Glean Unveils Autonomous Agents That Think and Act for Enterprises
The startup’s next-generation bots leverage full organizational context to make decisions, adapt dynamically, and support employees without rigid workflows
Glean is expanding its agentic portfolio with autonomous agents it claims are the first built specifically for the enterprise. These bots leverage organizational context to make decisions, adjust their decisions as circumstances change, and function within established governance rules. The startup asserts they’re designed to support employees rather than replace them.
“Most ‘agents’ in the market today are assistants that follow a linear script: They take an instruction, run a single workflow, and stop,” Emrecan Dogan, Glean’s head of product, tells The AI Economy in an email. “What Glean is introducing is different. Our agents can independently interpret intent, select and sequence the right actions across an organization’s systems, adapt when the underlying context changes, and verify their behavior against governance policies.”
Simply calling a bot “autonomous” isn’t enough—overusing the term risks diluting its significance. So what criteria determine whether a bot truly qualifies as autonomous? According to Dogan, it boils down to whether it can iteratively self-determine the next best course of action to complete its assignment and execute the actions required to accomplish its goal. No standardized score exists to determine if an agent falls into this category.
Has Glean Pioneered Autonomous Agents for the Enterprise?
The move to self-operating bots is indicative of a trend we’re likely to see emerge in the enterprise. Most software vendors, such as Salesforce, Microsoft, HubSpot, Lattice, SAP, ServiceNow, and Dropbox, already offer agents, but Glean claims its next-generation bots are more sophisticated. “There are agent frameworks that claim autonomy, but none operate with full enterprise context,” Dogan counters. “They either lack full access to a company’s systems and data, or they don’t have the reasoning capability to make decisions dynamically as they work.”
He argues that Glean’s new agents possess “the most comprehensive enterprise context,” bringing in data from across now more than 100 third-party connectors, including Microsoft Dynamics 365, NetSuite, Ironclad, Affinity, Procure, and Canva; indexes, memory, knowledge graphs, and actions. “That’s what enables reliability at scale,” Dogan points out.
There’s also a tool search capability being added to the platform, enabling agents to automatically select the most relevant tools and/or workflows based on the assigned task. Glean’s indexing infrastructure powers this feature, as well as helping to grow the native actions and MCP-enabled tools being used in the workplace.
Glean’s autonomous embrace shouldn’t be surprising. Earlier this year, the company announced its next-generation Glean Assistant, introducing AI that no longer handles simple task automation but now facilitates end-to-end workflow support. It’s capable of reasoning through multi-step challenges, can surface correct information, and refine answers based on new information until it generates complete and actionable outcomes. While existing agentic solutions will still be available, organizations that have put their faith in Glean’s technology can use the new autonomous agents to handle whatever work needs to be done.
“In September, we made a number of improvements to our Agent Engine that unlocked new reasoning capabilities in Glean Assistant,” Dogan shares. “After seeing what those improvements did for Assistant, we wanted to apply the same updated approach to Glean Agents. The resulting agents were faster and easier to build, while delivering high-quality conversational responses. We knew that these capabilities would be critical to unlocking the power of agents for our customers, and wanted to make them available as soon as possible.”
To assuage concerns that its agents could go rogue, the company emphasizes that all actions taken are automatically checked against the assignment’s original scope using Glean’s agent alignment models. Moreover, organizations can restrict which individuals and departments are permitted to create agents (“Departmental Agent Moderator role”), and there is team-level moderation to guide usage and ensure automation is trustworthy and well-governed.
The platform also enforces user-level permissions and displays sensitivity labels in Google Drive and Microsoft Purview, which extends to both OneDrive and SharePoint, to ensure that enterprise data is appropriately secure.
“We enforce full data permissions from the upstream systems that the agents have access to, so no user sees any information they aren’t allowed to see,” Dogan remarks.
Responding to Amazon’s Frontier Agents
But Glean isn’t the only one with a so-called autonomous agent. Last week, Amazon Web Services (AWS) debuted Frontier Agents, its take on autonomous AI. The first set of bots from the hyperscaler includes those for coding, security, and software operations. While Dogan concedes that AWS’s approach is considered autonomous, “its agents still lack the full enterprise context that Glean can bring to bear. For example, the security agent cannot rely on company documentation to understand security policies; all of that must be configured manually, and manually updated as policies change.”
Not one to gatekeep, Glean is also releasing a new feature in its agent builder that lets anyone create their own autonomous agent “in a matter of minutes.” Perhaps knowing that it won’t be able to address all the use cases that organizations may have, the company is encouraging customers to use the new auto mode to build these advanced, but specialized, bots.
For Glean, it’s less about having the “first” autonomous agent. “It’s about being the first autonomous system that can act inside an enterprise with a complete understanding of how that enterprise actually works,” Dogan says. “Everyone else gives you an interface. Glean gives you an engine that reasons and acts reliably at scale.”
Today’s announcement comes after the company, valued at $7.2 billion, revealed it has generated $200 million in annual recurring revenue.
Availability
The new connectors for Microsoft Dynamics 365, along with the Departmental Agent Moderator role and the ability to limit agent creation with specific actions to individuals and departments, are all generally available today.
As for Glean’s autonomous agents, they are currently in beta. Joining it are more than 85 additional new agent actions, connectors for Ironclad, Affinity, Procure, NetSuite, and OneNote, and the ability to hide content based on sensitive labeling in Google Drive, SharePoint, and OneDrive.
As for Glean’s agent alignment models, tool search, enterprise memory, and the Canva connector, the company states all of these will be “coming soon.”
Glean's agents interpreting intent, selecting their own action sequences, adapting when context changes—that's fundamentally different from script-following bots. Self-determination without human-in-the-loop for every decision. That's the autonomy that matters in enterprise.
The governance layer (self-evaluate against policies) is the unspoken requirement though. You can't have self-directed agents without guardrails. Enterprise risk departments won't allow it. Glean doubling revenue to $200M by solving that problem is telling—enterprises will pay for autonomy if it's bounded.
Impressive move by Glean here. The critikal distinction between autonomous agents and scripted assistants is spot on: can it iteratively self-determine the next action, or just follow a linear script? Most vendors call their bots "agents" but they're really just fancy automation workflows. Glean's approach of combining full enterprise context (100+ connectors, knowledge graphs, permissions) with dynamic reasoning changes the game. The real test will be weather these agents maintain alignment without going sideways when faced with edge cases governance didn't anticipate.