Ishan Mukherjee, CEO of Rox: On AI Agents Replacing Sales Tools | End of Limits Podcast by HUMAIN

We individuals have a choice to make. Do we adopt modern technology so that we can evolve or thrive, or do we want to be an island and not going to leverage that? Today's guest is a tech leader building the future of AI-powered sales. Ishan Mukherjee, co-founder and CEO of Rocks. With leadership roles at companies like Apple Inc. and Amazon and scaling growth at New Relic, his journey reflects innovation, scale, and sharp execution.

When you saw, okay, ChatGPT, the problems with existing CRMs, how did that idea come to fruition?

As a technologist, as an entrepreneur, Jensen gave us this kind of feedback, which is none of us create the waves. The job is to detect the waves early, try to build an amazing surfboard to ride it, and try to hire the best surfers, which is the team.

Amazon, you know, their plans for automation and they're looking to automate over 75%. Do you think that's a realistic timeline for them?

I think humans are in this part becoming orchestrators. Yes, there will be somebody who's kind of quarterbacking the whole operation. There will be a few people who are orchestrating the robot for it to be fully autonomous, I think it's going to take a lot longer than people think.

What do you think makes humans special?

A truth that I've uncovered now, which I didn't even uncover at Apple when we were doing Siri, is that the consumer in relation to prosumer expectations is this invisible asymptote that we will never meet.

Hi everyone, my name is Yasmin Khoyatbar and welcome back to the Humane Podcast, The End of Limits. Today, my guest is Ishan Mukherjee. He's the co-founder and CEO of Rocks. Thank you so much for being here today.

Thank you for having me.

Rocks is a Sequoia-backed AI-powered sales productivity platform. And before Rocks, he was the chief growth officer at New Relic, where he scaled the self-serve business from $0 to $100 million in ARR. And before that, you were at Siri Knowledge Graph at Apple. And before that you were at Amazon Robotics. And before that you were at Kiva, which was acquired by Amazon after graduating from MIT. So incredibly long and illustrious background, multi-time founder. Really excited to have this conversation with you today to learn more about what ROCS does and how we're enabling the future. So maybe to get started, we talk, before we get into ROCS, we talk a little bit about your background. In warehouses and logistics. Tell us about how you got into that space and how AI is transforming it today.

Yeah, absolutely. I think the common thread in my career has been I'm pretty obsessed about understanding kind of just this secular wave of technological progress that humanity has obviously leveraged throughout centuries, maybe millennia. And then how that really incorporates how we lead our life. So even before I joined MIT for grad school, I was always obsessed about being an entrepreneur that was at the tip of the sphere of automation. And the automation was how do we kind of create these companies and technologies where people can adopt new technologies and evolve to do higher order work. So this kind of riding this wave of technological progress and figuring out the products and companies to build that drive automation, have real-world kind of impact, has always been a common thread for me. So when it comes to Kiva and warehouse robotics, honestly, I did not start off with an obsession about warehouse robotics. I had— I moved to the States with this idea of understanding advanced manufacturing with a very coarse vision that I'm going to move back to India and become an industrialist. Like when growing up in India, like we didn't really know much about what entrepreneurship meant. We saw these great industrialists like the Tatas and the Birlas who had these huge manufacturing industrial conglomerates, and that was our kind of vision of an entrepreneur. So when I went to MIT, kind of did advanced manufacturing, the idea was, okay, how can I be a better entrepreneur? And while doing that, I kind of crossed paths with Mick Mounts, who is this HBS-MIT dual threat kind of founder. He started this company called Kiva Systems, and the core idea was, how can we automate the supply chain to make e-commerce a viable business? Just to go a little bit detail, e-commerce obviously now here in this region with Uber and Careem and all these things seems pretty like intuitive and everybody can assume it should work, but it took actually about 10 to 15 years of failures to make that industry work. Mik was there on the ground floor. There was a company called Webvan. This was kind of Web 1.0. This company raised a lot of capital. He was the head of supply chain. The first time they launched, he realized that this business would not work because it cost them more to actually package the order than they could kind of charge for it.

It's like the pets.com.

Exactly. So Webvan was the biggest kind of thing. So he started KeyUp with this idea. Of could we automate a lot of the kind of manual process in the warehouse so that we could actually make e-commerce happen. So my— so when I— my kind of paths crossed with him, my core idea is, okay, I'm going to work there, understand obviously automation, but also figure out how automation is changing the whole economy. And that's how I went in.

Interesting. And when we talk about AI transformation today, for example, this is a bit timely because, yeah, I don't know if you saw the New York Times article last week where there were a couple of papers there's some Amazon that got leaked about, you know, their plans for automation and they're looking to automate within, I think the next 3 years, over 75% of the workforce. And it's one of the largest employers in the US currently. What is your take on that? Do you think that's a realistic timeline for them?

