Glean:Go 2026 Dispatches from the Pier

Glean:Go – Dispatches from the Pier

Glean just wrapped up its second annual Glean:Go conference in San Francisco, where customers, partners, and analysts saw the company's latest enterprise AI advancements. The most important announcements were not simply new product features; they signaled Glean's move beyond individual AI assistants toward a governed enterprise AI operating model.  Three announcements stood out: Glean Intelligence, Glean AI Gateway, and Glean Transform.

Glean Intelligence

Tokenomics has become a common topic during any enterprise AI discussion today.  As models and agents move from pilots to daily operations, organizations are realizing that AI consumption can't be treated as fixed software expenses. Cost changes with model selection, reasoning effort, context size, tool calls, retries, and workflow design.

For the average employee, model selection is a distraction. A person trying to summarize a policy, investigate a technical issue, or draft a customer response should not need to know which model is best, how much reasoning it requires, or whether a different provider has become faster or less expensive.

This is where Glean's Auto Routing comes into play. The platform evaluates the work at runtime and routes the request to an appropriate model and reasoning level. That is more significant than simply offering a model picker. The system is making two decisions on the user's behalf: 

1. Which model fits the task
2. How hard that model should think

This matters because enterprise work is heterogeneous. Simple search should not consume the same resources as a complex analysis. Conversely, a high-risk or complex task may justify deeper reasoning and a more capable model. Routing allows those decisions to happen in the background while the employee experiences one consistent AI interface.

As open models become more pervasive, one concern we often hear from customers is how to test open models in their environments.  Glean's Model Hub addresses this by giving organizations access to multiple model options through a common enterprise experience. That creates a model portfolio that can be managed and governed. Efficient models can handle high-volume, lower-complexity work, while more capable models are reserved for tasks that genuinely need them. Importantly, new models can be tested without redesigning every workflow.

The ability to change the underlying models without forcing every employee to become an AI model specialist is the real advantage. Pricing, availability, latency, data-residency requirements, and model quality will continue to change. A model-independent architecture gives the enterprise room to adapt.

This is also where Glean Waldo is relevant—not as another destination for users, but as an abstraction layer for reasoning effort. If the platform can determine how much thinking a task requires before sending it to a specialist model, organizations can avoid making maximum reasoning the default. That helps preserve quality where it matters while reducing unnecessary latency and consumption.

The result is a simpler user experience and a more rational technology strategy: users focus on outcomes; the platform manages model complexity.

Glean AI Gateway

As AI spreads across departments, governance cannot remain an afterthought. Without centralized visibility, organizations can accumulate overlapping tools, inconsistent models, unclear ownership, and unpredictable consumption patterns. This is AI sprawl—and it is both a financial and a risk problem.

Glean's AI Gateway provides visibility and usage controls to address the problem at platform level. It gives organizations visibility into usage by users, department, and models, creating the basis for policies that reflect budget and business priorities. That enables practical management questions:

• Which workflows create the most cost per completed task?
• Where are premium models being used for routine work?
• Are usage patterns aligned with approved priorities?
• When should routing change as demand, pricing, or model availability shifts?

Many organizations already apply FinOps discipline to cloud spend; the same principles now apply to AI consumption.  The gateway provides a control point for managing the model portfolio, monitoring spend and adjusting policy without requiring every business unit to invent its own approach.

Governance can't just be limited to model access. Autonomous agents introduce additional variables: planning steps, retrievals, tool calls, retries, escalations, and the authority to act. Organizations need to define the intended outcome, scope credentials, set limits on loops and tool calls, track cost by workflow, and assign owners for value and risk.

The right governance model can't be "central IT approves everything." That approach stifles innovation and is unsustainable. A better solution is centrally managed guardrails with distributed, outcome-focused results. Employees and business units should be able to apply AI to work, while the organization retains visibility into what is being used, what it costs, what data it touches, and what actions it can take.

Glean Transform

Two of the most common AI questions facing companies today are: Where should we start? and How do we measure success? Organizations need to identify which workflows or processes are good candidates, but they also need to determine the appropriate actions or interventions and then provide a way to measure the results. 

Rather than starting with a list of targets, Glean Transform uses its deep enterprise context to examine how work gets done.  Transform evaluates calendars, documents, messages, CRM, and code—to build a view of recurring workflows, roles, handoffs, and friction points.

By connecting workflow discovery, opportunity prioritization, solution design, and ongoing measurement, Transform can reduce the time to realize quantifiable ROI.  It allows leaders to focus resources on the most impactful business problems even if they're buried deep in an organization. This reduces the risk of investing in poorly designed use cases and provides alignment between business units, IT, and AI teams.  Most importantly, it provides a mechanism to measure the business value derived from your investments.

The best of the rest of announcements from Glean:Go 2026

While Glean Intelligence, AI Gateway, and Transform garnered much of the attention, those weren't Glean's only announcements.  Glean Tau is a new AI desktop workspace and coding assistant.  It's built on an open-source harness and combines Glean's permission aware enterprise context with local files and applications.  This allows customers to move from search and in-browser assistance to secure, delegated desktop workflows.

Team Chat was another highlight of the event.  While most enterprise work is collaborative, the current AI chat tools are often private or single threaded.  Team Chat lets multiple teammates work together in the same shared thread or interactive AI artifact.  In short, it creates a path for AI-assisted work to persist across team discussion rather than hidden in personal chats.

From AI adoption to measurable performance

The gap between individual productivity and enterprise performance remains real. Most of our customers agree that workforce AI tools help their employees get work done faster and save time, but few can prove they provide more than personal productivity gains. Closing the gap between personal productivity and organizational value requires more than another assistant or another model.

It requires an operating model that makes the complexity invisible to users, the economics visible to leaders, and the controls enforceable across the enterprise.

That is the real significance of Glean's announcements. Model independence protects organizations from model lock-in. Tokenomics connects technical design to measurable value. The AI Gateway turns governance into an operating capability rather than a policy document.

The future of enterprise AI will not be won by choosing the smartest model. It will be won by building a system that continuously chooses the right model, the right reasoning effort, the right context, and the right level of authority for the workflow.