Arvind Jain (Glean): Solving enterprise search with AI

Arvind Jain spent a decade scaling Google Search and Maps before co-founding Rubrik, a cloud data security company that went public in 2024. As Rubrik scaled, he noticed how much time employees were losing searching for the information they needed to accomplish tasks. That’s when the idea for Glean was born.

As co-founder and CEO of Glean, Arvind is building the Work AI platform that unlocks knowledge already sitting inside a company. Glean connects to every system – from Slack and Jira to Salesforce and Google Drive – to help employees find the information they need and automate everyday tasks.

In this conversation with Pigment co-CEO Eléonore Crespo, Arvind explains how enterprise search was always a neglected problem rather than a hard one, what the superintelligent enterprise actually looks like on a Monday morning, and why trust is the only moat that holds.

Where enterprise search needed to evolve

When Arvind started Glean in 2019, the company’s focus was straightforward: help people find information their company had but couldn’t surface readily.

On the public web, finding information is relatively simple. Websites are accessible to anyone with a browser, and search engines rank results using visible signals like traffic, links, and user behavior. Inside a large business, none of that applies. Enterprise knowledge is scattered across hundreds, sometimes thousands, of systems. It accumulates over decades and rarely gets cleaned up. Some of it is current and useful. Much of it is outdated. Almost none of it is organized in a way that travels well across teams, tools, and time.

The products that existed weren't built for this reality. They transplanted the internet search model into a corporate environment without adapting it for how businesses actually work — no understanding of organizational context, no way to evaluate the authority of internal sources, no ability to tell fresh information from stale.

Glean approached it differently, treating enterprise search as a systems problem that required new signals for relevance, freshness, and authority.

What actually makes AI work

The core of Glean, Arvind explains, is understanding each business it's serving – not at a surface level, but deeply enough to know what products a company sells, who its customers are, and how work actually moves through the organization.

The way Glean builds that understanding is by noticing which documents are getting used, which questions employees are answering, and which tasks are getting completed. Those observations accumulate into an ever-updating model of the business: one that learns what's current, what's authoritative, and what's changed since the last time someone wrote it down.

The judgment it applies mirrors what a thoughtful person would do naturally. Faced with 10 documents on the same topic, a good researcher would weigh recency, consider who wrote each one, and cross-reference against what's actually in the code or the latest meeting notes.

Moving from search to agents was a natural evolution

In the early days, Glean was among the first companies to bring transformer-based models into enterprise search – pre-training on a company's entire corpus so the system could understand information semantically. And Glean pioneered the foundational infrastructure for enterprise AI, like vector search and embeddings, before they were popularized.

As the underlying models grew more capable, the product evolved naturally. First, Glean moved from surfacing links to generating answers. Then, it moved from generating answers to creating documents, running analyses, and completing work. Whatever journey AI was taking in the consumer world, Glean embarked on in parallel inside the enterprise.

In 2024, Glean launched AI apps – scoped, conversational experiences that let customers focus the platform on specific use cases. That meant an IT team could configure Glean to automatically resolve employee tickets using only its approved knowledge base. A sales team could build an assistant that knew every deal in the pipeline. These apps were, in practice, agents before the industry had agreed on the term.

Today, Glean operates as the context layer underneath the enterprise's entire AI stack. It's a Model Context Protocol (MCP) endpoint that other agent frameworks, from Copilot Studio to Amazon Bedrock, can plug into. Customers can build agents inside Glean's own platform or use Glean as the connective tissue for agents they build elsewhere.

Glean plans in monthly cycles now, down from quarterly. Engineers are rewarded for replacing technology with something better. Speed is treated as a survival mechanism, not a virtue. Arvind is direct about this inside the company: the teams that move fastest are the ones that stay relevant.

What companies should do today to scale AI

Most enterprises have accepted that AI matters. The gap, Arvind argues, is more about structure than it is about conviction.

The companies seeing the most impact have dedicated ownership – leaders whose only job is AI transformation, rather than executives for whom AI is one more initiative stacked atop an already full plate. A chief AI officer, or an equivalent center of excellence, consistently makes a measurable difference. AI-first is a good mindset, but Arvind believes that leaders are already fully subscribed. Dedicated ownership is what actually moves the needle.

Beyond structure, he points to one tactical approach that has worked consistently: pairing experienced leaders with recent graduates as AI chiefs of staff. People who just joined and have never done things the traditional way have no habits to unlearn. They reach for AI instinctively, and that instinct spreads.

On measurement, his advice is to anchor AI progress to business metrics that already exist, such as case resolution time, tickets resolved per day, or time to first response. Those are the numbers a CFO will believe. AI either proves itself against them, or it doesn't.

Setting a 50% efficiency target tends to paralyze teams. Glean tried it themselves and made little progress. Small wins, celebrated visibly, build the momentum that eventually gets you there.

What the superintelligent enterprise looks like on a Monday morning

Right now, AI is a tool for the few. Most employees are too busy to seek it out, and the habit of reaching for it has yet to form across most organizations. Arvind believes that will change this year. AI will stop waiting to be asked.

He offers a Monday morning scenario. On your commute, your AI personal coworker is already aware of your calendar, your open tasks, and the documents you need to review. It briefs you for the day in a voice interface, timed to fit the exact length of your drive. You arrive at the office prepared for every meeting.

Later, that same AI scans your calendar, your task queue, and your communications. It figures out who is waiting on you, assembles a prioritized list of what needs to get done this week, and quietly completes the 30% of those tasks it already knows how to handle.

For executives, the implications run deeper. Strategy shifts rarely cascade cleanly through large organizations. For example, projects that should be deprioritized keep running, and strong contributors who are quieter go unnoticed. The superintelligent enterprise is one where AI surfaces all of this, cataloging projects, flagging misalignment, and identifying the people doing exceptional work regardless of how loudly they advocate for themselves.

This is Glean's north star. It’s a coworker that comes to you with your needs and next steps already in mind.

How Glean hires for the AI era

The fundamentals of hiring have stayed the same at Glean. Arvind still looks for people who are passionate about what the company is building, motivated to grow, and strong collaborators. What has changed is how those qualities get evaluated.

Glean now designs interview assignments that require AI to complete in the allotted time. An engineering task that would have taken a candidate two hours has been replaced with one that would normally take a month, but can be done in two hours with the right tools. Candidates are not told to use AI. The question is whether they reach for it on their own.

The same principle applies across functions. Finance candidates get case studies that genuinely require modern AI tools. What Glean is testing for is a learning mindset – the disposition to stay curious, adapt quickly, and build new habits without being asked.

Why trust is the only moat that holds

When you’re building a company, the standard playbook has a lot to say about competitive moats. Arvind has largely set it aside. 

In enterprise software, the things that look like advantages tend to erode. Network effects rarely materialize the way they do in consumer products. And in a market that moves this fast, even hard-won technology can become a liability.

This is why Glean bets on trust instead. When users change jobs, they bring Glean with them. When they have a good experience, they advocate for it. That kind of loyalty has to be earned, and it compounds over time in ways that technology advantages simply do not.

In a market where every technical breakthrough becomes obsolete quickly, that turns out to be the most defensible position of all.

Conclusion

The throughline in everything Arvind describes is a willingness to do the unglamorous work – connecting every system, building accurate context, establishing baseline metrics before claiming AI is making a difference, and celebrating small wins before chasing larger ones.

Enterprise AI succeeds when organizations treat it as an operational discipline rather than a technology bet, and the companies moving fastest are the ones that have already internalized this.

The superintelligent enterprise is closer than it sounds. But it’s built on foundations, not breakthroughs.