Competing at the speed of AI

In the late 2010s, Jain felt a gnawing frustration. As a co-founder of Rubrik, the cloud and data security company, he should have been elated by its rocket ship growth. Instead, he pondered how the company had nearly tripled its headcount in a single year, yet barely moved the productivity needle.

Digging deeper, Jain uncovered a startling inefficiency that affected Rubrik and other large organizations: Employees spend up to one-third of their time simply locating information, according to various studies. At Rubrik, critical data was scattered across 300 applications. At larger companies, it can be dispersed among 1,000 or more systems. Jain went looking for a product that could solve this problem, and finding none, he resolved to build it himself.

For years, venture capitalists had heard pitches from founders promising to build a “Google for the enterprise.” Yet one startup after another either failed altogether or struggled to generate sufficient excitement to justify much investment. Even Google had little to show for its own efforts to penetrate the business market; its specialty was in web search, not in finding information buried in the varied and numerous proprietary data silos, each with its own set of access restrictions, inside businesses. 

So when Jain began shopping his idea around, the reception he got wasn’t exactly enthusiastic. “Nobody was excited about the problem I wanted to solve,” he says. “Enterprise search was viewed as this dead space, a graveyard where no company had ever succeeded.” 

But Jain was convinced that the ubiquity of SaaS applications, which have evolved to contain the vast majority of enterprise data, had created a new opportunity for enterprise search. And eventually he did find investors who were willing to bet on him, in part because of his pitch, but also because of his pedigree as a co-founder of Rubrik. He quickly assembled a founding team that included Vishwanath, an engineer who had spent a decade at Facebook and serves as Glean’s CTO, and Gentilcore, a UX engineer who had impressed Jain during their time together at Google.

From the outset, the three recognized that AI would be key to providing a compelling enterprise search experience. Drawing on his experience as a distinguished engineer at Google, where he focused on the company’s core search technologies, Jain understood the power of AI. And as other Google search veterans joined the company, AI became part of Glean’s core service. 

“We were an AI company taking advantage of all these deep learning advancements,” Vishwanath says. “We just weren’t a GenAI company.” Yet.

In the late 2010s, however, AI was hardly the selling point it would become by the start of 2023. “There was a deliberate decision not to use AI a lot in our marketing materials,” says Paloma Ochi, Glean’s head of product marketing since 2021. They may have been creating language models trained on an enterprise’s corpus, but pre-ChatGPT, most potential customers would not have heard about LLMs. “The general feeling inside the company was that it would come off as ‘AI washing,’” Ochi says. So the company went for a far simpler message that helped it gain unicorn status quickly: Glean was Google search for the workplace. “That was our initial North Star,” Gentilcore says.

With the arrival of ChatGPT — and with it, the generative AI era — Glean’s North Star changed. The Google comparison moved to Glean’s rearview mirror. “All of a sudden, we wanted to be as useful as ChatGPT is for the open web, but for all your company's information,” Gentilcore says.

The executive team understood that to get there, speed of execution would be paramount. Glean had a big lead on potential competitors, but the velocity at which the AI sector moved could wipe that out if they didn’t press their advantage. “I've been building products for over 20 years. I’ve never seen things change this fast before,” Gentilcore says.

The first of several war rooms the company would establish focused on an assistant. They assembled what Gentilcore called a “tiger team” that brought together engineers, product managers, and designers from around the company. They borrowed the concept of a “code yellow” used at Google and other tech companies. If someone on a code yellow team taps you on the shoulder, you drop what you’re doing and help out.

“The main thing was eliminating all distractions for this core group and giving them a clear set of exit criteria,” Gentilcore says. Deliberately, they made sure this team of approximately 15 included people in both its Palo Alto and Bengaluru offices. “That way it would be like an around-the-clock hackathon,” Vishwanath says. They had a prototype within six weeks, and a product that Jain describes as “very robust” and ready to launch by June 2023. They called it Glean Chat, but later rebranded it Glean Assistant.

