Kleiner Perkins Fellows Summit: Building Glean with Arvind Jain and Mamoon Hamid

Well first of all thank you for having me and I'm really excited to chat in front of all of you guys here. It's really exciting, such a great you know we actually have quite a few KP fellows in Glean as well so I know how awesome it is. Thank you so much. So anyway, coming back to 1997, 98, so that was actually when I was just starting my career. The first thing I did was after graduating I joined Microsoft. I was there for two years. It was a large company, good education for me, but wasn't fun enough and I wanted to in the middle of that dot boom and the internet was just sort of becoming big. Amazon was actually just a bookstore at that time, online bookstore. But there's a lot of promise. There's a lot of promise and everybody sort of just like you know how you feel about AI today, that's how all of us were feeling about the internet, that this is going to be the next big thing. And I wanted to be at a company which was foundational in terms of internet technologies. So I joined Akamai in late 99. And Akamai was the company that, you know, our mission there was to make internet work. In those days the internet was actually very, very slow. Most of us had dial-up modems, connect to the internet through the phone lines, and there was no way you could do things that you can do today, like videos on the web, not possible at all. Pages would take 15 seconds, 20 seconds to load, and all they had were a few lines of textual content on them. And so I joined Akamai. Their mission was to make internet fast and instantaneous for everyone. So that's sort of what I did. And that was my journey into the dot boom. And then later on, the past as well. It sort of reminds me of this app that I've been using where you prompt a photo that you want to see of yourself. It takes like a good 20 seconds to get a result, which seems so antiquated for anything. It just shouldn't take 20 seconds. And it's not too dissimilar from what you would encounter in the internet 1.0 era, which is you couldn't even watch a video and it took companies like Akamai to do all kinds of cool technology and caching. And so we're seeing a similar sort of paradigm that we're going through right now with AI.

Yeah, absolutely. Well, one of the things that I have always felt about technology is when something new comes, there is a lot of excitement and also hype. That's what we're going through with Gen AI today. We saw ChatGPT and suddenly we thought that it could do all these amazing things, and then after a little while, some reality sets in. But what I have found is that all these big technological moments, although there is this hype, we also still underestimate what's going to happen in the long run. When the internet was starting, nobody imagined all the things we could be doing on the web, even though there was a lot of promise. People's imagination only went to some degree. I remember one of the things I worked on at Akamai was basically trying to introduce video on the web. And I think our biggest success was streaming March Madness online. That was a big thing. But actually, if you look at the quality, you still had big TV screens in those days and it was a great experience watching the basketball match on the TV, but on the internet it would be like 1 inch by 1 inch, the size of the video, and we were happy with it. We were actually happy that now you can stream on demand when you want. But we didn't anticipate how video would completely transform and the only way to consume media and video would be on the internet. 20 years later, I think we're going through the same moment with Gen AI. Right now, people have seen ChatGPT and we feel like AI can actually do some great things. We can take all this knowledge that lives in the world or in our companies and make good use of it, make the knowledge accessible to people. So there are big use cases that people have in mind, but I think we're all underestimating what AI is going to do for us. I feel like in 10 years from now, 20 years from now, the way we all work is just fundamentally going to change. Like a lot of things that we do today is something that we won't be okay doing 20 years from now. One simple example is in 1995, if I were to ask you what was the GDP of United Kingdom 20 years back, you would be okay going to the library, spending one day trying to go through newspaper archives and get that data and come back to me. You're okay spending a day answering a question like that. Now today, you absolutely won't be okay doing that because you know that you can punch that question in Google and get the answer back. A true story for you: as a kid, I grew up reading encyclopedias, memorizing facts inside of encyclopedias. This is 1985, 86. And I wished for a place where I could just enter in my question because I had a lot of questions as a kid and get the answer right away. And to know that 10 years later you could actually do that with technology was a seminal mental moment for me when I first heard about it. I was still in college at the time. But that is the beauty of technology: 10 years forward, something I thought was impossible had happened. And I think we are in that moment with Gen AI. A lot of things that we think our machines are not capable of doing are very quickly becoming capable of doing those things. So two years from now, forget 10 years, two years from now, if I ask in a company, you have this question: 'Hey, you designed this system, why did we choose this architecture or why did we choose Java or Python?' You want to answer that question today, you would be okay doing that research, talking to subject matter experts, figuring out what happened historically. You'll be happy if you can get to an answer in a few hours or a day. But your expectation two years from now will be: 'Hey, that's a factual question, the knowledge is within the company, and I should be able to just ask that question to my assistant and get an instant answer back.' So a lot of things like that are going to happen. So my feeling always has been that with any new technology that is seminal and amazing, you overestimate what the technology is going to do in the short run, but you underestimate what the technology is going to do in the long run. That's sort of where we are with AI.