Before we talk about timeline, I think one thing we have to acknowledge is that for folks like us who are in this kind of industry, the topic of the impact of automation and technology on the workforce force, both blue collar or white collar economy, is an extremely heavy topic. It's extremely nuanced. And I've been through the whole cycle. We've kind of built robots in warehouses. You had the mainstream media kind of yelling at you, figuring out why are you taking out these jobs, to thinking about AI assistants like Siri, to now like thinking about how to make kind of businesses more efficient. But the But the underlying thing that I always go back to is the technological progress that humanity kind of drives and rides is unrelenting. And every human, whether you're in Saudi Arabia or in my home country in India or in the US, we individuals have a choice to make. Like, do we adopt modern technology so that we can evolve or thrive, or do we want to be an island and not kind of leverage that? And my general view is how can we build these companies and technologies in a way where we obviously drive kind of business outcomes and productivity, but also create avenues to reskill and provide kind of folks to reskill. I go deep into it because when you're actually kind of thinking about topics about job losses, people really forget like it's an extremely nuanced topic. So going back to kind of Amazon and kind of the rumors, it's hard for me to kind of speak to what the specifics are. I read the articles, I think it was coming out from the automation department. But the reality is that businesses small and large, whether that's a mom-and-pop store, a startup, a kind of enterprise business, or somebody like Amazon, we are in a position where everything is changing and individuals first and then businesses have to reskill a lot of their employee bases so that they can evolve and thrive and be part of the future forward economy. And as a byproduct of that, a lot of functions could become redundant. I think that's where all these things come from. I am an optimist, I'm a technological optimist, like my general view is when there are these kind of concerns of automation, there's always net new avenues to do hard order work. So going back to my original story, I was fresh grad in rural Pennsylvania trying to build robots in warehouses. We had literally the whole world kind of yelling at us. The reality is there's those folks who were in the warehouse that I personally worked with. They all ended up getting college degrees, became part of the Amazon technological company, and now have amazing flourishing careers, right? So I think that's kind of what's going to happen is that jobs will be affected and change, but hopefully there's avenues for them to like just evolve.

So hopefully Amazon won't become a net job destroyer, but will continue to be a job creator in the long term. One statistic that surprised me was that they were assuming that like around 30% of costs would decrease, which I don't know if I'm just being extremely optimistic, but it seemed almost a bit small of a percentage to me. Maybe you understand the unit metrics better. Is that a significant number enough?

So I think when we think about these 30%, 20%, or these are all gross numbers. So if you actually boil it down, there's absolute automation of certain tasks. Let's take the impact of AI and agents on the R&D or the engineering side. There is large swaths of the codebase which are now generated and reviewed by agents. So, so there is essentially tasks I wouldn't say like jobs which are no longer need— you need to hire. Then you have engineers because of that, they're actually evolving to do a lot more kind of higher order work. Right, so the net 30%, I think is a gross number. There's 100% kind of efficiency increase in certain tasks, and then there's certain functions which are growing by like 5 to 7%, right? So when I look at Amazon, Amazon is a nation state at this point. They're so massive. This is maybe like being a student of the game. I think they have 4 key businesses. They have AWS, which is the IT org, they have Amazon Retail, then they have a lot of like the other bets, and then you have media, which is a big part. If you look at AWS, that organization is probably going to get more efficient by having like compressing the org charts, but they still need a lot of engineers. If you look at retail, completely different economics. They have mass supply chains, there's gonna be kind of efficiency to be had. So it's very different like efficiency when they grow it up like 30%, I am on that piece, I feel like it's probably higher than what is going to be reality. And because some functions could be 100% automation, some would probably not be. Like actually every good engineering org is actually hiring more.

Okay. Do you then see humans being employed within warehouses in the next 5, 10 years at all? Or do you know what is their role then?

Yeah, so this idea of like lights out fully automated warehouses always existed in sci-fi, not just warehouse, like any industrial automation. I do think the wave of technological progress will get there. Like I think everybody wants to get to like as automated as possible. But if you look at when this journey started about e-com is about a 20-25-year-old industry. Initially it was all mostly manual. Now it's kind of gone down. I think humans will earn this part to becoming orchestrators. Yes, there will be somebody who's kind of quarterbacking the whole operation. There will be a few people who are orchestrating the robots and the number of humans per facility will go down. For it to be fully autonomous, maybe, but I think it's going to take a lot longer than people think.

Okay, I see. And moving on from that, so tell us a little bit about how you went from there to starting Rocks and then what Rocks is as well.

Yeah, absolutely. I think this would be relevant to the audience here, especially with Humane and all the growth that's going here. As an engineer, I'm a pretty, as they call it, hardcore engineer, but I always wanted to do engineering in the service of building businesses. I was an entrepreneur. So as you do entrepreneurship long enough, at some point you abstract away the certain market and you just get really obsessed about what are the transformative companies to build. So for me from MIT joining Kiva was a learning experience and I kind of learned kind of I guess what greatness looks like. After that I'm an immigrant in the US, so it takes me a lot of— takes immigrants a lot of time to get a green card, especially from India and China. So until I got a green card I worked in smaller and smaller companies to just learn the art of company building. So I was at Kiva, which became Amazon, and there was another company and then later which became Apple. So I had a really good kind of rounding. And then eventually I started, after that I started a company. This is pretty kind of hardcore, this company called Kubernetes is a big technology that we are part of. We started a business and that business was very successful. It got acquired by a public company. So when I ran a public company, it's kind of commercial organization, I stumbled into what is called a CRM space, or without going into the specifics, it's a software that you use to manage all your customers. And we kind of— and I deeply understood the space and I felt like with AI and automation, you could rethink that category in a way where now software could do the work. And that's where kind of Rocks came from.

Interesting. So tell us, what is Rocks?

Yeah, great. In plain English, Rocks is technology where AI or agents help you manage all your customers. So what does that mean? Let's say you are a mom-and-pop store and you are selling into other businesses. Traditionally, you would have to keep all the information about the customers in like pen and paper or some sort of software called CRM. Now with AI and agents, these AI systems can do all that work for you. So what Rocks does is we deploy these AI agents where there's one agent per customer and that agent does customer research, outreach. It actually talks to them. It helps you with all the conversations and meetings. It manages all your contracts. It essentially does everything that a human would do. And then it kind of from an admin perspective and it frees up the business owner or the frontline kind of person to actually deep and engage in deeper conversations with them. Yeah.

Okay. And this, this wouldn't have been possible even a couple of years ago, right?

Absolutely not.

So what do you think was the real enablement here?