The customer dynamic changed immediately. During Glean’s early years, the company had to educate customers on what a Google-like search could do for their employees. After ChatGPT burst on the scene, CIOs immediately grasped the enormous potential of pairing a chatbot with LLMs trained on a corporation’s data to deliver the proverbial needle in the haystack.

“We went from a situation where we were creating a market to suddenly being in a very high-demand market,” Jain says. But that enormous opportunity also presented challenges for a sales and go-to-market team that had been built for a very different environment. “For me as CEO, that was a moment where I had to say, ‘we’re a well-capitalized company, and this is the time to bring a lot more people here,” Jain says. The company scrapped its hiring plans and doubled its workforce targets in 2023. In one year’s time, the company tripled the size of its sales team.

The willingness to scrap well-laid plans is critical for any startup that wants to succeed in an environment that’s changing constantly, says Larry Downes, co-author of Pivot to the Future and former associate dean at UC Berkeley School of Information. “A startup can do their five-year and 10-year strategic planning,” Downes says. “But with the speed at which things are changing in AI, it’s useless.” 

Glean’s aggressive moves paid off. The Glean Assistant was precisely what enterprises were seeking in the post-ChatGPT era — and the company had an expanding sales team able to take advantage of demand. Glean’s revenues tripled in a year.

“The use case of Glean changed once they layered more generative AI into the product,” says Andrew Zloto, a partner at SoftBank Investment Advisers and an investor in Glean. “The use case went from being a search tool to being an answer tool, which is meaningfully different for a business.”

As it barreled forward at full speed, Glean had plenty of obstacles to navigate. “There was a lot of stress in those first months,” Jain says. Employees were stoked to be on the right side of the AI wave. “That really energized people,” he adds. But they also sometimes felt overwhelmed by the frenzied pace of change — and the pressure from new competitors creating answer engines for enterprise customers that could chip away at Glean’s value proposition. 

“All of a sudden, there’s a wave of these three- and five-person startups,” Gentilcore says. “Along with them, OpenAI, Google, and other giants were saying, ‘Hey, we’re building that into what we’re doing.’”

Engineering felt the pressure most acutely. “Our team felt the frustration of trying to operate in a very unstable environment,” Jain says. “We would build something and the next week it was already obsolete.” It was like constructing a house while someone was constantly reshaping the foundation beneath them.

Vishwanath credits Jain with having shown a steady hand through a turbulent time. Jain was a repeat founder but a first-time CEO who excelled at two things, Vishwanath says. The first was to continually stress Glean’s strengths, including a four-year head start on the competition. “His message was how we’re in this great position to exploit this opportunity given our really strong underlying product,” he says. Jain also focused on removing roadblocks to create the most efficient environment possible for the people he had hired, as he eliminated distractions like unnecessary meetings. 

Glean made other small changes to help people cope. Engineers had always been happy to help sales reps by jumping on a video call to go over some of the technical details for a potential customer or help an existing one troubleshoot a problem. But with so many incoming requests, engineering adopted a rotating on-call system so that no single team member became overwhelmed with requests. Engineering also learned to be flexible with anything they built given the shape-shifting environment of the past two years, where LLMs were constantly leapfrogging each other. “We had to build a product where there’d be nothing preventing us from using the latest and greatest model,” Vishwanath says.

Glean was also operating in an environment where the SaaS applications it was tapping into for enterprise data were building their own AI capabilities. Rather than try to compete with them directly, Glean packaged some of its AI capabilities into a platform that others could use to improve their own products. “Glean works very well with all the existing applications that you have inside your company,” Jain says. “That’s been our strategy from day one.” 

Because of its potential to improve the work experience across applications, Glean is now pitching itself as more than an enterprise search company. “Glean is a work AI company,” Jain says. “We are here to transform work for every employee in every company across the world.”