I'm very excited to introduce our next speaker, Arvind from Glean. Full disclosure, we are investors in Glean, so I'll have to be nice, but only for the first 10 minutes, and then we'll riff through some hard questions. Arvind, you've spent a decade building Google Search, Rubrik, which is a phenomenally successful business, and now Glean. What have you learned from your experiences building search, building data, and now the age of AI that helps you at Glean?
Well, first of all, I'm really excited to be here. Thank you for all the audience for listening to us. But so with our experiences, I think those have been really, really powerful learnings for me that I got from Google and Rubrik, which I'll be directly applying at Glean now. If you know, if you think about Google, that's been the largest AI application in the world for decades. And so, my experience there working for over 10 years, we learned a lot about search, information, knowledge, like how to actually do the needle in the haystack problem, how to figure out separating good information from poor information. And all of these challenges now, of course, at Glean, we have to solve as we actually roll out AI in the enterprise in a very broad basis. Like you have to make sure that all your agents are able to actually work with the right, accurate, high-quality information, which is ultimately a search problem underneath. So, all of that experience is directly contributing to our product at Glean. And same for Rubrik, a lot of my learnings there were around enterprise requirements around data security and governance, which is actually a key topic for businesses today. Like as we work with the largest enterprises, the number one concern they have with AI is, you know, it started with 'is my data going to be leaking out of my enterprise and people will train models on it.' That has sort of subsided now, but there's still a lot of issues around governance and security. How do you make sure that agents internally are restricted to only information that they should be accessing, the actions that they should be taking. So those experiences have been fundamental. I mean, for me just from a product perspective, but of course building a company before prepares you for and that this is a difficult journey and long journey and gives you that stamina. So that's what I appreciate from my past experiences.
Yeah. Well, you know, I started my career at Palantir and one of the things that we used to say around there is that your AI strategy starts with your data strategy. And so you've been paving that road for a long time. Okay, we'll jump into the meat of this Arvind. So ever since I have known you and since we invested, you have beat every single plan. Every quarter on repeat, you're crushing it. However, every single quarter it feels like you have a new competitor. When we invested, there was a thing called Microsoft Copilot. 2024, there was this thing called ChatGPT. 2025, Anthropic's already eating the world. You're still crushing your numbers. How do you do it? What's the secret sauce? Help us separate the fog of war.
Well, first, Glean as an enterprise AI company, we started in early 2019. So that's well before any one of these other players came into the same business that we operate in. And just to be clear for the audience, our business is helping you embrace AI inside your enterprise, like giving your employees a tool which is like ChatGPT, like Claude, but that's connected to all of your company's context and data. So you can go and ask questions or ask it to do work for you and it uses all of that company's context plus the power of all these amazing models like OpenAI's and Anthropic and Google's to do that work for you. So that's what we do. Now, I think as AI got — it became clear in the industry that this is the future, this is a very important technology. It was not a surprise that all software companies came in and they wanted to compete in it. So, partly, the fact that we've been the first ones, we always get that credit. When Google comes in and builds, launches a product that looks like Glean, the market will always talk about us in a favorable light as being the pioneers. And so, that helps us with our customers. But otherwise, it's about hard work. It's about building a great product and achieving success with our large enterprise customers who have a great brand and then the word of mouth carries. So, that is what has been the key success driver for us is keeping focus on that one problem, only that one problem which is how do you build the best context graph and retrieval technology inside the enterprise and then partner with all the other companies innovating on other fronts. For example, we chose not to actually invest in training and building models ourselves because we saw that Anthropic and Google and OpenAI had a lot more capability and expertise and they were doing it much better than us. So, that's been the model for Glean is to not reinvent, is to not compete but rather partner with the best innovators and then fill in the gap because we're nowhere close, like you think about enterprise AI and its potential and what we can do, we're nowhere close to achieving it today. So, there is enough white space and we pick our own swim lanes to work in.
Yeah, focus is the answer.
So, Arvind, since you know, 3 years ago when we were studying Glean, I remember the biggest question, summer of 2023, ChatGPT's come out for 6 months. The biggest question was: will enterprises adopt AI? The question's changed quite a bit. There's companies who are spending half a billion a month on AI now and honestly people are now starting to limit how much AI or how many tokens you can spend. Tesla just announced a $200 a week limit. Other people have a monthly limit. So clearly in 3 years we've gone through the full cycle of will people adopt to will people ration. How does Glean fare in this? How do you think about pulling ROI or making sure that you're within the envelope? And I know your products have changed a lot over the 3 years as well. Talk a little bit about that journey from adoption to retention.
