While the internet argued about which chatbot writes the best poem, a company called Glean quietly built one of the fastest-growing enterprise software businesses in history — by making an AI that reads everything inside a company and answers any question about it. On June 9, 2026, Glean raised $150 million at a $7.2 billion valuation, and the numbers underneath that headline are the real story: $300 million in annual recurring revenue, tripled in 15 months, and an AI agent system now performing 100 million actions a year on its way to a billion.
I want to walk you through this one because it's the clearest example yet of where the real enterprise-AI money is going — and it's not where most people are looking. It's not the flashy consumer chatbot. It's the boring, invisible layer that knows where everything in your company lives and does the looking-up for you. Let me show you how it works, why it's worth $7.2 billion, and what it means whether you work inside a big company, run one, or invest.
What Glean actually is (and why it's not a chatbot)
Here's the problem Glean solves, and you've felt it. In any company bigger than a few dozen people, the information you need is scattered across a dozen apps — Slack messages, Google Docs, emails, Jira tickets, Confluence pages, Salesforce records, that one spreadsheet someone made in 2024. Finding anything is a treasure hunt. Studies have long pegged knowledge workers as spending around a fifth of their week just searching for information.
Glean connects to all of those systems and builds what it calls a permissions-aware knowledge graph — a living map of everything your company knows, who created it, how it connects, and crucially, who's allowed to see it. Then it lets any employee ask a plain-English question — "what's our refund policy for enterprise customers?" or "who's working on the Henderson account?" — and get an answer grounded in the company's actual documents, not the open internet.
That last part is the moat. A generic chatbot like the ones I compared in ChatGPT vs Claude vs Gemini knows the whole internet but nothing about your company. Glean knows your company cold but respects every permission boundary — so it won't show an intern the executive comp spreadsheet. That combination of deep internal knowledge plus airtight access control is exactly what makes it sellable to a Fortune 500, and exactly what a raw model can't do out of the box.
The numbers that justify $7.2 billion
Enterprise software valuations live and die on a few metrics, and Glean's are the kind that make investors trip over themselves. Look at the trajectory:
| Metric | Value |
|---|---|
| Valuation (June 2026) | $7.2 billion |
| Latest raise | $150M Series F |
| ARR | $300M (up 3x in 15 months) |
| ARR path | $100M early 2025 → $200M Dec 2025 → $300M May 2026 |
| AI agent actions/year | 100 million → targeting 1 billion by year-end |
| Fortune 500 customers | Nearly doubled year-over-year |
| Customers deployed in 5+ departments | 85% |
That 85% number is the one I'd underline for you. It means once Glean gets into a company, it doesn't stay in one corner — it spreads across the whole org. That's the holy grail of enterprise software: land in one team, expand everywhere, and watch each customer's spend climb year after year. It's the same "net revenue retention" engine that powered Anthropic's run to a $965B valuation and Cursor's leap to $29.3B. When existing customers keep paying more without you spending a dime to acquire them, revenue compounds — and compounding is what a 20x-plus revenue multiple is really pricing in.
The shift from "search" to "agents" is the whole game
Glean started as enterprise search — a better way to find things. What makes it a $7.2 billion company instead of a $1 billion one is the pivot to agents: AI that doesn't just find the answer but does the work.
That's what those 100 million agent actions are. Instead of just telling you the refund policy, a Glean agent can draft the refund email, update the ticket, and notify the account manager — across all those connected systems. The company is targeting a billion such actions by year-end, a 10x jump. This is the same leap I described in the Claude Fable 5 launch: the frontier of value moved from "AI that answers" to "AI that acts." Glean's advantage is that its knowledge graph gives those agents something most agents lack — full, permission-safe context about the specific company they're working inside.
It's also the productive, non-scary version of the AI-at-work story. Glean's agents don't replace the employee; they hand the employee back the fifth of the week they used to lose hunting for information. That's the amplification pattern that actually generates returns, the exact distinction at the heart of the Gartner study on why AI layoffs don't pay off but AI augmentation does.