Well, that's actually interesting. Google was actually anything but hot. Or rather, it was not the hottest company at all when it originally started because there were plenty of search engines in that day: Lycos, Excite, AltaVista. There were a range of them, and none of them were actually the way they marketed, the way people thought about search engines. That was not the hot thing. The hot thing was, for example, I joined Akamai because it was actually hot, being rewarded in the public markets with massive valuations. Google was actually private and nobody took it seriously for some time. The company grew after a few years of existence. So in 2002, all the hot companies of the internet actually had most of them go under. There was a big bust. A lot of these technologies which actually made sense were just a little bit ahead of their time. And Google was actually one of the only companies left standing and doing well. And to me, whenever I picked which company I should do a job at, I was sort of lazy in terms of evaluation. I never had fancy mechanisms to look at the TAM or market analysis to figure out that this company has big opportunity and therefore I should join. I always went with very simple observations. And my observation with Google was that the internet is growing and there's a lot of information out there, and today my only way to get to that information is through Google. This is going to be the gateway of all the information on the web. It seems like an important problem. All of us are going to use it. So I've always made my decisions that way, in very simplistic terms: if something adds value to a lot of people, that means if you execute well, it's going to be a big thing.

Yeah. So in 2005, Google was growing very rapidly and we felt that we had hired all the engineers in Silicon Valley that we wanted to hire. And now to keep building the talent, the company felt that it had to go outside of Silicon Valley and go all around the world because the best engineers are all over the world. So Larry had this mission of establishing presence wherever big talent is. And India was always seen as one of the key locations for amazing tech talent. So they were looking for expanding. It was actually an interesting story. When I went to establish Google India, I had never managed any person in my career, not even an intern. So I had no experience managing. I only knew how to code. So when I was asked to go and establish Google India, I told our executives that they were probably mistaken, talking to the wrong person. I had never managed anyone and had no idea how to do this. But they said no, that's fine, we want to grow and nobody in the company knows how to manage, so let's learn on the job. That's how it came about. They asked me to go and build the organization, and I went there and learned as I tried to execute.

Well, I mean, when I was 22 years old or 27, I always had this dream or fascination with starting a company. I wanted to, in fact, but I never did. I realized that oftentimes I would come up with ideas but then kill them myself, in the sense that it's hard to be an entrepreneur. When you are not ready mentally, you'll find some reason to find holes, gaps in your own ideas and decide not to do it. Then I joined Google and Google was a great place for me. I got to build amazing products that I would see everybody around me, my friends, my family, using those products. So that thought of starting a company disappeared from my mind. And when we started Rubric again, it was actually not on my mind that I want to be an entrepreneur and I have to do it. I think successful companies get built when you run into a problem and your desire to go and solve that problem gives you 100% conviction because you've seen it, you've experienced it, and you're going to persist through the ups and downs of starting a company. So that was my journey. I was not intending to start a company when I wanted to, I didn't start one then. I didn't want to start one, I just sort of ran into it.