Short answer is like, I think everybody here obviously uses Humane Chat, hopefully, and have heard of OpenAI. So when LLMs happened about 3 years back, or large language models, which is sometimes called like GenAI, with that, what was unlocked was If you have all the right information and context, you could get software to do a lot of analytical work. As you hear from McKinsey, for example, so like the way we think about our technology is what if you had a team of McKinsey analysts kind of helping you, working for you so that you had a deeper understanding of your customers, you engage with them thoughtfully and actually kind of manage that business. And that's all was made possible by LLMs. And that, so that was the ChatGPT moment now about 2.5, 3 years back. And then we started to work on the problem about 2.5 years back and then officially started a year and a half back. And with that now we've been able to actually like kind of bring this to market.

Right, and you saw that opportunity I think early on you sort of jumped on it when you saw, okay, ChatGPT, the problems with existing CRMs, how did that idea come to fruition?

So great question. Maybe like one takeaway, I'll go back to this kind of wave of like technological progress. I think it's I would say it's an important point that I've learned the hard way that especially early folks in kind of high school or college or even experienced professionals have to, the deeper they understand that, the more they would be technologically forward. So what I mean by that is as a technologist or as an entrepreneur, Jensen gave us this kind of feedback, which is none of us create the waves. The job is to detect the waves early and try to build an amazing surfboard to ride it and try to hire the best surfers, which is the team, right? So we like, for me personally, I'm always thinking about where is the world headed. I think the more people think about it, like people like you who are sharing knowledge, it kind of makes you more, it just increases your ability to kind of ride them. So let's say New Relic, I was running a business, it was $1 billion of revenue. I was just obsessed about running that business. So I was not thinking about AI or other LAMMs or CRM. So I was thinking about how do I keep my business growing? And so I was using CRM tools. So once, as we went through it and we sold that business to private equity, I was thinking about, okay, what are the spaces to go after? So up until this point, I didn't even think about AI. I thought about What's the business? What do I need? So that was an enduring problem. Like we could start a CRM business even without AI. But then once AI came in, we felt, okay, we could build something that's generational. So the idea was kind of keep working, figuring out market insights and pain points, and then look at these waves which are coming and feel like, okay, if we get to ride that wave, like this could be a truly generational company. Yeah.

Amazing. And for you, you know, I would say that you guys are riding the wave obviously in a lot of ways. Yeah. But in some ways you're not. So to expand on that, I think there's a lot of running jokes now that, you know, people are trying to raise in the Valley who are, we have a B2B SaaS company, but we're pre-revenue, pre-product. And you know, they're raising like hundreds of millions in valuations. And then the other thing is a lot of people, instead of creating AI tools are trying to market AI that just replaces the salesperson altogether, whereas you guys are firm on the, we're creating tools that will augment you and not replace you. How did you think about, you know, sort of these anti-waves?

Exactly. I think both are underpinned on what is the enduring value to be delivered and what is the enduring business to be built. So And then I guess through experience you end up learning, for me, for us, this is our life, right? Like we're not first-time entrepreneurs. I'm married, I have kids, and when we do things, we do things with an extreme kind of long-term mindset. Like what if we could build something that could last for decades, hopefully serve customers globally, like here with Humane. So this idea of optimizing for the long-term is very important, but it also really changes the types of bets and investments one can make. If I was a first-time founder, unproven, with a lot of experience and looked at these ways, I would optimize for the short term. Let's try to come up with something where I can raise capital and can learn on the fly. Again, every founder's journey is different. So for me it was, okay, what is a long-term enduring business to be built? And then that led us down to two things. One is what are the types of companies we want to support? So we at the very start believed we want to be, we want to earn the right to help businesses that will survive this AI transformation. So we really focus on old school businesses that have been around for generations. So our customers are people like Siemens and Philips and NVIDIA and Intel and hope as Adronco and Humane, right? So that was always very important. And then within that, it would be hard for me to come to Humane or Saudi Aramco or Intel and say, hey, everybody's going to get automated. That's just not the reality. And so the idea, going back to this thing, like the best will get better. So how can we be a good partner in their AI transformation journey where we build technology and services where the best get better and we actually reskill and retrain everybody else? So as going back to the original point, like for me, it's a very heavy responsibility is We're going to go into these enduring businesses. We want to be an essential partner. As a partner, we cannot be transactional. So we will provide the technology to make your best employees more productive. But at the same time, we literally do new employee training, kind of do all that reskilling and retraining, and that's kind of part of the gift kit.

And the other aspect of the business model that you're changing, I think, is going from commitment-based pricing to usage-based pricing. Correct. So tell us a little bit about what that means.

Absolutely. So in plain English, I think everybody uses Netflix and streaming applications. When you're using these applications, you can pay per month or you can pay for every content that you consume, right? So generally the global economy, the business model has changed from a salesperson selling you a 3-year contract with X dollars per year. To usage-based pricing, which is pay for what you use. So we use that model for ROCS where these large businesses install ROCS and they don't have to pay per employee or per seat. They pay for how much these agents are doing the work. So let's say I have 10 customers, I get 10 agents. If these agents do an email, that's 1 action. If they create a slide deck for me, that's 20 actions. Do a campaign, it's 100 actions. If they don't do anything, there's no actions. So that's how we kind of monetize. And I think it's a win-win because customers are looking at AI and agents as a way to augment work or automate work, and they want to pay for the work that's getting done. From our perspective, it basically aligns kind of the ROI is the more work we're doing, The more we can command the right to charge. If they're not doing it, you don't have to like pay, like that's the point.

And I think you're touching on the other aspect of the business model that's fascinating, which is that you have agents that are assigned to customers. So you work on a per-customer business model rather than traditionally, I think software systems have been sold as like a per-task thing. So you have one software that you have for research, another one for like warehouse filing, et cetera. But you guys have one. Assigned per customer, which I think is something else that's really enabled by the AI era where you can have like hyper-personalization. So one agent just goes really deep on one customer.

