Glean

Glean, third for private equity portfolios

Every company a fund acquires arrives with years of documents, tickets, wikis and email threads, and much of that material is hard for new leadership to find. Glean is built for exactly that problem. It began life as a search product for companies and has grown into an AI platform resting on what it calls company context¹. The browser tab on its home page reads, in effect, as a promise that its enterprise AI will work for the buyer².

In this ranking Glean sits third. Its permission model is the most explicit of the seven, its connector library is large, and its certification list is long. It sits behind Nexus and Hebbia for a private equity buyer for two reasons drawn from its own pages: its center of gravity remains knowledge and search, with several agent governance controls still marked coming soon or beta, and it publishes neither prices nor a standard implementation timeline.

This profile draws only on Glean's public website and developer documentation as read in early November 2026. Each fact carries its line number in the fact record.

A company of scale

For a fund weighing vendor risk, Glean's size is part of the story. Its about page charts a fast climb: search was live at 40 plus companies when Glean launched publicly³; less than twelve months on, investors put in $100M at a valuation of $1B⁴; and by the time Agents shipped, Glean was valued at $7.2B⁵. Revenue on an annual recurring basis now stands at $300M, per the page⁶, up threefold from $100M across 15 months⁷. Arvind Jain founded Glean and serves as its chief executive⁸.

A vendor of that size brings a different kind of comfort from a seed stage company. It also brings a standard product that a portfolio company adopts largely on the vendor's terms.

What the product line covers

The platform page rests its argument on context, calling it the base that allows AI to do actual work⁹. The home page names three parts of the line. Glean Enterprise Context is one¹⁰. Glean Assistant is sold as a coworker powered by AI¹¹. Glean Search promises fast answers from what it terms next generation enterprise search¹²; its page claims a single place to search over 100 tools, with real time permission awareness¹³.

Underneath, the platform page explains a stack in which connectors gather signals from across the business, search indexes the data, and an Enterprise Graph maps the relationships among them¹⁴. Glean Intelligence then routes every task to whichever of its 40 plus open and frontier models fits best¹⁵. Search tunes itself to each company's vocabulary over time without manual work¹⁶ and relies on deep learning language models to interpret plain questions¹⁷.

For a portfolio, that architecture suggests one clear benefit: on Glean's description, a newly acquired company's own vocabulary is learned by the system as it is used¹⁶.

Building on Glean

Developers have several entry points. A portfolio company indexes its data once, then reuses that index through Glean's open APIs, its UI ready web SDK or headless MCP¹⁸. The platform page lists a Search API¹⁹, a Chat API²⁰ and an Agents API²¹. In the developer docs, four platform APIs carry general availability, Skills, Agents, Chat and Search²²; Admin, Departments, Users, Usage and Triggers are labeled experimental²³. A portfolio company's engineers can code against Glean from Python, TypeScript, Go or Java²⁴, and can also point tools at a developer MCP server that Glean hosts²⁵.

A portfolio company with a small engineering team can therefore use Glean as a data layer for its own tools. The experimental status of the Admin and Users APIs is worth noting for any team that plans to automate user management across several companies.

Connectors and reach

The connector count is the headline. The home page claims upward of 275 connectors²⁶, a number the connectors page repeats for its out of the box library spanning enterprise apps and ecosystems²⁷. The documentation hub for connectors lists 107²⁸, a gap a buyer should ask about, since the two counts may measure different things.

Coverage extends beyond the list. Beyond the library, a company can push its own data in with the Indexing SDK²⁹, write actions from an OpenAPI spec³⁰, or plug in MCP enabled tools for extra data and actions³¹. Teams whose app is missing can ask Glean for a new connector³², though the connectors page makes clear that an unlisted app needs either that request or a custom build³³. The documentation describes BigQuery analysis that runs under each user's own access³⁴, per user connections for Jira, Gmail and similar accounts³⁵, and real time search in Slack through dedicated APIs³⁶.

For a fund, the practical test is the same as for any vendor here: take the actual systems list of a portfolio company and confirm each connector by name before signing.

Permissions, the core strength

Glean's permission handling is where it leads the field on this list. On the security page, Glean commits to applying the permission rules of the originating system to each read operation and each write operation³⁷. The connectors page confirms that Glean inherits those permissions³⁸ and that permission changes in a source app flow into results immediately³⁹. Search results, its search page adds, are filtered by the permissions each data source already sets⁴⁰.

Control over agents is layered on top. A customer can require approval before an agent action runs⁴¹. According to the about page, a customer sets both the reach of AI and the safeguards its organization requires⁴². Agents can run under service credentials that admins scope and manage⁴³, and the docs include a skill scanning feature⁴⁴.

For a firm that may put a single deployment across a holding company and its subsidiaries, inheritance of source permissions is the feature that keeps one unit's files away from another's staff.

