Glean

Glean on cost: deep deployments, no public price

Glean is the hardest product in this catalog to price from the outside. Its pricing URL redirects to the front page, and we found no plan names or amounts on any page we read¹. Yet Glean also publishes customer stories at a scale well beyond 1,000 employees². For a finance buyer, that combination produces a specific task: build the cost model from the vendor's quote, and use the deployment record to judge whether the value side of the case is credible.

The home page calls Glean "Enterprise AI that Works"³. Its about page tells a story that begins with search inside companies and ends with a platform for AI resting on company context⁴. This profile places Glean fourth, and it explains how a 1,000 employee company can approach the cost question when the vendor publishes no numbers.

What the pages say about price

The direct answer is nothing. Pricing is not publicly stated¹. One platform page line promises AI tuned for price and performance; it describes the product and quotes no figure⁵. The pricing unit (seat, usage or another basis) is not publicly stated on the pages we opened.

The model choice behind that price performance claim is published. Glean routes every task to one of over 40 models, open and frontier alike⁶. For a buyer, that suggests model costs are managed by Glean inside the product, but whether they are included in the price or billed through is not publicly stated on the pages we opened.

Building the cost case from scale evidence

What Glean does publish is a record of large deployments, which helps a buyer judge whether the value side of a business case is realistic at 1,000 people.

1. Booking.com: AI rolled out to 14,000 staff². 2. Zillow reports 80 percent adoption⁷. 3. Ericsson reports more than 2,700 agents⁸. 4. DBS Bank: its chief transformation officer appears as a reference⁹. 5. T-Mobile: fewer customer service handoffs to live agents¹⁰.

Headline figures on the home page include 110 hours a year saved for each user¹¹. That number is reported by the vendor and has not been measured by us. A finance team using it in a business case should replace it with a figure from its own pilot.

Adoption matters to cost under any seat based model, because unused seats are pure overhead. Zillow's 80 percent adoption figure⁷ is the kind of evidence a buyer should ask Glean to reproduce for a company of its own size and sector.

What the product does for the money

Glean's product lines frame what a buyer would be paying for. Glean Enterprise Context is one named product line¹²; Glean Assistant is billed as an AI coworker for the enterprise¹³. Glean Search claims to be the next generation of workplace search¹⁴, covering upward of 100 tools from a single box¹⁵. Context, says the platform page, is what lets AI take on real work¹⁶; connectors gather signals, an index holds the data, and the Enterprise Graph records how people, content and processes relate¹⁷.

Two features bear on the cost of ownership. Relevance improves as the search learns each company's vocabulary¹⁸, and LLMs interpret questions typed in plain language¹⁹; both should reduce tuning work after launch. And the open platform lets a company index its data once and use it everywhere through APIs, a web SDK and MCP²⁰, with Search, Chat and Agents APIs listed^21,22,23. If a company plans to build other AI tools on the same index, that reuse can spread the cost of Glean across more than one project.

The documentation adds connector features that matter for knowledge work: BigQuery queries can run with each user's access²⁴, users can bring their own Jira, Gmail and other accounts into answers²⁵, and Slack can be searched in real time through real time search APIs²⁶. A developer MCP server endpoint is published²⁷.

Company scale

Glean's own figures describe a large vendor. Annual recurring revenue stands at $300M by the company's account²⁸, three times the $100M of 15 months earlier²⁹. Within a year of launch it raised $100M and was valued at $1B³⁰; the Agents launch came with a $7.2B valuation³¹. When it launched publicly, its search product already ran at over 40 companies³². Arvind Jain is the founder and CEO³³.

For a multiyear cost commitment, vendor scale reduces continuity risk. It does not, by itself, say anything about price.

Implementation time and cost

A standard implementation timeline is not publicly stated beyond customer reported timelines. The customer reports are useful. Three weeks is the time to running that one customer cites³⁴. TIME went live after 3 weeks³⁵, while the full enterprise integration took 5 months³⁶.

Those two TIME figures describe different milestones, and both belong in a cost model. Go live is when users begin to get value. Full integration is when the implementation effort ends. A finance team should budget internal staff time for the whole period, since a packaged implementation service with its own price is not publicly stated on the pages we opened. The admin documentation includes a setup page for organizations³⁷.

Integrations and their cost effect

Every system a product connects to out of the box is an integration project the buyer does not fund. Glean offers more than 275 connectors³⁸, described on the connectors page as out of the box across enterprise apps and ecosystems³⁹. The documentation hub lists 107 connectors⁴⁰, and we found no explanation for the gap between that and 275.

