open up Claude Code or some other tool like that and say let me see if I can do something here. And they will say I want to have an extremely complex workflow automated like I want to know my competitor's pricing at checkout. And I want to know that on all 50 states at different zip codes. So I want to run 300 sessions in parallel on my competitor's website. I want to buy from Kansas this box and from Florida the same box and I want you to pick all color options and then tell me when pink color is not showing up as available in Florida because I have a bunch of pink stuff. If it is out of stock for them, then I want to push and get a promo before the replenish. That Claude Code now will say let me find out how to do that and one of the things that will do is actually spin off a bunch of agents and they are going to say I want to know how to run this in parallel so I need serverless computing, browser infrastructure, and then I'll say I need human-like navigational skill because my competitor will never open up an API and give me the data. It'll say it's going to have infrastructure problems like bot detection and proxy and stuff like that. And what we have done at Tiny Fish is we turned all of that into an API. So even though the user themselves may not know that they're actually creating agents through chat interface, they are actually spinning off hundreds if not tens of thousands of agents and these agents are coming to your website. So yes, they're not directly running agents, but agents are approximately 28 to 43% of the internet traffic right now are agents.

28 to 43. It's a range depending on which data source you're looking for. But it is happening so much, all getting spun up by these models.

So this is the thing, like you are here in San Francisco. I'm sure you took a Waymo, right? I'm pretty sure you took a Waymo a few years ago as well. Two years ago would have been science fiction. And today, first time I'm sure of this. When you got into the Waymo, you probably had your eyes. You're probably scared, right? You can't relax because you're excited and you probably have your phone out and you probably took a video of the steering wheel automatically turning. But after second or third, I'm pretty sure you're used to it. You took your iPad and then, okay, what am I going to let me do some research. That is something to remember. Things are happening faster than we think and this industry, this space, it's moving so fast. I hate to make predictions but I do not think we are talking about multiple years, probably two years. But in the next 12 to 24 months there will be significant movements in autonomous robots. The amount of work that is being done here and one of the first use cases that will go viral is domestic chores. And we already know, these are some fascinating stories, some of the Global 2000 companies we're talking to, they did some research and you know what is the number one thing people want to automate is not cooking. It's laundry. So people still like cooking, some people, no one likes laundry. So I do think from autonomous cars which is nothing but an agent that will be making calls to the web every time it says I need to go find a coffee shop versus order something. These are all going to be agentic transactions to home robots to everybody using and building apps. I mean software, it's not going to be something that we think of today. Today software is built and hosted in cloud, Salesforce, ServiceNow, and then we all go to it. Software will be something like American Idol, built by a few, for a few, for a few times. You will build a software app that you and your wife and your family might use three times and that's it. You'll never use it again because the cost of building software will be so close to zero. Which means that the process of building software also will be done by agents. Now multiply that with a billion ChatGPT users. You can already see we will pretty much reach the kind of scale that I'm talking about within months if not a couple of years.

Web is not ready not because of infrastructure, of course there are infrastructure problems and we will talk about it. I think the primary reason is because commerce is broken. Then because you remember I just talked about the detergent aggression. If you take one of the largest in the US, Procter & Gamble, and they make amazingly beautiful smelling, beautiful color detergent, you know, Tide and others. And they spend a lot of money on that brand and they spend even more money on digital shelf so it shows up in the right place with the right color with the right vanilla fragrance and all of that. And it's all built for humans because we like the color. If it is not in the right eye level when you walk into Walmart, you will not see it. So they fight to make sure that is properly placed. Again, coming from Amazon, you know that game. Now imagine that robot that did the laundry is making that decision. It will say Siddharth and your wife Nancy like organic detergent. They like vanilla fragrance but it should not have any toxins because Kabir is small. And it takes that and breaks that detergent into molecular level and it'll understand exactly the components required and then say okay this is what I need to buy. I zero care about the color or the branding or all of that stuff. I'm going to be unemotional about finding that exact composition across a thousand websites. And I will never get tired. I will never get bored. And it doesn't matter if it's on the 50th page or 100th page. I'm going to click on all of them. And then I'll only buy when I know exactly it'll be over. I won't pre-buy and keep it next to me like stupid humans do. And by the way, when I buy it, it'll be the best deal for my family. Now think about the implications on commerce, branding, web companies where brand itself is the entire business. All of that vanishes. So when I say the web isn't ready, it is not because of technology companies like Tiny Fish, now we can actually solve some of these problems, but the commerce behind that is going to be the thing. And by the way these things will happen so fast and it is going to disrupt a whole series of business models.