Absolutely. So there's a lot of like technical and kind of business strategy behind it, but the most kind of simple way to think about it is we stumbled into this by working with customers. But the story that kind of encapsulates it, I was sharing kind of how, what we're doing and somebody said, Imagine like Barack Obama's going to enter like a meeting. What happens before the meeting? He probably has a team of McKinsey analyst types who give him a brief like, hey, here's what's going to happen in the meeting. That analyst is probably the best at their game and they research everything. What if you had this kind of virtual like team or army of AI agents kind of working for you? And that's kind of been the mental model is How can we have agents one per customer and each agent, all it knows, it knows everything about the customer and it's going to work really well for us because a little bit of detail is when you look at AI and agents, we are prebuilt agents. You can go to rocs.com and try it. It's available at ksa.rocs.com or by email and is useful out of the box. There's no configuration, you don't need consultants. It's only possible because we have these prebuilt agents one per customer. It has a lot of context and it gets better with time. Can I end this answer with saying, it's a little kind of technical here, but AI is only as good as the context of the data it has.

You literally predicted my next question.

Exactly, so once we tried to create this kind of universal agent or task, it just didn't work in our domain. Once we pinned it to a customer and then we had all this information, it started to work.

I would then push though, because as a consultant in my former life, I spent a lot of time with some of the largest companies in the world and you would expect them to have great data. And when you look into their Excel sheets and you see actually what they're working off of, it's terrible.

Exactly.

So how, you know, you're given these terrible databases and you need these agents to somehow extract high-quality information from it.

Well, yeah, it's a phenomenal question actually. So there's kind of two ways to look at it. One is like technology up and then business problem down. I think business problem down, the major shift is when you use AI and agents, you're moving from deterministic recommendations to probabilistic kind of recommendations. So before LLMs, an analyst would do research and they would have a robust kind of Excel model and a slide. This is true. When you're using AI and LLMs, it just by the nature is probabilistic. It will not be perfect. It won't be 100% true, but it can get you 90-95% there faster, like 100 times faster, and you'll be able to explore a lot more. So when we kind of build these, we say, look, that's why we really want the human. The human needs to use AI to make them dramatically more productive, maybe over time 10x more productive, but you still need the human, a human to make the judgment and be that. It cannot be autonomous. And then on the technology up, these agents are really good with dealing with kind of messy data. The old history of having Salesforce and SAP and Oracle, these systems of record, has actually broken. Most of the context that you need to make these agents are no longer— they're there, but they're also the warehouse and other tools. So we built a system where we can make all this messy data available. We use the LLMs to reason about it. What do you think is right? We just ask it like, hey, what's accurate? Then the server responds and the response, because we cannot be supremely confident, we say, hey, here's what we figured out and here are the sources. And we always say like, hey, you should first review it and then kind of present it. I think that is just the reality of where the products are. Now the con is it could be wrong a few times. The pro is you could, the scale is 10, 200, 1,000x. Now what would take a McKinsey analyst 100 hours can be done in an hour, for example.

I wonder though, and maybe this is me just trying to retroactively justify the pain I went through in my initial years of consulting is I learned so much through going through those raw files and like trying to extract insights from them that eventually I wasn't the, I was managing someone else who would go through those raw files, but I was able to direct them better because I had done it myself. So for someone entering the workforce today, how do you become that amazing sales agent without that grunt work before?

Yeah, it's a great question. I think it's, when you look at AI and agents, they're going to go in through, I would say in some way in our domain, if I may call it like white collar work, like office desk work, right? Coding, customer support, sales and marketing, finance and accounting. And each one of those, like if you believe like our mission, how do we make the best better? We believe like the top employees will get 10% more productive and those people are best because they've kind of gone through the ranks. So if you look at experienced employees today like yourself, myself, who've gone through it, a lot of our intuition is based on us doing all that grunt work on our own. We can see pattern match and we can do that. I think that will die out, like it'll decay out. So we're already seeing engineers who are 18, 19, 20 who are phenomenally sharp. They grew up with Snapchat, TikTok, now ChatGPT, and then so next generation of people, especially here I learned 70% are I think below 25 or something. 35, yeah. 35. These folks will grow up with ChatGPT, it's just that they will build intuition in a different way. They will not— it'll be hard to force them to go through like 100 data rooms and read decks and, and she used the word intuition, but intuition will be important, but it'll be very different. And very specifically what I'm learning is with AI you can just explore the search space a lot more efficiently. Let's say you want to know how to do accounting. With AI you can learn all the concepts, you can search through accounting files, find patterns in a dramatically more efficient way. And then kind of you will kind of build these attentions. But if you ask an accountant, no, no, you can't use ChatGPT, you have to be in a room and go through spreadsheets, it's just not going to happen. Yeah.

Okay. So do you think, you know, the future of becoming a more experienced salesperson is just based on who can use AI systems better and like sort of just trains themself on searching better? I think what your system is built on as well sort of mirrors that, right? It's like 3 layers, I believe, of going through data, sourcing that intelligence, and then customer outreach as the final step. So talk us a little bit about how you arrived at that 3-step model.

Exactly. So I think it goes to like the— these are these kind of, I guess, I'm not a liberal arts major, but now I have to be because a lot of this is really deep philosophical questions. Like our philosophical conviction, mine is I've seen consumers with Siri, warehouse workers, blue collar workers, white collar workers is I'm very pro-humanity, I'm very pro-technological progress, and I'm very pro-the best getting better. And so our view is always like how do you find the best humans in these organizations who are change agents and how can we make them like 10x better? So the hero or the heroine of our story was always like that, that human. So the idea was how do we kind of empower them with a team of agents who can kind of make them more productive. To make them more productive, yes, we would automate a lot of things that maybe analysts would do before or other support people would do, and that's where like automation comes from. So that was the core idea, and then from there we laddered down to these 3 steps. First is to deliver these agents, we need to build a system where you can manage hundreds, millions of these agents. So that's the agent swarm technology that we pioneered. Now it's pretty, pretty, it's a global standard. We build something where some companies have millions of them. We have about 30 million or more agents now. So that's the top layer of the cake, or maybe it's the icing on the cake. The below is a context system. So to make agents work, you need to proactively tell them or kind of reactively tell them to go to work. And it's an event-based system that's looking for changes. And in the bottom level is the data layer. So the data layer has all customer information. So public data, private data, my emails, all that's feeding in and it's all in one place. And then the context layer is looking for changes. As the changes happen, like the agents do work. And that's the kind of three-layer system. And now that three-layer system is pretty much kind of the standard pattern now to build agentic applications. We were very opinionated about the data and the context layer, the UI layer, the agent layer. We actually started off with this kind of, we call it the God mode agent, we did everything and we kind of broke it down into these one agent per customer pretty early, about mid last year.

