Building an AI Unicorn: Glean CEO Arvind Jain (Part 7)

Sramana Mitra: Language models have evolved a lot.

Arvind Jain: There is that vision for the future. The nature of knowledge work is going to change. We are going to have these powerful assistants that are going to take care of most of the repetitive time-consuming tasks as well as the tedious parts of search. While the technology is powerful, it’s also very hard to make it work for you. There are problems with instability in the sense that you can ask the same question four times and it comes back with a different answer each time.

Sramana Mitra: Is that less in a constrained dataset that is specific to an enterprise?

Arvind Jain: There are techniques that you deploy to reduce hallucination problems. We use language models in a very constrained manner. When you come and ask a question to Glean, we’ll use a score search technology that we built over the last five years. It assembles the pieces of knowledge that is relevant to the question you entered.

Then we take those nuggets of information and we’ll use a language model to say that for this question, use these nuggets of knowledge. We constrain it that way. It avoids that hallucination problem. Then there are techniques we deploy. Even with that, you can’t really explain how they work. It contains the output of that model. You can take the input and you can do some correlation and be more confident that the knowledge that we provided is the knowledge that is used. There are techniques that we deploy to reduce this issue.

Sramana Mitra: We are close to the end. I want to ask you one question that is different from the more technical questions. It’s about how this whole unicorn mania is shaping up. The whole venture industry has gone after unicorn ventures in a big way. We went through the adjustment. Right now, we’re still going through the corrections.

How do you view that? In your experience of building Glean, you have an enterprise product which is making substantial revenue right from your second year. You’re in good solid revenue early on. You raised a Series C while you were still making revenues. It’s a lot of capital. How important was this unicorn status to you as an entrepreneur?

Arvind Jain: When we got the unicorn status, people were trying to avoid it.

Sramana Mitra: Really?
Arvind Jain: What’s important for us is that you have access to capital which can help you invest in R&D the way you want to and scale the business. When we raised our capital, that was important for us. We are now more comfortable making R&D investments.

Sramana Mitra: The way I look at it is, there are real unicorns and there are pseudo-unicorns. Real unicorns are businesses that are rooted in strong business models and strong ability to monetize. Then there are pseudo-unicorns that are pumped up with big liquidation preferences and got a lot of capital but it’s not really validated and not anchored in a good business model.

Yours seems to be a steady enterprise software company. In that sense, it’s not something esoteric. It’s great technology. It’s a simple business model.

Arvind Jain: It is. We have a long journey ahead.

Sramana Mitra: It takes a long time to build a significant company.

Arvind Jain: Right. We have a good start. Our users are very happy using the product. All things are pointing in the right direction. We hope to keep growing from here.

Sramana Mitra: Are you going to include in your vision this enterprise personal assistant being available in a broader scale?

Arvind Jain: When I say enterprises, I was not referring to super large companies. We use it ourselves internally ever since we were 20 people. It’s useful for all sizes.

Sramana Mitra: Enterprise software strategy and pricing is different. If you want to make something like that available to a broader audience, everything has to change.

Arvind Jain: We started with mid-market companies, and then we’ve gone up market. We serve both midsized and large. We’re focused on companies with 200 employees upwards today. We don’t have plans to bring it to individuals yet.

Sramana Mitra: You can’t do everything. Great! It’s a wonderful story. Thank you for your time.

This segment is part 7 in the series : Building an AI Unicorn: Glean CEO Arvind Jain
1 2 3 4 5 6 7