- The Ravit Show
- Posts
- Breaking: Glean launches third-gen Assistant and Enterprise Graph
Breaking: Glean launches third-gen Assistant and Enterprise Graph
Ravit Jain
September 25, 2025
What happened
Glean just announced its third-generation Assistant - now delivering deep personalization, contextual awareness, and true agentic intelligence. Glean Assistant now completes multi-step work, powered by a new Enterprise Graph that understands people, content, and workflows. New platform capabilities also make Glean Agents more powerful than ever, adding richer actions, MCP directory and host support, and more intuitive ways to build and deploy agents at scale.
The core: Enterprise Graph
A living map of enterprise context that combines Glean’s existing knowledge graph with a personal graph for each employee that reflects their projects, collaborators, and work style. The latest Enterprise Graph gives AI the context to answer, plan, and act.
The Assistant: personal, contextual, agentic
A personal graph that truly knows you, understanding your tasks, projects, collaborators and unique working style
Learns tone-of-voice and writing style for email, docs, and chat
Canvas: a new interactive workspace for creation and quality content
Plans tasks, asks for missing info, and finishes the job
Fast mode for quick asks, extended thinking for deep work
Understands images when answers need visual context
No advanced prompt engineering required!
Agents: end-to-end work, open by default
Chat-first builder so anyone (not just developers) can create and refine agents
100+ new native actions across Glean-connected applications
Curated directory of 20+ validated, popular MCP servers that are easy to configure and ready to use
Version history and safe rollout for every agent
Live today: Enterprise Graph, personal graph, third-gen Assistant, Agentic Engine 2 in Assistant, deep research, contextual image understanding, schedule agents, featured agents, knowledge queries, fast and extended thinking modes, enhanced Chat API, agent version control, agents respond to user inputs.
Beta: Personalized writing, conversational agent builder, Assistant routes to agents automatically, agents toolkit, remote MCP servers, agent looping.
Coming soon: Canvas, 100+ actions, MCP Directory, LLM model choice in Assistant, Agents with Agentic Engine 2.
My take
Context is the unlock. The Enterprise Graph gives AI the structure to move from summaries to outcomes.
Prompts get lighter. Planning inside the Assistant lowers the bar for adoption.
Openness matters. MCP, SDKs, and model choice reduce lock-in and speed real use.
Try in week one
Have Assistant prepare a QBR brief that pulls the right sources and drafts the outline
Ship a Slack daily-summary agent that DM’s you updates
Best,
Ravit Jain
Founder & Host of The Ravit Show