Accelerating the journey to the superintelligent enterprise: New Enterprise Graph, Third-generation Assistant, and Agent superpowers in Glean

Modern companies are complex systems, operating every day through countless networks of people, applications, and interactions. No single person, not even the CEO, sees the whole picture. However, that whole picture is essential to achieving true enterprise superintelligence—an end-to-end understanding of the entire organization that enables everyone to do more work with AI, at a higher quality, than any human can do alone. 

Today, Glean is accelerating everyone’s path to enterprise superintelligence with the Enterprise Graph, a graph which turns scattered signals into a connected map of people, teams, and projects for each business. It’s a secure, algorithmically built, continuously refreshed model of what makes your company unique, adapting as new initiatives launch and strategies evolve by reading and understanding live work data.  

However, great work is also inherently personal. To make AI truly helpful and authentic, every employee needs a personalized solution that understands their projects, collaborators, and communication style. That’s where Glean Assistant comes in—powered by a per-user personal graph that captures each employee’s work habits, writing style, communications style, and memory.  

This superintelligence shows up directly in the new product experience: our third‑generation Assistant is personal and agentic, it takes on more work with fewer instructions, and feels like a natural extension of each employee. The same agentic capabilities power Agents, so anyone can “vibe code” automation that executes everyday work. And for developers, Glean exposes this enterprise context wherever they build—through open APIs, MCP, and the agents toolkit—so there’s no need to master context engineering; Glean handles it all for you.

Assistant: Your personal agent at work

The third‑generation Glean Assistant is more than an AI tool—it feels like an extension of you at work. Glean is uniquely capable of delivering this because it closely observes and understands data, interactions, and daily tasks and goals, which all build into each individual’s personal graph. 

Assistant leverages this personal graph to deliver deeply personalized experiences that genuinely feel tailored for you, since they’re based on what you work on, who you work with, and how you work. Assistant remains simple, intuitive, and easy to use, and has become even more capable of anticipating your needs, with no prompt engineering required. This means you can speak to it without providing extensive details about precisely what you’re looking for, and it will still be able to deliver relevant, high-quality responses. 

The Assistant is also personalized to follow your specific writing style. For enterprise work, tone and structure matter as much as content. Rather than a one‑size‑fits‑all “temperature” knob with users manually specifying their preferences, Glean tailors responses to your exact context because it understands where and what you write. You can experience this more clearly than ever through Glean Canvas, a newly dedicated writing surface for drafting polished, personalized, and professional content step-by-step alongside AI. Canvas makes it easy for everyone—engineers, sales executives, support staff, and others—to interactively edit text inline and collaborate conversationally.

When we say the Assistant is agentic, we mean it can reason about how to approach multi-step tasks, plan adaptively, adjust as it learns, and leverage the power of search to automatically identify the right data and tools to reach a goal. Users can take advantage of these capabilities—ideal for complex tasks requiring additional time to thoroughly think through the question—by selecting Thinking mode within Assistant. 

‍Agentic Engine 2 infuses enterprise understanding into planning (grounded in how you actually work), execution (finding the best tools and iterating toward completion), and memory (persisting durable learnings to the Enterprise Graph and personal graph). The Engine now better addresses direct end‑user feedback, supporting a complex query class that represented 20% of feedback. This boost is reflected in Assistant engagement, with 21% more queries run.

Agentic Engine 2 even automatically pulls from both your company and web data sources to deliver the best responses for your single query using the most relevant inputs. Better yet, tune the experience even further by selecting your model of choice directly from the Assistant, matching the model to the task you want completed. 

Consider a real Assistant query: “Create a plan for a CIO dinner in LA for the manufacturing industry. Create a messaging plan, identify prospects to invite, and research venue options.” The Assistant taps into 30+ data sources across Salesforce, Slack, Google Drive, and the web, and orchestrates an agentic flow to deliver a work‑back timeline, a short‑list of venues, strategic prospects, and brand-aligned messaging and case studies. A single, simple prompt becomes the framework for the entire field event.

Assistant is also capable of understanding and surfacing relevant images (charts, diagrams, screenshots) from your enterprise documents and data—enabling users to quickly grasp insights and make decisions faster based on the responses Assistant provides. 

“Glean’s third‑generation Assistant feels like an extension of our team. Its agentic engine intuitively plans, pulls context across all knowledge sources, iterates on its work, and delivers complete, personalized results without our team needing to over‑engineer prompts.”
- Joe Gardener, Director, Information Technology PMO at AHEAD

Agents: Vibe code agents to automate work

Vibe coding has transformed software development by showing the power of leveraging LLMs to create software through a fully conversational experience. So why not bring vibe coding to agents? That’s exactly what our new conversational agent builder does. In addition to defining an agent, users can now iterate on the agent’s steps and instructions simply by chatting with Glean—no technical expertise required. This makes advanced automation accessible to anyone, not just developers. 