Oh, interesting. Yeah. And you were able to iterate on the model fairly quickly.

Exactly. So kind of one kind of advice is the market and the consumers and the prosumers just do not care about our opinions. So the best way to kind of figure it out is to work with customers. So we build all of Rocks in the open with working with customers. Traditionally building these kinds of hardware systems or software system or consumer applications like you were telling me about your family, you would spend about 12 to 24 months just building and then you kind of launch it. In AI, there's not an option, right? So we worked with customers from day zero and they told us, look, if you have this one kind of God mode agent, it's just not very useful. So we kind of decomposed it, broke it down into agents, but the underlying kind of data and context layer stayed the same.

Okay. And then I'm curious about at the very top layer, so like the customer interaction. What the AI excels at and what you think the humans are still better at. So my hunch for some reason is that humans would be better at finding totally new customers. You know, they would be better at the cold call aspect and the AI would be good at sort of the maintenance. But I'm curious what you've found out with your customers.

Yeah, I think that's real intuition, right? In general, humans are better at strategy. And LLMs are better with execution from known playbooks. Again, strategy, a professor used to say, is the word you use when you don't have data. Well, strategy is like coming up with new playbooks or hypotheses, and LLMs are, you can, they're pretty generative, but they're much better in our context to like execute a playbook. So what does that mean? So what are agents really good at? If you know how you want to monitor your customers or map the market is really good. So we do full market mapping, market monitoring, market segmentation on kind of known players. You can say, okay, look at the Saudi market, break it down into government, PIF companies, sports and entertainment, and let me know when there's an RFP out. Like those things it's really good at. Then once you reach out to them, find them people, draft me, like draft an email, don't send it. So it's really good with that. What people are like, humans are really good at is figuring out, okay, I need to break the market down into PIF companies versus sports and entertainment versus something else. Like how do I think about the market sizing? So that is strategy work. And then actually like emails, emails are very good ones. So I think email seems super simple. I'm going to draft an email and send it. Actually, it's probably the hardest thing to do. Because when you actually send an email or text, you're representing the human, it's coming from their inbox, people actually view it very— take it very personally. Like they don't want their communication to feel robotic, especially in our kind of end relationship kind of game, like especially Saudi, like it's a big relationship market. If you get one CEO to talk to another CEO and the email was super robotic, you will lose trust. So what we do is we actually draft the email and actually have the human review it. Over time, if they can get comfortable and say, okay, this is good, autonomous, go with it, great, but we don't automate it. Actually, that's not yet happened. And that's kind of where I view it again, like humans as the orchestrator and these agents doing kind of 80-90% of that work.

And AI, I think, is especially good at that because you're only going to get more knowledgeable with time. The context window will grow larger. They'll learn more about the universe they're in versus legacy software, I guess, sort of stays stagnant and eventually falls behind.

Yeah, absolutely. Just to give an idea, I have 6,000 enterprise customers that I personally manage with Rocks. We have, at Adam's here, we have a 6-person BD team. We have no analysts.

That's incredible. For a point of reference, how many people would that usually take?

Tense. A lot, because imagine if I have 6,000 enterprise relationships, I would— these are large businesses, they're global companies, a lot of them. We would have a dedicated analyst team, a data team, an ops team, we're constantly like kind of doing work for me, but a lot of that work is done by the agent. And then we would have a large sales team that's doing like all that work. Now we have a small elite team of people who can build the relationships, can fly.

So they're each managing like 1,000 customers.

Yeah, we're an enterprise company, but tens of them. Yeah, traditionally they will use something like tens, hundreds of them, right? Yeah. And that's the scale, right? Like now it's not that we're not growing or hiring, we're actually growing and hiring like crazy. It's just way different. Like we could— we've been only public for a month, we've been in valid beta. Like we could now grow faster. I could grow into Saudi Arabia, I could grow into Europe, just the business can grow faster. It is because instead of hiring folks to do things which are probably not customer-facing, we can hire people who can go work with customers. And I shared that example and we see that across our customer bases. People who are using it, it's like my brother's a surgeon, like he's probably the most, he's a power user of generative AI tools, like the best folks are so supercharged, they can just do a lot more. And that gives me a lot of pride, like I feel like wow, humanity can just go do so much more now. And there's just no going back. And that's kind of what gets us excited is like how can we help like Humane in this market and other markets like really adopt these better so that that company obviously grows, but more importantly, those change agent, those companies, they can get supercharged.

How do you feel about the sales relationship getting affected at all? Because obviously the human element of it is really important and we touched on that a little bit with, you know, making sure they read the draft of the email before it's sent. But when the customer gets the sense that, you know, in some way or another, a lot of their interaction is getting automated, do you think that there's a fear or a dislike of this? Like, or do people say, I want to have a person dedicated to me?