The money math for a company that buys it
Let me put real numbers on why CFOs sign these contracts, because that's what turns a cool product into $300M ARR.
Say a knowledge worker costs a company $120,000 a year fully loaded, and they lose ~20% of their week to searching for information and re-creating things that already exist. That's roughly $24,000 of wasted time per employee, per year. Glean enterprise pricing runs in the ballpark of a few hundred dollars per user per year. Even if the tool only recovers half of that lost time, you're turning a few-hundred-dollar cost into ~$12,000 of recovered productivity per person — a 20x-plus return.
| The Glean ROI question | Per knowledge worker |
| Fully-loaded cost | ~$120,000/year |
| Time lost to information hunting (~20%) | ~$24,000/year |
| Glean cost | a few hundred $/year |
| Recovered value (even at 50%) | ~$12,000/year |
| Rough ROI | 20x+ |
That's the same brutally simple math driving every real enterprise-AI win right now — the $37M-a-year documentation savings I broke down in healthcare and the cost collapse across document processing. Find the expensive, invisible time-drain, automate it, count the recovered dollars. Glean just did it for the single most universal office task there is: finding stuff.
Why this is harder to copy than it looks
You might wonder why Microsoft — which owns Office, Teams, and Copilot — hasn't just crushed Glean. It's trying. But Glean has two things that are genuinely hard to replicate.
First, it's neutral. Glean connects to everything — Google Workspace, Microsoft, Slack, Salesforce, Atlassian — without favoring its own ecosystem. Microsoft's Copilot is brilliant inside Microsoft's world and awkward outside it. Most real companies run a messy mix of tools, and they want an AI that sees all of it, not one that quietly nudges them deeper into one vendor.
Second, the permissions-aware knowledge graph is years of unglamorous engineering. Getting an AI to know everything and never leak what someone shouldn't see is brutally hard, and it's exactly the kind of trust infrastructure that risk-averse enterprises demand. That's the same defensibility lesson from how the most valuable AI companies are built — the moat isn't the model everyone can rent, it's the proprietary data layer and the trust around it.
What this means for you
Depending on where you sit, here's the takeaway.
If you work inside a company, this is the AI that's actually coming to your job first — not to replace you, but to sit beside you and end the daily treasure hunt for information. Lean into it. The people who become fluent at directing these enterprise agents are the ones who get more valuable, the trend I track in the highest-paying AI jobs of 2026. The worker who masters the company's AI brain outperforms the one who keeps searching by hand.
If you run or advise a business, Glean is the template for getting AI ROI that actually shows up in the numbers: don't buy "AI transformation," buy a tool that kills one specific, expensive, universal time-drain and measure the recovered hours. And if you're building your own company, the lesson scales down — the same connect-everything, know-everything approach is what lets one person run an operation that used to need a team.
If you invest, Glean is a clean read on where enterprise AI value is concentrating: not in the models themselves, which are commoditizing, but in the application and data layer sitting on top of them. Just remember the catch from the AI wealth divide — at $7.2 billion and private, the easy gains here are being captured in rounds you can't buy into. The realistic public play is watching which of these enterprise-AI champions, like the ones in the $3 trillion IPO pipeline, eventually reach the market.
The honest take
Glean is the quiet proof of where enterprise AI is actually paying off. While the headlines chase consumer chatbots and trillion-dollar model labs, the company tripling its revenue is the one that solved a boring, universal, expensive problem — nobody can find anything at work — and wrapped it in trust infrastructure that big companies will actually buy. $300 million in recurring revenue isn't hype. It's thousands of companies deciding the recovered time is worth far more than the price.
What I'd take from it is a pattern you can use anywhere: the biggest AI money isn't in the most impressive demo, it's in the most expensive boring problem. Finding information, processing documents, writing clinical notes, searching a codebase — these are the unglamorous time-drains that, once automated, print real returns. Glean found one of the biggest and built a $7.2 billion company on it.
So here's the question worth carrying into your own work: what's the boring, expensive, everybody-does-it task in your world that an AI could quietly eat — and who's going to build the Glean for it before you do?
