Glean Teardown — Arvind Jain's $2.2B Enterprise AI Search Platform

TL;DR + Quick Facts

Glean is what happens when one of the engineers who built Google's search infrastructure decides the worst search experience on Earth is the one you use eight hours a day at work. The product is deceptively simple to describe and brutally hard to build: one search box that searches across every SaaS tool a knowledge worker touches in a given week. Slack threads, Google Drive docs, Notion pages, Jira tickets, Confluence wikis, GitHub PRs, Salesforce records, Zendesk tickets, ServiceNow incidents, Workday HR documents.

Quick facts. Founded 2019 by Arvind Jain (CEO, ex-Google distinguished engineer, Rubrik co-founder), T.R. Vishwanath (ex-Facebook), Piyush Prahladka (ex-Google), and Tony Gentilcore (ex-Google). Total raised through 2024 is north of $260 million, with the most recent round at a $2.2 billion post-money valuation led by Kleiner Perkins, Lightspeed, Sequoia, General Catalyst, and Citi Ventures. Revenue is not publicly disclosed but external estimates put it at $50 million ARR by mid-2024. Roughly 350 employees, headquartered in Palo Alto.

Pricing is per-seat. Public-ish anchors put it at $30 to $50 per user per month at typical enterprise volume. A 2,000-seat customer is doing somewhere between $720K and $1.2M ARR.

The thing to internalize: Glean is not a wrapper. It is a search infrastructure company that happens to ship an LLM front-end. The defensibility lives in the indexing layer, the permissions model, and the five-year head start on integration depth.

The Product — Federated Search, Permissions-Aware, AI Synthesis

Most people describe Glean as "ChatGPT for your company's internal data." That description is wrong.

Glean runs an indexing service that connects to every major SaaS tool via the tool's own API and OAuth scopes. For each connector — and there are now more than 100 of them — Glean does a full historical crawl on initial setup, then maintains a live delta index via webhooks and incremental polling.

Three engineering problems make this hard:

The first is permissions. Inside a 5,000-person company, the same Google Drive folder might be visible to 12 people, the same Notion page to 200, the same Slack channel to a single team of 8. Glean's index is permissions-aware at query time. When you search, Glean intersects the result set with your permissions snapshot in every source system simultaneously.

The second is freshne