Glean’s $300M ARR milestone puts token economics at the center of enterprise AI buying

Glean has crossed $300 million in annual recurring revenue, up from $100 million just 15 months ago, a pace that says as much about enterprise AI procurement as it does about product-market fit. In a category that once traded almost entirely on breadth of model capability, Glean’s pitch is increasingly about efficiency: if a system can understand a company’s internal context well enough to avoid wasting tokens, it can be easier to justify at scale.

That matters because enterprise AI budgets are no longer abstract line items. They are being compared against the cost of integrating internal data, governing access, and proving that the system produces useful output often enough to offset both the software bill and the operational overhead. Glean’s growth suggests that buyers are starting to evaluate AI not just by model quality, but by how much work it takes to make the model useful inside a real organization.

At the center of Glean’s argument is what it calls a context graph. In the company’s framing, the graph learns from internal systems to understand a business’s structure and needs, then uses that understanding to shape prompts and retrieval so the model receives less irrelevant material. The practical claim is simple: better context means fewer tokens spent on noise and more tokens spent on information that improves the answer.

That distinction matters because token usage is now part of the economics of enterprise AI deployment. A model can be powerful and still be expensive to operate if every request pulls in too much unhelpful context. A system that can narrow the prompt to what is actually relevant may reduce cost per task, improve latency, and make usage patterns more predictable. Those are not flashy benefits, but they are the kinds of benefits procurement and finance teams can model.

The technical catch is that this only works if the underlying context layer is accurate and current. A context graph has to map messy internal systems into something the AI can use, which means ingesting and linking data from across an organization’s tools, repositories, and workflows. That is not a one-time setup problem. It is an ongoing integration and maintenance problem.

For adopters, the deployment questions are as important as the model questions. How much internal data can be connected? Which sources are authoritative? What access controls apply to which users? How are sensitive documents handled? How often does the system need to be updated as teams, permissions, and workflows change? The more heavily a product depends on internal context, the more attention it demands from security, data governance, and platform teams.

That creates a useful but uncomfortable reality for vendors: token efficiency is only one part of the ROI equation. If the cost of integrating data, managing permissions, and monitoring behavior is too high, the savings from reduced token use can be partially or fully offset. Buyers will need to measure net economics, not just raw usage reduction.

Glean’s timing is notable because it is now competing in a market crowded with heavyweight entrants. Google, Microsoft, OpenAI, Anthropic, Salesforce, and Atlassian are all building enterprise-facing AI search and assistant capabilities, and each brings distribution advantages that smaller vendors cannot match. In that setting, Glean is arguing that its differentiator is not just being early, but being structurally more efficient at understanding the enterprise itself.

That is a plausible near-term moat, especially if the company can demonstrate that its context graph consistently reduces token consumption while improving relevance in large deployments. But incumbents have their own advantages: they already sit close to the productivity suites, cloud platforms, and identity systems where enterprise AI is often deployed. If they can deliver “good enough” context retrieval within bundled products, the market could quickly shift from technical differentiation to packaging and price.

The broader implication of Glean’s growth is that enterprise AI buying criteria may be changing. A year or two ago, teams often started with model quality and ended with integration details. Now the order may be reversing. Buyers are increasingly asking whether an AI system can be made context-aware enough to be worth its ongoing operating cost.

That does not make the context-graph approach a guaranteed winner. It does, however, suggest that the next stage of enterprise AI competition will be judged on more than benchmark performance. The winners may be the products that can align model capability with internal knowledge, keep token spend under control, and survive the governance burden that comes with being embedded in real company systems.

Glean’s $300 million ARR milestone is therefore more than a growth headline. It is a sign that token economics, data integration, and administrative control are moving from implementation details to purchasing criteria. If that shift continues, enterprise AI vendors will need to prove not only that they can answer questions, but that they can do so in a way finance teams, security teams, and platform teams can actually sustain.