Yeah. So in fact, the idea for Glean came when I was at Rubric. At Rubric, we had the good fortune of building a product that I think everybody in the world wanted to buy. So we were able to scale the company very rapidly, and within four years we were more than a thousand people in the company. But as we experienced that hyper growth, it also came with a significant reduction in per person productivity. People were not able to do as much work as they used to be able to before. There were a lot of concerns around that. We felt that we tripled our R&D team but it was not resulting in the same increase in velocity in our product development. As we tried to dig into that, we did surveys and asked people what's wrong, what are the things we can do better as a company. The number one complaint we would always get was: 'I cannot find anything in this company. I don't know where to go and look for things. I don't know who to go and ask for help.' So we were not enabling our people. With fast growth, we didn't invest enough in organizing our company's knowledge and creating the right structures where people could get help when they needed help. So this was the problem. And for me, I'm a search engineer by training, so the first thought that comes to me is: of course we can't find anything in this company because we're using 300 different SaaS applications and our knowledge is all over the place. People use the tools they're comfortable using. Somebody uses Dropbox, somebody uses Box, somebody uses Google Drive or Confluence. This was happening across the industry. All companies were completely changing and redefining how technology worked for them, adopting SaaS. So there was this big problem in front of us. And me being a search engineer, the first thought that comes to my mind is: let's build a search layer on top of all this enterprise content across 100 different systems. As part of that, the first thought was to just buy a product for that, and we realized there's nothing out there. That got me excited. And this is what I was sharing before: again, I had no intention to start another company, but a problem came to me and I felt it was so important. Every person who works in any company in the world, we could help them so much if we build this product. That's sort of how Glean started.

Yeah. Well, first, thanks Moon for letting us be your guest for a long time. We definitely overused the hospitality from Kleiner Perkins, but that was amazing. The initial days of building a startup are fascinating. We built a great team. Tell me the first few hires you made because it was pretty rapid fire. You went from one to five of really amazing, incredible people. Yeah. So when we were starting Glean, I had already been working in the industry for quite some time. A lot of people who I really admired, amazing engineers, and I had already tried to hire many of them from Google when I was starting Rubric, and I failed because they were not interested in building a data protection product. It was not exciting enough for them. So I reached out to them again when we were starting Glean, and actually all of them said yes to me. I was surprised, but I think it made sense because at Glean we were solving a problem that actually matters to all of us. So it's easy to relate to the problem we solve. I think that gave us the ability to hire the best people. I never had to explain or convince anyone that this is an important problem to solve. If you have that kind of luxury, you'll have a great time hiring people. So that's how it went initially. I got back with colleagues I had worked with at Google and convinced them to join us.

Yeah. So search is actually a very high bar when you build a search product. You're always competing with Google even though they have nothing to do with enterprise, but still the bar for people when they come and ask questions is that all you have to do in Google is put in two words, you can misspell them, and you don't have to share anything about what you're looking for or the context. Somehow there's this magic that happens. Google understands what you're looking for and shares the right piece of information. So that's the bar we have to build to within the enterprise. Now Google has the luxury of understanding you because they have billions of people coming in every day and they learn a lot from that human behavior. We have to figure out how we are going to learn because in an enterprise there are a few hundred people or a thousand people, and you just don't have as much learning that can happen just from people coming to a search engine and running search queries. So that's one of the key challenges we had to solve: understand how we're going to build signals that will help us understand this enterprise better, the knowledge better, the relationships between people and knowledge. So a lot of work happened in that area. Another one was just how are we going to bring all of this enterprise content together in one place? All of these systems had APIs with quirks, and sometimes you couldn't get the information you need. So these are complex engineering problems that take time to solve. But I would say the biggest issue for us was convincing people to go and use our product. Imagine you're the CIO of a large enterprise and I come in as the CEO of an unnamed startup that nobody in the world knows about, and my request to you is: please give us all of your data, all of your information. We're going to take it and make it searchable for you, and we have the right security controls when you have no security engineer in the company. So that's the challenge. You run into these kinds of challenges when you start a company, and you have to solve them all. We figured out very quickly that nobody was going to share their data with us. So in this modern world where everybody was building SaaS products, we had to make a call that we're going to build an on-premises kind of product. It runs in the cloud still, but it has to run in the customer's environment. We tell them that your data stays within your cloud VPC and we will not have any access to it. And to be clear, there was no way this could be a PLG bottoms-up approach. That was another one. PLG was a fashion, and we couldn't execute on that because to deliver value to a single person in the company, you have to connect to all of the company's content. So these were some of the interesting challenges. It's interesting because the problem is not new. People have tried to solve this for a long time, but these are real big headwinds that you run into and give up. So a lot of people gave up trying to solve this problem. And I think that's the lesson for all of us: if you're going to try to solve a big problem which is not solved yet, it's because it's actually hard. If it was easy, it would have been solved by now. But you can solve it. All you need to do is persist.