Yeah, absolutely. Like, I kid you not, like, it's the most common thing that I get asked by media and investors, like, oh, like you have millions of agents, it's the fastest growing agent in the world, we consume like 3 billion tokens, everything's going to be automated, agents talking to agents. Going back to me in a warehouse, I was like, I just don't believe in that. Like my view is that your most important relationships, humans will be important. Like how do we supercharge them? But they still will be very involved. Because the con of it is just complete loss of trust. So for example, you reached out to me or in your relationship than mine, like if I am building a relationship with somebody else and they got to learn that I used an AI or my AI reached out, they'll never trust me again, right? Like I think so that's why kind of our focus is how do we make the best better so that they can kind of scale that human relationship that they're there. And everything in the background is kind of automated versus full automation, at least in our market. Yeah.

Interesting. And you work with some of the biggest companies in the world, which is super exciting. I mean, only a year after launching Copitley, so congratulations on that. Only a month. Only a month, sorry, sorry. You guys came out of stealth mode last—

Last year we went from private beta to public beta in November.

Incredible. So I'm wondering also what is the value that you're providing for— I mean, obviously you're providing a lot of value for your customers, but in an age where software creation is becoming easier and easier for everyone, how do you build that defensibility, that moat where people still want to come and use your product instead of trying to build it in-house?

Yeah, absolutely. There's many ways to answer that question, but I'll go back to your really like heavy but difficult kind of example like Amazon. What we're seeing is yes, we've been able to acquire these customers because we're a really good applied AI team, we're some of the best AI engineers, we have good references, we're backed by Sequoia and GC, and we have great partners like Humane. But more importantly, you have business leaders across the world who have this mindset that they have to transform or they risk losing behind. There's literally not a business leader in the world who's not thinking of this as an existential moment where they have to transform or in some ways say transform or die. That's kind of the core mindset. So and when— so every customer that we acquired is all inbound, it's all word of mouth, is because they have an ongoing board-level conversation about productivity. Hey, like I've rolled out coding agents in R&D and engineering, what do I need to do on my go-to-market side? So they all kind of came inbound. And once you're in there, The differentiation I'll talk about, kind of technical, is ultimately time to value. When the business owner, whether that's small, small media business, mid-market enterprise, if they have the mindset that I need to transform, they will build trust with partners who can help them get there faster. So the email and text I generally get is I have incumbents who are massive companies and that's going to take me 9 to maybe 18 months. I have horizontal systems where I need to go build my own things. I need to get some early like ROI or wins in a quarter or two quarters so that I can change my plan for next year. So I'm still going to do those. Can you all come in and actually like move fast? It's like the Humane example. It was one conversation. We were live in the market and deployed in a few months, right? So our core differentiation is how can we build a company that from the ground up, from the team, the product or delivery model, we can basically do the fastest time to value for AI transformation outcomes. So to kind of recap, like a Fortune 500 company reached out, the CEO's saying, hey, like to your point, like I'm being asked to increase productivity by 5 to 7%. I have all these folks who are just not moving fast enough and I just have 2 quarters. I need to kind of lock my plan for next year. So we came in within a quarter, deployed, we proved more than 10%, and then now we're kind of rolling it out globally. And we were able to do so because we can deliver like really fast. And on the technical side, we're able to move really fast because from a technology perspective, we run inside the customer's environment. So we don't have to wrangle all the data and our agents are pre-built. So because of those, we can really deploy and now we build a forward deployed kind of field team that can go in and try these kind of transformation projects.

I see. And along that similar vein, obviously people are insatiable when it comes to AI. As soon as they see one model, they're amazed by it in the first week, and then immediately they're like, okay, what? I want it to be better and faster. So how do you track and make sure of the improvement of the system over time?

It's a great question. It's one of the best questions I got. Yeah, I didn't do it a lot. So I try to kind of ground it in like fact truths which are like hard to argue. And then a truth that I've uncovered now, which I didn't even uncover at Apple when we were doing Siri, is that the consumer and in relation to prosumer expectations is this invisible asymptote that we will never meet. Whether that's our grandmoms or parents or our kids, they expect software to do everything for them. Because they've been sold the vision of singularity. They use ChatGPT and it seems like magic. When they come with the workforce, they want software to literally do everything. So I don't even expect any consumer, prosumer to come and say great job the second time. The first time, oh, this is interesting, because by the next week they want some, like the expectations change. So to answer your question, how do we think, how do we do better? First is because our core place in the market is we can outship the market. We're the fastest movers. We have to constantly be delivering the highest quality models, right? So if OpenAI or Anthropic or Groq here, our partners, are able to like push the envelope, we have to be able to productize it fast. So speed really matters. Second is from metrics perspective, what we track is essentially how much work are we doing. If people are using ROKS, our core unit of work is agent actions, like how many actions are they taking? If we are seeing it as those best power users are using us voraciously, they're using us constantly, then we feel like, okay, these agents are doing a ton of work for them. And if they're not, that means either they don't trust it or we didn't train them well enough. So yeah, so kind of at a high level, we literally have to be the first to market for every new model, every new capability. If you're behind, then why would they trust us? They would trust somebody else. The second is we try to see, can we be a daily driver? Are people doing work with these agents and is the agent actually improving?

Okay. Is there something that you think would lead to a significant breakthrough for you guys and what do you think that would be? So if we narrowed it down maybe to the data, the quality of the data of the customers, is it the underlying foundational models you guys are building on or maybe it's the the business model itself, out of those three, what do you think would be an advancement and create the greatest unlock?