Yeah. So we started as a search company. In early 2019 you could think of us as Google for your work life. That search transformed into a work assistant like ChatGPT as models got better. And from there we evolved into becoming an agent platform, a horizontal agent platform that allows you to build agents across all of your different business, different departments like finance and legal, HR, sales, marketing. And from there we evolved into a fundamental horizontal AI platform that not just our own products but every other AI product that you buy or agent building system that you buy, they rely on Glean as a context provider to those systems to build better and high quality agents within their own frameworks. That's sort of has been the evolution of Glean. Now we work with large enterprises and the number one thing that we hear from them, and in fact this has changed in the last two or three months. Before that, the conversations were always around, okay, help us educate, like how we should be embracing AI. But now it is actually, as you said, most companies have actually brought in AI tools and now they're worried about the cost and the spend. The cost of AI has accelerated and everybody's trying to figure out how do I actually manage this? How do I actually even create, find the budget for it? And for those who actually did find a budget, they saw their annual budgets getting consumed in 1 or 2 months. So, this is a huge issue right now and this combined with the fact that it's so hard to measure returns and value from it. Like most businesses are spending and they're not seeing the value back. Our strategy there, and so, I think we have happier customers and part of that is because we anticipated that AI is going to be expensive and we shifted our attention to that we have to go beyond personal productivity. Personal productivity is great. In fact, there is millions or billions of dollars of value that you get by giving people great AI tools, but it's extremely hard to measure. You cannot actually figure out if it's actually meaningfully moving business metrics and it's going to take time, too. So the shift that we made was we will go department by department and we will look at business processes that you already know how to measure, where you know that you're spending, let's say, tens of millions of dollars supporting customer queries or filling out RFPs or reviewing legal contracts. So, all these business processes across different departments where you're good at measuring the cost of those processes. So, now you actually build an agent and plan for those departmental use cases and you see an immediate impact in how fast you can actually do that work now and it translates directly to dollars. So, that's actually the key. AI is a good technology. Nobody should actually stop investment in AI, but ultimately you have to — when you're actually able to directly measure that ROI and it happens through departmental use cases, that's when you now can actually start making room for budget. For example, you can say that, well, in customer service if we are now 40% more effective in resolving cases, then we may not have to actually grow as much as our business grows. And that sort of creates budget to actually put towards AI.
Can I double click into that? So, software is dead as everybody knows. What used to be good is now bad and what used to be bad is now good. Two examples, one is, services used to be bad 10 years ago. If you were a software company and you were delivering services, people were like, what are you doing? Today, and by the way, I'm a former forward deployed engineer from Palantir, so I can say that FDEs are back again. The same is true of models. If you were a software company training your own models back in the day, like, guys, what are you doing? But now, everybody's training open-source post-training open-source models. Alex Karp said it, Satya said it. Maybe start with FDEs. What do you think of the forward deployed engineer strategy? What do you think about services to go along with Glean?
Our strategy with FDE is, you know, it's a very important — you cannot deliver some AI technology to your customers and hope that magically they'll be able to figure everything out and drive business value with it. There is without a question help that they need today. And so, if you have these teams that can take our platform and then actually help our customers build some of those high-value agents that drive business value, that actually certainly helps. So, we do have it as part of our selling process, but it's not — like it's not like, for example, how you would think about you mentioned Palantir. It's not like that. For us, that element is actually a very small part of our revenue base. And in fact, mostly we have that motion to learn like how to make our platform more effective. And we bring it to some of our most strategic customers. But from those patterns that we learn, then our strategy has been to then partner with services companies, consulting companies, and go with them to our clients where we bring the software platform and then they bring the services in a more scalable fashion. So I still believe in — it's not like software is bad or services is bad. They're two different things that require two different types of companies to be built and we still believe in that segregation personally.
Yeah. 10 minutes are up. So now I get to ask you the hard questions. You've said, Arvind, in the past that OpenAI, Anthropic, and Google are partners. We're not competitors. They're going to deliver the intelligence. We're going to deliver the enterprise context. But since then, every lab, OpenAI, Anthropic, etc., they've launched enterprise connectors, MCP, context and memory. Your biggest supplier is potentially now a competitor. How do you think this plays out?
One thing I can tell you with extreme degree of confidence is that there is no enterprise customer today that would tell you that we're happy working with just one model company. And I think that is fundamentally the structural advantage for players like us who — at Glean, we use our AI platform, we bring all of those models. Actually, there's nothing missing from our portfolio because we get to partner with OpenAI, with Anthropic, with Google, with Nvidia, with all of the providers and open source and we also get to bring open source models to our customers. And so that's fundamentally something that we can do which will be hard for a model company to do on their own. And so that puts us in a really, really good position from that perspective. And also, it's interesting that if you have to really think about AI as layers of technology in that stack. This intelligence layer, and as you mentioned, it's coming from open source increasingly more and more. In fact, we believe that in the next 2 years majority of enterprise inferencing workload is actually going to be with open source models. And so there's a lot coming from there. There's a lot coming from that.
Open source models, yeah. And then there's a lot that's actually coming from of course the foundation model companies. And so that's the layer that they get to bring to customers either directly or through us. Our layer is the one above that which is how do you make this intelligence actually work inside your enterprise? So integrating with your enterprise systems, understanding where all the data and knowledge and information is, understanding the security and governance of that information, figuring out — understanding your company and business at a deeper level. So who are your employees? What roles do they play? What work that they do on a daily basis? Figuring out the different product lines, different teams and departments. And finally understanding how today people are doing work inside your enterprise, how any given task is getting done. This is a very deep sort of knowledge and enterprise graph that you have to build and that's our specialty. That's what we focus on. That's where most of our R&D resources go. In fact, we actually put probably more number of people on that compared to any other company in the world. And so that's the value add from our side. And then on top of that you have AI experiences, and that's going to be delivered by many players as well. We have one of that experience, but Cloud, ChatGPT, they have another one. And then of course every other SaaS company in the world are also going to deliver great AI experiences within their own vertical applications. So that's the layer of the AI stack. And in that layer everybody plays a role. We play the biggest role in the middle. And that's how we're going to actually survive and thrive.
You know, I — the way I answer this question, Arvind, for our team is there's four seasons of the year. OpenAI, Anthropic, Google, and SpaceX. If you're a CIO, you're planning for a climate, the seasons will change, but you need somebody neutral. You need somebody neutral who lets you go through the seasons and prepare for a climate. And I think that's Glean.
That's us. Yes, exactly. Yeah.
