The future that I see with AI is that every person who works is going to have this amazing team of assistants, co-workers, and coaches around them that is going to actually make them a lot more effective. Welcome to the Matt podcast. I'm your host, Matt Turk from Firstark. Today, my guest is Arvin Jane, the CEO of Glean, one of the hottest enterprise AI companies in the generative AI wave, widely rumored to be valued at $7 billion in a new round of financing. Early in his career, Arvin spent more than 10 years at Google, which he joined pre-IPO as an early engineer on search. He then went on to be the co-founder of Rubric, a data security startup that's now become a public company with a market capitalization of over 11 billion. In this conversation, we go deep on Glean's AI search product, and then spend plenty of time on AI agents in the enterprise, both technically and from a broader industry standpoint. Given his life's work, Arvin has a unique and very pragmatic perspective on the evolution of search and what AI can actually do within organizations. There is real value being delivered to customers, but agents are still better run in a supervised manner where a human is in charge and looking at the work of the agent. Please enjoy this great conversation with Arvin of Galen. Arvin, welcome to the Met podcast.
This feels like a very timely moment to have the conversation about Glean. It seems that you guys are firing on all cylinders. You had a well-publicized remarkable year in terms of AR acceleration. You made just a few weeks ago what feels like a really important release around agentic reasoning. And of course there are rumors which I will not ask you to confirm nor deny that there may be another big round of funding with a rumored $7 billion valuation. So thanks for making the time in the middle of all this to chat with us. We're going to unpack all things at Glean: enterprise search, RAG, agents. But before we do that, I thought it'd be a fun way to start to sort of riff a little bit on what's going on in the industry. So, it's been yet another wild week in AI. You know, over the last few days, we had the release of O3, O4 Mini, and then just a day after there was a release of Gemini 2.5 Flash. So, I'm curious as an industry participant and observer, what do you make of all of this and where do you think we are heading in terms of what the AI model world is going to look like in a few years? Are we going to have a bunch of models? Are we going to have a small number of very powerful model and model companies?
I think all indications are that there's going to be a very large number of models that are going to be available to us in the industry. It's going to be both closed models from companies like OpenAI and Anthropic and Google, but also there's going to be a lot of open domain models, models like Llama and then a lot of derivatives as well that are going to be available to us. So that is already happening. It's just going to be more and more of that in the future. What it means for us is that we will be in this world where it's going to be hard to always keep up with the model technology for all of us. So I think it'll become increasingly more a deeper infrastructure level thing that then software developers actually make sense of and figure out for a given application, for a given use case, what's the right model technology to use. And hopefully for enterprises, for customers, the LLMs become a technology that they don't really think about, it's just something in the background.
And you just mentioned open source, and I think you've said somewhere that 80% of the AI work may rely on open models. What's your sense of the reality in the enterprise today of the state of preparedness and adoption of open source AI models?
Today, for most customer use cases, the companies are using the closed models at the moment, for the ones that at least we are familiar with. As we go and work with large enterprises and we think about their core business processes that they want to automate with AI, right now the closed models are the ones that tend to dominate, and I think it's for the right reason. As you are in these early stages of journey and you're trying to build an application where you're going to actually use AI as a key component, this is not the time for you to optimize and figure out what's the best model or what's the most cost effective model, because you're building an application and nobody uses it. So it doesn't matter how much it costs. First, go and use the best, the state-of-the-art model, which today are still the closed domain models. So for that reason, you're seeing that largely people are starting with these larger models from OpenAI or Anthropic or Google. But as things reach scale, let's say you have an application that now has millions of users and hundreds of millions of daily interactions, now cost becomes a real factor, and that's when open-source models start to shine because you can actually take that technology and make it work very specifically for your application for your use case. You can distill a model which is much smaller. It becomes not only a lot faster for you, it becomes more accurate, but most importantly, it becomes much cheaper for you to run. So that's sort of how we see the transition happen: initially you start your exercise, you work with the state-of-the-art largest models, and then you start to fine-tune, distill to the right model as you start to get to scale.
Mhm. And do you think that cost is the main driver for open source adoption as you just mentioned? Is there a part of the reason why enterprises may want to use open source also something around the fact that they want to run it on their own infrastructure and possibly have a reluctance to send data out to commercial API companies? Is that something that you hear or is that something that people make a big deal of but in reality none of that is actually true?
Well, for certain businesses, there is actually real regulation and restrictions where you cannot send your data out. Having everything be running in an air-gapped environment is actually a must. So for those, you have no choice but to work with open domain models. So that's one driving factor. Another one is cost, as I mentioned. When you have a large scale application, you have to look at the needs that you have from LLMs and use very specific models for them, because they are going to be both better for user experience because they become faster for you, and they will hallucinate less for you because they are trained to do a very specific thing, so their range of errors actually also reduces, and finally they are more cost effective. But there's another reason, the third reason, why people actually want to work with open domain models is also control. You feel like you don't want to have dependency for your core business processes on other companies sometimes, and you want to run your own models. You also feel like over time you'll be able to customize them better and better if you have the expertise to do that, which you won't be able to do if you were working with a closed domain model.