Yeah, maybe the second one, obviously. So we are in some way the fastest moving surfers riding the wave of the foundation models. So as the foundation models capabilities increase, our product kind of gets the benefit faster than any incumbents or so. So the biggest unlock will always be number two. Very specifically now we've made, because our kind of product does work and does work out of the box, you can actually tell it to do, hey, I'm gonna meet, I'm gonna go do this podcast, what should I talk about? Can you send them an email? And after that send them a recap document. You can command, we got a command. As the kind of underlying kind of models improve, that is going to get a lot more capable. So imagine a world where whether you're on WhatsApp or Slack, where all you have to do is like speak to it and/or kind of text to it and it'll go do like really complex work. It could, the work could be multi-step to the point, go reach out to them, tell them I want to come tomorrow, send them a contract. Once they sign it, can you just talk to finance? Like that kind of work. So that's probably number 2. Is the most important thing. In terms of the business model, I think it's awareness. In some way, like, people like us, we are looking for these transformative leaders across the world, like you all, like Tariq here, like there's not that many. I think people talk about tech and innovation a lot, but the reality is an extreme power law. There's a handful of people who can see around corners and detect the waves. And most people are not that. So it's kind of driving awareness for what we're doing, our mission, so that we can find those people who are looking to make that change more effectively. I think that's the second part, right? So first is product, it's kind of foundation models, then we just have to be, as they say, get on planes, try to meet these people. The first one you talk about data, that we feel like The company's actually called Rocks Data Corp. It's not Rocks AI. It's because like our core thesis was always that it's all, everything's about context. We are a product application company and we work with intelligence layer. So we feel really good about, we've already built something that runs in the customer environment. That's why we can deploy in Saudi Arabia, like FaaS or any other kind of area. So I think that's kind of more execution. I would say like foundation models and awareness. Okay.

I could talk about this forever. I want to get also some insights from you from, you know, more about life in general before we run out of time. Because you've obviously have so much experience and you spent so much time building different products at different companies. Would love to know, you know, what were some of your biggest wins, biggest losses, and what were the takeaways you had from that?

Professionally, personally, or anything? It could be anything. Yeah, good question. Biggest wins and biggest losses. I think the high level, like the long, I feel like I'm just starting. So it's kind of crazy, I think, okay, I have gray hair, I got 2 kids and you extend. So I think my wife's amazing and we've been together for a long time and it's kind of a team effort honestly. So it feels like, oh damn, like time, like Number one thing is I still feel like I'm just starting, like there's many hills to climb. I'll probably keep building companies till my last breath, like that's what I'm here for. But yes, I am 37. So the biggest learning is time does not stop for anybody, so everything has to happen today. So that kind of drives the urgency. I think biggest wins and losses, before wins and losses, the longer you do it, you learn that you have to enjoy the process. And not go— the highs can't be too high, the lows can't be too low because the whole game is to kind of keep the slope high and keep this going. Like I'm somebody who obviously comes from a different background. My mindset is I'm going to outlast everybody else. Like how can I be persistent over a longer period of time where as long as I'm growing, I'll kind of keep doing, being a better parent, better husband, better founder. So the losses and wins, I don't think a lot about honestly. Having said that, most days are bad. That's a reality of an entrepreneur or somebody who's gonna drive a lot of change. There's 10 times more L's or down days than there are up days. And the way the brain works for people like us is you don't celebrate the wins and you actually like really like get obsessed about losses. So I think that that's kind of the frame. Just to call out kind of wins, I would say like unquestionably being a parent is like incomparable. I think it's just, that was probably like nothing compares. I think that's number 1. Number 2 is through my experiences, the number 1 thing that kind of motivates me is as in someone entrepreneur or a capitalist, like the biggest thing I can do for society is to build companies and through companies it does change the lives of our employees. And so everybody who I've worked with and then being able to go build, buy homes and have families, that's probably extremely rewarding. And the third thing is, at least for me, I like to go after large existing markets and take a generational swing. Like how can we define the next generation? Whether that's Kiva and robot, robotics, or Siri or Kubernetes or here is like, These are all massive industries and how can we build the next thing? So I would say those are the 3 highs, like parenthood, being able to help my team members get some sort of financial stability. And then the third is kind of leaving a mark in every industry. I think the downs, I would say immigration is the problem. Like you just have, if you're an immigrant, if you've been that journey, you've been in the US, It just, you just have to be patient.

The lottery, particularly for Indians.

It's crazy. Yeah. So you have to have a level of patience and with my wife, like a support system where you can persist and you kind of figure out how you just focus on your own lane because you can easily look at LinkedIn and Twitter and look at everybody else like, and those things do mess with your brain. I would say like that's a pretty big one. Like the previous company that I started, and there was a visa issue and I had to change my role in that company. So that's number one. Number two is like the loss is not a specific example, is kind of compartmentalizing is probably like the hardest thing is when you're kind of going back, especially now with AI is the level of urgency is so high. You only win by execution at like crazy velocity is you just don't have a lot of time to just sit and sulk about your losses. You just don't have, like you need to just snap out of it, move on, be a better parent, be a better operator, like really fast. And I have a really hard time with it, yeah.

Do you think being a parent has changed how you do business and look at business?

Yeah, unquestionably. Awesome. Yeah. There's like, I would say, my wife's gonna listen to this, like we like, I didn't think a lot about parenthood before because we're kind of obviously building companies and doing what we do, but like nothing compares, like nothing. It's just that, like I can't even imagine like pre-parenting or parenthood life. So I think it changes everything about people, so So as like an example who stumbled into parenthood, I think it's like, it's completely transformational. So what does it change practically as an entrepreneur, as an operator, is just the value of time. And I think everything is doable. Like a lot of people ask me or others, if you look at Steph Curry, he's an NBA player, he has I think 4 kids. Like you have like phenomenally successful people who are great like family men, women. The way you kind of perfect that is literally every second matters. There's a line, I think Rocky 3, which is there's no tomorrow, get everything done today. Right, so like every day has to get executed like in a tick-tock fashion and there's just no time to just hang out. So I think that changes and from a work perspective what you give up, hobbies, friends, a lot of things that you obviously like are fun, but you just don't have time for.

You have to be brutal about your priorities.