For systems outside the list, the options carry their own effort. An Indexing SDK loads custom data⁴¹, an OpenAPI spec defines custom actions⁴², and MCP tools can supply either⁴³. Teams can request new connectors⁴⁴. Developers get client libraries in four languages: Java, Go, TypeScript and Python⁴⁵. The Search, Chat, Agents and Skills APIs are generally available⁴⁶; five more, including Triggers and Admin, are labeled experimental⁴⁷.

A buyer should list its systems, check each against the connector list and price custom work for the gaps.

Security and compliance

Glean's certification list is broad. Six attestations appear on both the home and security pages: GDPR, ISO 27001, ISO 42001, HIPAA, TX-RAMP Level 2 and SOC 2 Type II^{48,49,50,51,52,53,54}. An always on audit trail can stream events to a SIEM⁵⁵. SSO verifies who is making a request before AI runs⁵⁶. Before an agent acts, Glean compares the action with what the user meant and is allowed to do⁵⁷. A bug bounty program handles vulnerability reports⁵⁸.

As for hosting, SafetyCulture runs Glean inside its own cloud tenant⁵⁹, and Glean says it supports several clouds⁶⁰. Data residency region options are not publicly stated.

Several security features are described with accuracy figures, which are vendor claims. Glean's security page gives two accuracy numbers: 96.9 percent on prompt injection⁶¹, and 95 percent when scanning unstructured data for risk⁶². Skills are scanned today, agents in beta and memory later⁶³; the docs cover skill scanning⁶⁴. The Ericsson story carries the label hybrid AI governance⁶⁵.

Governance and permissions

Permission handling is Glean's most consistent message. Reads and writes alike obey source permissions⁶⁶. Connectors inherit them⁶⁷, and changes made in the source app appear in results immediately⁶⁸. Agent actions can require approval before execution⁶⁹. Agents may run under narrowly scoped service credentials that admins manage⁷⁰.

Some governance controls are not yet generally available. Policies governing agents are still labeled coming soon⁷¹, and so are separate policies for particular agents or users⁷². Controls on agents using a browser remain in beta⁷³. For cost planning, agent level and user level policies are the kind of control that would let finance limit spend by team; a buyer should ask when they ship.

Use cases relevant to a 1,000 employee company

Use cases on the home page span departments. Engineers get code, tickets, documents and incidents in one place⁷⁴; retailers get help with merchandising and pricing⁷⁵; healthcare is covered⁷⁶; and professional services firms reuse old proposals and deliverables in new client work⁷⁷. Customer Service appears in the department list⁷⁸. The breadth suggests a seat population that could span most of a 1,000 person company, which makes the unknown pricing unit even more important to pin down.

Questions to put to Glean

Because the public record is silent on price, the quote request carries more weight than usual. A 1,000 employee buyer should ask:

1. What is the pricing unit, and how does the price change between 1,000 and 1,500 employees? 2. Are model costs across the 40 plus models included⁶? 3. What implementation support is included, and what does the 5 month integration seen at TIME involve³⁶? 4. Which of the connectors we need are in the out of the box list³⁹, and what would custom connectors cost? 5. When will agent level and user level policies leave coming soon status⁷²?

On the about page, Glean says the customer decides what AI can see and what protections apply⁷⁹. The same principle applies to the contract: the buyer should control which parts of the cost are fixed and which vary.

Limits a finance team should record

1. Pricing is not publicly stated¹. 2. A standard implementation timeline is not publicly stated. 3. Data residency regions are not publicly stated. 4. Several agent governance controls are coming soon or in beta^71,73. 5. Headline time savings are vendor reported¹¹.

Why it ranks fourth

Glean ranks fourth on cost because it pairs a deployment record that includes a 14,000 employee rollout² with no public pricing at all¹. Its scale evidence (14,000 employees at Booking.com², 80 percent adoption at Zillow⁷) makes the value side of a business case more credible than for the products below it. It ranks below Gemini Enterprise and Copilot Studio because both publish at least part of their pricing^80,81 and Glean publishes none¹. It ranks below Nexus because Nexus has a pricing unit, per role automated and not tied to headcount^82,83, includes consultancy support with no per diem or consultant time⁸², and puts the first agent live in 4 weeks inside a 3 month proof of concept⁸².

A buyer considering Glean should ask for a written quote at 1,000 employees that names the pricing unit, states what model usage is included, and sets out the implementation support and timeline.