More employees can create agents and make them easily discoverable across their organization. While we’ve long invested in the agent library—where customers have grown collections to hundreds of agents—we also recognize that Assistant is agent zero: it’s where users start their work. Now, organizations can curate which agents they want highlighted so they become sharable as part of the Assistant experience with agent routing.

Agents are not only easier to create with vibe coding—they’re more powerful, too. We’re introducing 100+ new native actions across Slack, Microsoft, Salesforce, Jira, GitHub, and Google so you can close the loop on work. Instead of hopping between systems, you can delegate end‑to‑end tasks to agents grounded in your company’s context via Glean’s Enterprise Graph so they take the right action. 

Beyond Glean native actions, we’re expanding the action space further with the Model Context Protocol (MCP) and a new directory, where you can enable more than 20 pre‑loaded servers to add entire families of actions.  We’re thrilled to welcome in the initial directory cohort: Asana, Canva, Notion, Box, GitHub, Thoughtspot, Hubspot, Databricks, WisdomAI, and PagerDuty, and many more coming soon. These servers are enterprise‑grade—centrally managed, secured with OAuth, and vetted for safe use at scale. 

In addition to making it simpler for everyone to create agents and automate their work, we’re introducing advanced capabilities, including scheduled triggers (run agents daily, weekly, or monthly), looping (iterate over datasets to handle more complex work), and version control (manage an agent’s lifecycle).

Platform: Context is the next data platform

AI chatbots and agents rely on enterprise context, so the next data platform must connect your content, people, relationships, and processes to truly understand how work gets done. That takes a new foundation—search and knowledge graphs—built over your existing structured and semi-structured data. Glean is already that platform, powering the next generation of enterprise tools.

While context is what sets AI apps apart, connecting and making data useful to AI is also one of the biggest hurdles in the enterprise. Enterprise AI must connect to proprietary sources, crawl for updates in near real time, chunk data, generate embeddings, tune ranking, secure access, and more. Glean handles all the hard work of aggregating enterprise data together and takes your data a big step forward by turning data into context with the Enterprise Graph. 

Glean is committed to openness and agent interoperability, firmly seeking to be the best agent development platform while also recognizing it won’t be the only one. Agents will communicate cross-platform to get work done, so we’re providing developers with new interop options: 

• A more powerful Chat API: An API to embed Glean’s enterprise-aware chat anywhere, with secure context, citations, and built-in tool use.
• Remote MCP servers: Use Glean wherever you work by bringing the power of full enterprise context to Cursor, Claude Code, Windsurf, and more using secure, centrally managed OAuth MCP servers. 
• Agent toolkit: Build, run, and observe agents using your preferred frameworks- CrewAI, Google ADK, LangGraph, OpenAI, and more—while natively tapping Glean’s search, Enterprise Graph, and permissions framework.

Glean delivers enterprise superintelligence

Glean makes it possible for everyone to leverage AI that fully understands their enterprise, their work, and how they like to get things done—turning AI from a generic chatbot into a compounding advantage for each unique enterprise. This is enterprise superintelligence.

With Glean Assistant, every employee can understand what teams are working on, the goals and strategies they’re driving toward, and the processes that move work forward so they can act on the same ground truth to accelerate progress. The same intelligence powers better‑informed agents that reflect how work actually happens, now easier to create with vibe coding and to automate end‑to‑end with 100+ new actions.

After 6 years of building knowledge graphs, we’ve reached a new clarity in how enterprises work. Next, we’ll become even more proactive, surfacing insights you never thought to ask and pinpointing ways to make processes more collaborative and effective.

You can learn more about the state of enterprise superintelligence along with everything we announced at Glean:LIVE Fall'25 here. If you’re not a Glean user and want to get started with your very own personal agent at work, sign up for a free demo today. 

Availability
Generally available: Enterprise Graph, personal graph, Agentic Engine 2 (Assistant only), third-generation Glean Assistant, Deep Research*, all knowledge queries, fast and extended thinking modes, contextual image understanding, schedule agents*, agent version control, agents respond to user inputs, featured agents, enhanced Chat API (Agentic Engine 2)*.
Beta: Personalized writing, conversational agent builder, agent looping, Assistant routes requests to agents automatically, agents toolkit, and remote MCP servers.
Coming soon: Glean Canvas, 100+ new actions, Agents (Agentic Engine 2), LLM model choice in Assistant, and an MCP directory.

The decision to purchase our services should be made based on features that are currently generally available.

*These generally available features are subject to usage-based pricing.