Exactly. But what you do get is focus and urgency. And just like, okay, what is the main thing? How do you keep it the main thing? And the urgency is that like, I could rather go play with my kids and like, how do we really like focus? I think But a lot of it is like I'm brainwashing my brain. I think if you were not a parent, you would also, you could be an amazing founder, right? Like everybody chooses their own paths. Yeah.

And as a father now, when you think about your children's path in life in the future, how do you think that would differ from what you did or what's conventional now, you know, sort of like getting a degree or multiple degrees and working at these like large companies, do you think that path changes, how would you advise people who are entering the workforce today?

It's a great question. Honest answer is I don't know. Honestly, that's probably the most common question I ask like other parents who have kids who are slightly older than us or going through that journey. There's actually in some way no— obviously like I do ask my own parents and people who've gone through it, but the folks who have kids now in kind of high school and kindergarten—

There's no precedent.

There's no precedent, right? So this is probably the number one question I ask anybody is because I'm trying to like absorb and figure out like what's going on. Will it change? 100%. Like I would say like a shout out to my team. Like there's a lot of folks who are early career for Jawada School. They're just smarter. They have a lot more exposure. They're just so sophisticated. The best ones can also complement that being like patience and focus. I think a lot of times You can also get into like overexposure where you can't really decide what to do. Everything's— the world's an oyster. So I think if you kind of play that forward, like my kids who are like 4 and 1, so very young, they will have more information. They have a lot more tools. They will have a lot more exposure. So that's only going to expand. There's obviously like fear about iPad and YouTube and things like that. I think that's kind of the gray area to kind of figure out, but that's where parenting comes in. But there's some truths as well. The truths are, I think hard work is what we've learned is probably the most important thing that will always be true. If you can't work really hard, like insanely hard as Jensen says, nothing's going to be— you just won't excel. Second is curiosity, is even if you do have ChatGPT or all these tools, or you and I, like we learned English by watching like American TV, we would still have to go to that channel and look at Friends, right? So is it ChatGPT? Somebody has to go ask those questions. I think the curiosity will always be there. The last thing I think has worked for me is I think the value of being well-rounded, being a global citizen, having exposure will always be true. I think humanity is a lot more similar than people might think. So I think that's kind of that gray area is what my wife and I and our friends are trying to figure out, okay, how do we do that? But those two aspects that they'll have a lot more information is true. And then kind of hard work, curiosity, and just being built up on balance, I think.

I think that's fair. I was talking with Anjad Nassad, the CEO of Maplit. Yeah, he has kids too. Yeah, exactly, two days ago. He also has kids asking him similar questions, and he really believes it's also the age of the generalist. So I think probably the best advice you can give is to, you know, like make sure your kids are curious about anything and everything and just let them sort of explore the earth and figure out what excites them and what they can excel at.

Yeah. You think, you know, the harder part is just like, which I don't know, actually my wife and I and a bunch of friends, we were not, we didn't excel early because we were like generalists. Like I, we liked everything from soccer to TV to math to physics to pretty much anything.

Which is human nature.

But we grew up in India. There's only 3 lanes. You got to be an engineer, go to IIT. No, that's 2. A doctor. I mean, my brother was obviously super sharp and successful at both. My parents went to IIT. So you were kind of these kind of black sheep or whatever you want to call it. So I think US, I'm also seeing it like when you're especially in Silicon Valley and there is this race to get in the best school and get CS at Stanford or Princeton and there is that kind of race. Making sure people are generalists and just kind of well-rounded. At the same time, they're able to go where they want to get to. It's kind of a balance, like the competitive thing, which is I think we'll figure it out, but I kind of think about, okay, like when they're doing something, that competitive urge is coming in, like, okay, you got to be the best. But I don't think that's the right way to think.

Yeah, the Taipei behavior of like, okay, like stop and smell the flowers and like be the best at flowers.

Exactly. Yeah. So that's kind of the breadth versus depth thing is I think breadth is key, but then once you choose something, then I think that thing comes in. But again, it's to them, they'll select what they're good at. As long as it's a viable path for them to be happy. I really truly believe, I think Amjad as well, as immigrants, we had a great upbringing growing up because we're different world. But when you go to some other country, you are a sovereign individual. You have to figure everything out on your own. I think as long as people, they can figure out their own lane and be the best at it, I think that's fine.

Amazing. And then one last question, I think, and it kind of wraps up maybe a lot of what we've been discussing today is what do you think makes humans special? So what remains unique to us in the age of AI?

I'm like an eternal optimist. Like I'm probably the most anti-Luddite you would see. No, not remains for me. It's like these are new frontiers. Like every day when I build this company or see my kids, I feel like, wow, there's so much to do. There's a universe to explore, there's so much to learn. So my view is what we have about 7 or 8 billion people on this planet and we're probably trying to grow them. And a lot of them are in India. A lot of them are in India, China, the Middle East is growing. Like everybody with Replit and ChatGPT and hopefully in the business with system like Rokst, they'll be able to do so much more. So what really excites me is I do believe in the power law. I think there's 10, 20, 30% of the people, I hope we can increase that, who are just in our DNA is to detect those kind of waves and kind of ride them and build new things and do art and do sports and do new things. And I think humanity is just we can do 100 times, 1,000 times more. So for me, what excites me is we have our time on this planet, but like our kids, they have probably like 80, 90 years out and like what would that world be? And I'm not that, I guess this is naive of me, I think humanity is a lot more similar than they are. I think there's these truths and facts which is a lot more similar and most people kind of are more center than they are on the extreme. And those kind of 10, 20 kind of outliers, those folks will just build a lot more. And yeah, that's kind of a fuzzy answer, but I'm a bear, so I'm pretty excited about it.

I believe that being optimistic makes sense.

Yeah. Only good news is the way to think about it.

Yeah. I really appreciate this conversation. It's been a real pleasure, Ishan Mukherjee.

Yeah, absolutely. This is like phenomenal questions and hopefully it's helpful for you.

Thank you. Yeah, thank you.