Not as much as you think. And I'll tell you there is a reason. So for example, automation, when we do automation the primary reason people automate is because it's more efficient and frees me up and it does faster than I can. If a car is slower than walking, you'll never take a car essentially, right? If you look at when you think of web specifically, you think of like OpenAI Atlas or Perplexity Comet and all of that. It's sort of like that Waymo example the first time because first time watching this mouse move is kind of fun. But when you are trying to book a hotel and it is hovering over the calendar for 3 minutes and thinking and it's not clicking properly, you will just cancel and say I'm going to do it myself. That's the problem. The reason why consumer-based agents are not ready to go is because it still assumes humans are watching and human will correct. For example, book me the next hotel in Palo Alto and it booked Four Seasons. Maybe you don't want Four Seasons, right? So you're afraid of that. That is why. But what it has changed is on industrial size for enterprise. So for example, one of our clients is a very large insurance company. And this insurance company wants to make sure that every, there is something called pre-authorization. You know in the US health insurance is so complicated where if your primary care doctor said Siddharth you got to go see a specialist, the insurance company now has a clock running because before Siddharth gets pissed off I need to make sure that the specialist that your primary care has to see, he or she, are they in the network, are they available, are they the right kind of frame, do we have a new contract, have the accreditation is proper and all of that stuff. This is a lot of paperwork involved in the back. It's all in the internet but it requires a lot of work and it can't have failure and there are hundreds and thousands of these things happening in parallel and usually what happens is it takes few days and let's say somebody wants a scan done and after getting that scan only I will have peace of mind because it could be really bad or not bad at all and the patient is like on handstrings and back end is holding everything back. These are the kind of places where agents are already making a difference because imagine taking 10,000 of this parallel transaction and the entire workflow around pre-authorization is deterministically executed and that is important because it can't have errors. It has to be reliable. It has to be deterministic and it has to be done with the kind of scale and cost that it delivers as opposed to the slow moving. So agentic transactions built for this kind of use cases are working. That is why we have focused only on that. We don't want to book your hotel. But if the tour company, the Expedia or Booking.com wants to compare 10,000 hotels at the same time, we are there for them.

That, but look, consumer is a freaking hard game, Siddharth. I mean, there was that period where I was thinking what should I do and I thought maybe consumer sounds good. In fact a friend of mine and I, we tried to even do something in the world of sports but I quickly hit a reality that when you think of companies like MyFitnessPal and Duolingo and others, the way they do demand and then constantly keep feeding the top of the funnel, it's unbelievable science. Like growth business that they do will put all enterprise to shame. It is a very different type of skill and science. I don't have the skill for that because one of the things with consumer business is they keep dropping, people sign up and drop, sign up and drop, sign up and drop, which means that the bucket has a lot of holes which means you better have a huge demand gen pipeline. What these people do is so data-driven, so unbelievably good and so scaled and every message, every pixel is perfectly built. I have a lot of respect for them. So partly like I said it's easy but otherwise like this is not something I can do. I mean quickly realized that.

Yeah. So look, I think sales has always had this bad rap and it is true, it deserves bad rap because sales is like a mercenary profession. It's one of those jobs where you hire people and fire them every 90 days if they don't produce and there is no place to hide. Like if your software code didn't release on February 1st and slipped by a week, nobody gets fired, you know, because it is considered to be an artistic profession. Whereas sales, March 31st comes, the quarter ends and you're a hero or a zero. There is no place to hide. If you say like the product failed or support wasn't good, nobody cares. So first of all, it starts with different types of mentality. And I say that in the trenches there are no board seats. Like let's say you're in World War I and there is trench building and the bullets are flying. There are no boardrooms or discussion going on. The only world is to survive. Often enterprise sales teams feels like they are in the trenches and they don't have a lot of appeal. I mean they could finish a quarter, come back into a QBR and say hey the product sucked. But that is an excuse post-facto. What makes it noble in my mind is those companies where the sales team feels like they're not alone in the trenches. When the bullets are flying, the competition is trying to kill them, the product team is with them, the engineering is with them, the support, the marketing, the CEO, they're all in the trenches. They are learning from wins, learning from losses, losing opportunities. But at the end of it, the accountability is squarely with the sales organization. And if you can properly do that, what happens is that this sales team will be so directly focused on customer outcomes as opposed to figuring out how to report back up to the management so they won't get fired. So organizations where the sales teams feel like they have transitioned from being mercenaries to missionaries, sales is a noble profession and you can clearly see the difference. Customers will see that, sense that and that's the kind of sales organizations that I like to engage with, I like to build and I guarantee your customers like to interact with.

I mean look, I'll tell you Europe is pretty slow but US absolutely. And but here's the thing though, if you're an AI company there's a good chance that pretty much every, if you're a decent AI company with a good team and good funding, chances are pretty much every large enterprise will do a free POC or maybe even do a pilot, paid pilot. But do not think that that means you landed. I mean we talk about ARR and all that very loosey-goosey right now. You absolutely will get the attention and you might even actually get a few dollars for pilots but unless you are able to really deliver value, not just deliver value but deliver value at very low risk. So the numerator, the value, has to be so high and the denominator has to be so low that you deliver the value without saying you know we're going to be here for 10 years because remember this industry is so crazy no one knows what 2 years from now looks like. So if you do that, enterprises are definitely converting into production and that's one of the reasons Tiny Fish we have decided to focus only on internet today because of course we can actuate any browser which means we can easily go into internet, we can use CRM systems, ITSM systems because they're all browser-based applications and we can do anything, payment processing, working capital, any of that. But the moment you go inside then you have to deal with data quality, security, audit, governance and all that. What we have done is we just focus on internet but within internet imagine focus on freshness of data. If freshness is value, we can quickly deliver. So morning we have the meeting and usually in our first call itself we show something that they should see and then if they like something by afternoon we have finished the pilot or the POC part of it and then we go into a paid pilot and from there we can move. So this ability to move means you have to show high value and then the low risk part is let's say Tiny Fish stops existing tomorrow, they ask us that question, we'll say you will just go back to the way things were and it'll be slower and probably not the highest quality data but you didn't break anything. So if you apply that kind of extremely scientific way of thinking on behalf of customers, you can absolutely crack into enterprise.