All right. So more commercial models, more open source models. And it's actually fascinating because there's even more companies being launched as foundation model companies. So most notably thinking of machines and SSI. I'm again curious with your industry observer hat on what you make of all of this. You know, we were talking about those rumors, unconfirmed rumors about a $7 billion valuation. And you guys are a fantastic business, reportedly well over now 100 million ARR. Meanwhile, you have companies that show up sort of out of nowhere and raise at staggering valuations, you know, $10 billion valuation for one, $32 billion valuation for the other before they even have a product publicly in market. How does somebody like you think about that state of things?
Well, the more the number of LLM companies that are out there, the better it is for the industry. So we actually like seeing new names. We like seeing new models. And in fact, we benefit from them today in a big way. There are so many model choices that we have today compared to last year. Models that are of comparable quality, models that are all getting better at certain things. For example, today for a lot of coding use cases, people tend to use Claude a lot more. It works better for reasoning. They're using GPT for others. So you're seeing these models get specialized to specific skills. So for application companies like ours, where we're trying to use this innovation and solve real business problems, the more innovation that we see, the better it is for us. Not just from a point of view that we have more technology to use, but also it creates competition and it ensures that we can actually get the LLM advances in a more cost effective manner that we can then bring our products in a more cost-efficient manner to our customers.
Obviously the big flavor of the day in the world of AI is agents. So we're going to unpack what that means in the specific context of Glean in a minute. But to start at a sort of high industry level, I'm curious about your view on the sort of state of agents in 2025. And I'd love to start with an actual definition of what an agent is just to ground the conversation, because there's so many people talking about agents in so many different ways. So, what is your definition of an agent?
Well, I don't know if the world is looking for one more definition, but you had a great one on Twitter, which is what I'm riffing off of. I think for me, an agent is simply an application. It's a unit of work that is being done. It used to be done by a human before, now it's being done with the help of AI. So typically, an agent is going to work on some piece of information, some data, some knowledge, which can be brought from the public internet or the web, or it could be brought from within your company's systems. The agent works on this data, it has a task to perform, it uses LLMs to reason and perform those tasks. There may be steps in this agent to self-reflect so that it can actually do the task better, and ultimately it's going to produce some work which is going to get saved in your internal enterprise systems. So an agent is nothing more than an application, like we're all used to what software applications are. It's just that AI is being used in a fundamental way inside to do a lot of this work. So that's how we think of agents. They come in a wide shape or forms, from a simple workflow to a very deep complex application. People are choosing to use the word agent to describe all of them.
What's the why now of agents from your perspective? Where do they suddenly become such an important topic, and also your sense of reality of what can be done today versus hype of what people think they might be able to do in the near term?
So ultimately, an agent is automating a business process which was before managed by a human. I'm talking in the context of agents in businesses. Of course there are agents that you can use in your personal life as well, but for a moment let's stick to business agents. So these are agents that are taking a specific business process and transforming them and automating them using AI. Take some examples: in your legal team, when you get third party contracts, you have to redline them, make sure they conform to your company's guidelines. Today you have somebody in the legal team, a lawyer, who's actually reading these long documents and they're going to redline it and pick all the places in the document where you need to change it to conform to your company's policies. This process requires a human to go through a long document, they read it, they know the standard to which they are trying to align this document. And you can now replace that whole process with an agent that automatically redlines your third party contracts. The way you build this agent is to tell it exactly the process that the human was following before: read the document, compare it with your version of the document, and then highlight or redline the problem areas in the document. And AI is pretty good at reading instructions, reading documents, and is able to do this automatically for you. You can imagine that for many business processes today, like customer service, one of the key activities is customer service agents resolve tickets. To resolve tickets, they are reading knowledge bases inside your company looking for where the answers to these questions could be, and then use that to respond back to the customer. Again, AI can come and do that work. It can search over all of that information, understand the question, and answer that question automatically for the end user. So you can start to imagine that all these business processes that you have, you can start to automate them with AI. This has been a desire of the industry for a long time. Before the AI wave, you had robotic process automation. It was a big thing. Now you can think of AI agents as just a much more advanced version of RPA. And there's a lot of business value from it. You can actually double the efficiency of every person in your company with it. And that's why there's so much excitement. The thing that's happening is that when you think about AI in this specific way, that you're going to build agents and you're going to automate your business processes, it's super appealing because you can see direct ROI impact from building these agents, which is causing all the excitement in the industry today.
And do you think that we are there or very close from a technology perspective, or do you think there's something missing for the full realization of the vision, whether that's open industry protocols, MCP style stuff? What do you think?
