Glean Assistant 3.0 Turns AI Into an End-to-End Workflow Partner
The upgraded assistant adds agentic AI, a vibe coding builder, and developer tools to help enterprises automate complex tasks.
Two years after debuting its Glean Assistant, Glean is rolling out a major upgrade that it claims moves the AI from simple task automation to full end-to-end workflow support. Alongside the enhanced assistant, the company is launching a new conversational AI agent builder and a suite of developer tools designed to make it easier to extend and integrate the platform.
“Over the last year or so, knowledge work has become a lot more AI-driven. That was a time when people were using AI here and there to help with a subset of the work they did day in and day out. But now, the expectation, if you are a Glean customer and you have access to Glean Assistant, is that the number of tasks that you don’t use Glean [for] is…a smaller bucket,” Emrecan Dogan, Glean’s head of product, told me. In other words, people expect AI to handle a majority of work processes. “Glean Assistant v3 is very much focused on being a helper end-to-end along the way in every step of how our users get work done on a daily basis.”
A More Predictive Work Partner
As an individual assistant, Glean Assistant builds a personal graph for all employees of an organization, categorizing projects, collaborators, and work patterns. It’s a feature that mirrors what Dropbox, HubSpot, and Atlassian have also added to their respective platforms. Additionally, Glean states that it can adapt to the type of work and the medium, such as when drafting an email, document, or message. It also offers an interactive workspace called Canvas, where employees can edit content inline with real-time guidance to create clearer, more equitable, and publish-ready work.
Lastly, Glean Assistant gives users complete control over their experience. They can choose the speed of the AI’s response and the level of detail in its thinking, based on the depth and complexity of their prompt. Employees can also designate the model family used based on their company’s preferences or the LLM that generates the best response, such as OpenAI’s GPT-5, Anthropic’s Sonnet 4, or Google’s Gemini 2.5 Flash or Pro.
But beyond being a more reactive helper, Glean’s bot is also proactive (also known as agentic). Powered by the company’s second-generation agentic engine, it’s capable of reasoning through multi-step challenges, can bring together the correct information, and refine its answers based on new information until it generates complete and actionable outcomes. Glean Assistant can also supervise sub-agents to help complete complex problems and deploy scouts to uncover new information.
Should Glean Assistant not have the necessary information to form an answer, the company said it will know to ask clarifying questions. It’s also intelligent enough to determine what knowledge is needed, pulling from sources across the organization and the web, consolidating them into a single answer.
Making Glean Agents More Powerful
Besides Assistant 3.0, Glean is introducing new tools to strengthen AI agents on its platform, beginning with a revamped agent builder that turns development into a no-code, conversational experience.
“We call it like vibe coding an agent,” Dogan said. “You not only start building an agent with conversation, but then you can keep giving iterative feedback to make that agent better over time.”
The company is also adding more than 100 new agent actions across Slack, Microsoft Teams, Salesforce, Jira, GitHub, Google Workspace, and other applications. Dogan shared that this is in response to users wanting Glean Assistant to be more than just a companion helper, but as an agent that autonomously does the work end-to-end. Although Glean is doing just that, it’s not cutting humans out of the loop—employees will still play an active role in the process.
For example, a Glean agent could process a customer support ticket in Zendesk after conducting deep research and diagnosing the root cause. The bot could prepare a response and have it reviewed by the employee before being sent to the customer. If approved, Glean would then execute the appropriate action within Zendesk.
“Whether these actions map to…HubSpot, Marketo, Outlook, JIRA, etc., Glean is able to recognize which action the user wants in which use case, and either take that action autonomously or seek confirmation from the user and then take that action afterwards.”
In addition, upgraded APIs and SDKs are being released. Developers can now access all of Glean Assistant’s personalization and reasoning features to incorporate into their apps using Glean’s ChatAPI. According to Dogan, enterprise customers have been pushing the company for more robust API endpoints. In the meantime, many have cobbled together custom agents, stitching Glean Search, Assistant, and existing agents into their own makeshift stacks to meet diverse use cases. That’s no longer the case, as all of Glean’s agents and its Assistant are available as API endpoints.
Moreover, with the adoption of the Model Context Protocol (MCP) growing, Glean is making its entire platform accessible through the MCP interface. This means that it can bring its enterprise context into a host of different applications, including GitHub Copilot, Zendesk, or Salesforce’s Agentforce. Not only that, Dogan disclosed that Glean can function not just as a server but as a host. This means Glean can coordinate other services, such as Canva, Box, Linear, and Grok, serving as the brains behind the agentic system rather than an arm.
“We are operating truly open, whether or not you use Glean as the UI, as the MCP host, or MCP server,” he asserted.
Ultimately, Dogan said these AI updates are meant to help users—whether product managers, designers, marketers, or lawyers—work smarter and elevate their craft.
“The future of work is the superintelligent enterprise, where AI understands people and organizations deeply enough to transform how work gets done. At Glean, we’ve built the Enterprise Graph to make that future real, giving Glean Assistant the context to act as a true teammate,” Glean Chief Executive Arvind Jain remarked in a press release. “With today’s announcements, Glean is setting a new standard for AI at work—assistants that are deeply personal and adapt to every employee, and powerful, interoperable agents that drive work forward with trust and reliability.”
When Will This Be Available?
Not everything announced today will be immediately available. That said, Glean disclosed that customers can now use both the enterprise and personal graph, the Agentic Engine 2 (Assistant only), its third-generation Glean Assistant, the scheduling and featured agents, all knowledge queries, fast and extended thinking modes, deep research, contextual image understanding, agent version control, agents respond to user inputs, and the enhanced Chat API.
However, personalized writing, the conversational agent builder, Assistant routes requests to agents automatically, agents toolkit, remote MCP servers, and agent looping are in beta.
Lastly, Glean Canvas, the more than 100 new actions, the MCP Directory, LLM model choice in Glean Assistant, and Agentic Engine 2 (Agents) are all listed as “coming soon.”