We got to take a step back then because sales, if you think of it, I said it's a noble profession, right? If you really boil everything down to why people buy, it's three things. They buy time, they buy money, and they buy promotions. It's that simple. Let me, I don't need to explain but I'll tell you again. You will buy a product if it gives you more time to spend with Kabir. You are not in an organization so you don't care about promotions and all of that but in large organizations that matters. People want to get promoted. You know there used to be a saying, you don't get fired for buying IBM. The simplest thing is you also don't get promoted for buying IBM. And this is the same thing applied to Salesforce and all of these companies now, they're all like safe choices. Sometimes you have to step out and if you do it right you'll get promoted. And that's the second. And the third one is, I mean there is a social angle to AI which I'm not qualified to speak about but there is going to be significant transformation in the workforce and labor. There is going to be a lot of disruption to white collar jobs but as organizations go through that there will be dark days so they have to make sure that cash flow and operational efficiency of the business is increasing.

I wish I could say that it's the quality of the tech and it stands itself, but the reality of our industry is that experience and network really does matter. And I've been fortunate enough to work with, so I have two co-founders. They both exceptional people. One is Shuha Zhang and the other one is Kit Zai. They both came from different life experiences. So between three of us we have non-overlapping network. We also jokingly say between India and China and US now we connect like more than half of the world. But seriously though, this is one of the things I think Siddharth, you know this, when I and others reached out, I was hesitating. But then you reached out to Vijay. And Vijay is a dear friend. And when Vijay says Sudheesh you should do this, I'm like doing it. And that is the example of everything. So what are the lessons? Number one, this might sound philosophical which I try not to do but you know you have to ask why are you starting a company. And sometimes people say oh start for the right reason, don't start for the wrong reason. I actually have a contrarian view there. If you want to start a company, do it whether it is the right reason or wrong reason, I don't care. What like quote unquote what is the right and wrong reason? The right reason could be I want to make the world better, I see a problem I feel like nobody's solving, I have to solve it. These are all the things that we tell others. But the wrong reasons which are truly sometimes the reason is because I think I can raise money. Somebody's giving me money. I know Siddharth, I'm sure they will fund me the pre-seed. So I might as well start something. Or worst case sometimes it's like I went to a Diwali party. All my friends have started. I better start. Or you come home and your spouse is telling you hey you went to IIT and see the other person started all of them, right? Or they're thinking oh god everybody's making money. I don't want to work for somebody. Or you had a bad day. Your boss yelled at you like screw this. I'm going to start my own company. My advice is like we are living in such a beautiful age for us in this bubble. If you want to start a company, no matter what the reason, give it a shot. However, you take some time to think honestly like the why. Just be truly brutally honest with yourself about that reason and say this is why I'm doing it. I want to never work for anybody else. I want to make a ton of money so that I don't ever worry about money. I just want to quickly do something so that when I go to my next college reunion I don't look like a loser. Doesn't matter. And then you don't want to show anyone. Put that and say this is why I'm doing it. But then realize and say for us to be a company this is not enough. So I need the missing pieces. One of them could be like a good company can never be built by one person who is so intensely focused on what they want. There has to be a better equation and no one is overnight going to become enlightened about all of this stuff. Which means that I say I do need somebody who really is like manically focused on solving this problem for the right reasons and I'm going to bring that person and then I need one more person who genuinely thinks like they can easily raise money. Now what happens is you have a team.

It is important to know that you are watching the shows that others have put out. Like I'm sitting here and putting out a show and you're probably on the other end watching it. You will never get to see the backstage of my company which is a mess, right? But I go and see the best every day and then what happens is this contrast that building, I get one backstage pass and a bunch of front row seats and all the front row seats are showing amazing performances and then I'm comparing that with my backstage. If you do these few things what happens is you're built and calibrated in a way that actually makes you feel like I know what I want, I know why I'm doing it and I'm willing to grit through it because I guarantee you no matter how good the story is, like Cursor, one of the biggest success stories, my company used to have around 83% of the company engineers were Cursor, today it is like less than single digit. I'm not saying that Cursor is doomed or any of that stuff but the reality is nobody's safe.

I think if you understand the customer psyche, the customer developer is an individual customer, they are really fickle. Wherever they get the maximum value for the buck, whichever helps them write the most accurate amount of code for the least amount of cost, they'll go for that tool.

And they're also constantly trying to see if any product tries to keep them without portability, they will also not use that. Yeah, this is the thing. So every company will have dark days. The question is, in the old days it used to be like you'll have 'I'm on the top of the world' like holy nothing is working, was probably every 3 months and today it's happening three times a day.