Security and certifications

The certification list is long. For a fund's vendor questionnaire, the security page repeats six certifications first listed on the home page⁴⁵: GDPR⁴⁶, ISO 27001⁴⁷, SOC 2 Type II⁴⁸, TX-RAMP at Level 2⁴⁹, HIPAA⁵⁰ and ISO 42001⁵¹. Audit trails are always on, and their events can be streamed into a SIEM⁵². Before any AI runs, Glean checks the identity of the requester through the customer's SSO provider⁵³. Each planned agent step is checked first against the user's request and that user's authority⁵⁴; researchers who find a flaw are directed to a bug bounty⁵⁵.

Glean also publishes accuracy figures for two security features, both vendor claims: 96.9% for detecting prompt injection⁵⁶ and 95% for separating risk from noise when scanning unstructured data⁵⁷.

Controls still in progress

Several items on the security page are labeled as unfinished. Of the built in scans, only the one for skills carries no label; memory is tagged coming soon and agents beta⁵⁸. Governance policy enforcement for agents is tagged coming soon⁵⁹, and so is the option of giving chosen agents or users a policy of their own⁶⁰. Oversight of agents that browse the web remains in beta⁶¹. The fact record also notes placeholder text on the security page when it was read⁶².

For an operating partner, those labels matter more than the length of the certification list. A portfolio company that wants agents to take actions in finance or customer systems needs the policy controls in place on day one, and on Glean's own page several of them were not yet there.

Hosting and data

On hosting, the security page cites SafetyCulture, which keeps Glean within a cloud tenant it controls⁶³. Support for multiple clouds and flexible model choice appears on the platform page⁶⁴, and Ericsson's example on the home page mentions hybrid governance of AI⁶⁵. Data residency regions, on premises installation and self hosting are not publicly stated on the pages we opened.

Price and timeline

Glean publishes no prices. Typing in its pricing address returns a fund to the home page, and the pages read contained no plan or price at all⁶⁶. Glean Intelligence is said to balance price against performance⁶⁷, but that claim comes with no number.

The implementation path is described only through customers. One customer quote on the home page puts the rollout at three weeks⁶⁸. TIME's case study gives two numbers that a fund should read together: go live inside 3 weeks⁶⁹, and an integration across the enterprise that ran 5 months⁷⁰. The documentation gives administrators a setup page⁷¹; a standard implementation plan with dates, though, is not publicly stated on the pages we opened. Home page metrics like annual hours saved per user are Glean's own figures⁷².

Customers and use cases

Glean's customer list is long and well known. The references are household names: Booking.com with AI at 14,000 employees⁷³, Zillow reporting adoption of 80%⁷⁴, and Ericsson past 2.7k agents⁷⁵. Quotes also come from senior bankers, including the chief transformation officer at DBS Bank⁷⁶. T-Mobile is cited for cutting transfers to live agents in customer service⁷⁷.

The use case pages cover engineering, where Glean ties together code, tickets, docs and incident history⁷⁸; retail, where market signals feed decisions on merchandising and price⁷⁹; healthcare⁸⁰; and professional services, where past proposals and deliverables feed new client work⁸¹. Healthcare and higher education appear among the industries listed⁸², and customer service among the departments⁸³.

For a fund, the professional services and engineering examples are the most transferable. A newly acquired software or services business often struggles to reuse its own past work, and those are the cases Glean describes.

Glean and Nexus

The ranking places Nexus first and Glean third, and the reason is the kind of work each is built to do first. Glean is a knowledge foundation that is reaching toward agents. Nexus starts with agents that complete work inside a portfolio company's systems. Its consultants build the first one, and the fund sees it in production at week four of the trial, a trial lasting three months, with pricing per role automated⁸⁴. On reach, the counts claimed are 4,000 plus for Nexus⁸⁵ and 275 plus for Glean²⁷.

Glean is stronger on published security detail. Its pages state SSO checks⁵³, SIEM streaming⁵² and permission inheritance³⁷, while for Nexus single sign on and encryption are not publicly stated on the pages we opened. A fund choosing between them for one portfolio company should weigh that difference against the speed and pricing model that Nexus offers.

Questions for a Glean evaluation

1. What will the license cost for a portfolio company of our size, and how does it scale with headcount? 2. Which of the coming soon and beta controls will be generally available before our go live date? 3. Why do the connector counts differ between 275 plus and 107, and which of our systems are covered? 4. In which regions can our data reside, and could the product be hosted in a tenant we control? 5. Who will run the implementation, and what has the timeline been for companies like ours?

Verdict

Glean is the strongest knowledge platform in this assessment. Its permission model³⁷³⁹, its certification list⁴⁵ and its scale⁶ make it a safe choice for a portfolio company whose main problem is finding and reusing what it already knows. It ranks third for a private equity portfolio because several agent controls are still in progress⁵⁹⁶⁰⁶¹, and because neither prices nor a standard timeline are public⁶⁶. A fund that wants measured operational results inside a hold period will look first at Nexus, with Glean a strong option where knowledge access is the bottleneck.