We are in very very early stages of this journey. I mean, if you think about agents theoretically, an AI agent is promising you to do any piece of work. In fact, that's how we describe our product Glean to our customers. We say you can come to Glean and you can ask any question or you can give it any task, and Glean will use all of the world's knowledge as well as all of your internal company's data and knowledge to answer those questions or complete those tasks for you. Now when I say that, I'm making this assertion that I can do everything in the world, and that's not the reality. Today agents are still quite basic, I would say. They all still need a significant amount of supervision. You need to, so I would think of agents as more being like, I have a piece of task, I still own it as a human, but an agent can potentially do it for me, but it'll do its work and I'm still going to go and review it. We barely see any application where customers are running agents in a fully unattended, unsupervised setting. So we are early, but the impact is actually very clear. The agents, but remember, even if human supervision is required, that doesn't make the agent worthless. That same example I was giving you before about a legal person reviewing a 100-page contract and redlining it was a two-week activity. If you actually get an AI agent to redline it for you, it's going to do it in one minute. And it won't be perfect, but now I can go in and look at the work that AI did and potentially spend 2 hours instead of a week to actually redline that, and that's more than 90% savings for me. So that's where things are today: there is real value being delivered to customers, but agents are still better run in a supervised manner where a human is in charge and looking at the work of the agent.
All right. I'd love now to go into Glean itself very specifically, and perhaps rewind back to the beginning which I believe was in 2019. So you started as an enterprise search company for the first few years of the company, which I find to be a fascinating idea because there was this whole generation of companies in the 2000s and perhaps all the way to the early 2010s. I'm thinking of Fast, Transfer that was out of nowhere and acquired by Microsoft for like a billion, something there was bought by Oracle for a billion, famously Autonomy which was acquired by HP for 10 billion and then HP wrote off 8 billion a few months later, and Verity, that whole generation. So it sort of feels like it was an industry that came and then went away in the early 2010s. So what was the vision and the insight then that the industry was ready for reinvention of it when you started?
I would say it was not so much a vision or a realization that this was the right time to solve a problem. It was just frustration, personal frustration that I had, which forced me to start this company. Finding information at work has become increasingly difficult over the years because number one, there's so much knowledge that we have in our companies these days, we live in a data-driven world. Number two, that information has increasingly gotten fragmented as we went through the SaaS revolution. Every business has ended up with hundreds of these cloud-based applications because buying applications is so simple now. So businesses will end up with hundreds or sometimes even thousand plus applications, and all of your company's data and knowledge is spread across all of those systems. And my company before we started Glean was one of those companies. We built in the modern SaaS world with tons and tons of information and data with 300 different applications, and nobody in our company could ever find anything. That company famously was Rubric, now a wonderful public company. So that was a personal pain that I had. And I did feel that this problem had to be solved because it was not unique to us. Every person who I talked to would all say the same thing: it's hard to find information, it's hard to get the answer that you need to do your work. And even before Rubric, I was used to at Google, and ironically we were making it easy for everybody in the world to get answers to their questions, but we were not helping ourselves internally. Inside Google also it was really really hard to find information. But in 2019, there were a few key technical trends which actually made this problem more tractable, or actually more acute. The same thing I talked about how SaaS fragmented your information so much making it hard to find, but SaaS also allowed you to tap into enterprise data and knowledge more easily because we have open APIs. So if you're building a search engine, it's much easier to build it now because you can tap into these standard APIs to these products and actually get that enterprise content all in one place and then make it searchable. But the second big trend that we thought was super powerful was transformers. It's 2019 and nobody as such is really talking about transformers and language models. This was still largely within the domain of search, like Google people were using these models to try to advance Google search. And our team, me included, most of us actually came from Google, and we were seeing this big impact that the transformer technology was having on search. Suddenly you could move from that keyword-based word matching technology to find the right information to conceptually understanding questions and knowledge and doing the matching at that semantic level. So we saw a big opportunity and we decided to use that to build a really good enterprise search. That's the origins of how Glean got started.
So that was phase one. And then there was a phase two I believe around the ChatGPT moment when you went from AI-powered search to something that's more like RAG. Is that fair?
That's right. So these models kept getting better over the years. I think for us, using transformers in 2019 made us one of the first companies to bring LLMs to the enterprise. But we didn't necessarily have the foresight. I should correct myself. We didn't have the foresight and we didn't know how fast these models would get better. In 2019 they were good at understanding content but not good at generating, not good at reasoning. And these capabilities started to come over the years. So by 2022 with GPT-3, GPT-3.5, we were seeing this amazing ability for these models to actually generate their own answers to questions that people have. And we felt that that allowed us to really advance our product in a significant way. So what used to be, we started out with being the Google for your work, for your workplace, to now we could actually become the ChatGPT for your work environment. When people come and ask questions in Glean, instead of just surfacing the right information back to them which we think is most relevant, we could actually make AI read that information and generate its own precise answers so that you could save our users even more time. So we saw that opportunity naturally as the models got better and transitioned and launched our Glean Assistant, the product that you can think of as a more powerful version of ChatGPT inside your company. And the timing was perfect, right? Because you had done all the piping before of integrating the data sources, building the connectors to...
