Glean Agentic Engine 2.0 combines the enterprise graph and personal graph to create context aware agents
News

Glean Thinks Its AI Understands Your Company Better Than You Do

5 minute read
Siobhan Fagan avatar
By
SAVED
Glean launched Enterprise Graph, combining company and personal context for AI agents, plus Assistant 3.0 with personalized writing at Glean:LIVE.

In Brief:

  • Glean Enterprise Graph launches to provide complete picture of how a company works.
  • Third generation of Glean Assistant, a personalized and agentic assistant that executes complex tasks.
  • New SDKs and MCP support broaden agent capabilities and interoperability.

Glean wants to know everything about how your company works — and it's betting that knowledge will set it apart in the competitive enterprise AI market.

On September 25 at the Glean:LIVE event, the company launched Glean Enterprise Graph. The Enterprise Graph builds on its six-year-old Knowledge Graph, which mapped people and corporate knowledge, to include insights into projects, processes and data. Together, these layers provide the context AI needs to deliver accurate responses to questions and clearer marching orders for AI agents.

As Glean CEO Arvind Jain put it, the goal is "AI that fully understands how your company works, even better than you do."

visualization of Glean's enterprise graph
Glean

Table of Contents

Glean Personal Assistant Version 3: Add Context and Action

Context was the central theme running through all of the announcements made at Glean:LIVE.

Case in point: the release of version 3 of the Glean Personal Assistant. The updated version combines the Personal Graph introduced at Glean:GO in May and the Enterprise Graph to move beyond being a knowledge finder to become a personal assistant with the ability to execute complex tasks and re-assign tasks to more specialized agents. 

Founding engineer Chaitanya Asawa compares the difference between typical AI assistants and the new Glean assistant to that of a first day hire vs. an experienced colleague providing support.

Glean founding designer Cindy Chang cited an example of how the assistant recognizes context by using a question about PR. PR can mean very different things based on whether the question came from a software engineer or from a communications manager. For the former the response would surface GitHub pull requests and surfaces actionable tasks needed in these instances. For the comms manager, the assistant surfaces deadlines for public relations in advance of an event, flags the feedback organizers are waiting on and pulls all relevant materials and dates from across an array of internal working tools (e.g. calendar, Google Drive, Asana to reassign tasks, etc.).

Each assistant is for one individual only and as such, can perform pro-active tasks on that person's behalf, such as creating briefs in advance of a meeting. 

The company is currently internally testing a feature where the assistant pro-actively surfaces critical tasks and deadlines on the Glean homepage of an individual worker. Jain described a future state where the system becomes proactive, "Instead of flagging that a product launches off track, it will pinpoint the real issue, like a dependency that was missed, and drive the resolution to get things back on track."

Strategy won't be a quarterly exercise. It will be continuous as AI scrutinizes and optimizes processes in real time, always learning, always improving how you work.

- Arvind Jain

CEO, Glean

Other updates to the assistant included:

  • Personalized Writing Assistant — Glean Assistant now learns an employee's preferred tone depending on audience and context. It adjusts the style, tone and length of the messages based on the medium — e.g. a Slack message vs. a quarterly report — and adjusts this knowledge as your own writing style changes. It holds three to five writing personas for an individual employee.
  • Glean Canvas — A space for the employee and assistant to work together directly on any generated writing, avoiding the need for further prompts for refinement.

All of these updates aim to "lift the responsibility away from the user," including the need to identify the right agent for a given task.

Glean Agents and Developer Platform Updates

Agents form the backbone of  Glean's broader automation platform. The company teased a forthcoming "vibe coding" conversational agent building capability. With it, the company states any employee can chat in natural language in the Glean interface to build and deploy agents. The builder provides the steps involved in creating the agent and adapts based on employee follow up requests. 

creating a simple agent with Glean through natural language prompts. user can fine tune, switch models and has oversight into entire process
Glean

Over one hundred new actions for agents are coming soon to further support agent-building efforts. Notable among them are native integrations with Slack and Salesforce, access which appeared threatened following the decision in May by Salesforce to shut down access to Slack's API to companies outside its approved marketplace list. Glean didn't provide specifics around the scope of the agreement it reached with Salesforce, but for now, the access question appears resolved.

The company also introduced a new MCP directory with 20 pre-loaded servers to commonly used workplace software including Asana, Notion and GitHub. 

Availability of Features Introduced at Glean: LIVE

The company packed a number of announcements into the relatively short Glean:LIVE event. Below is a high-level view of what was announced and when the feature is expected.

Feature AreaSpecific Feature/ComponentAvailability Status/DescriptionSupporting Sources/Context
Foundational ArchitectureEnterprise GraphGenerally AvailableA living map and central intelligence layer that understands people, content, workflows and connects enterprise graph to personal graphs.
 Personal GraphGenerally AvailableA private, per-user knowledge graph capturing employee's work habits, style, projects and collaborators for deep personalization.
 Agentic Engine 2Generally Available (Assistant only)New architecture boosting performance for complex queries, integrating enterprise understanding into planning, execution and memory.
 Deep ResearchGenerally Available*Creates reports with citations from information found in your company's knowledge and the web. *Subject to usage-based pricing.
Glean AssistantThird-generation Glean AssistantGenerally AvailablePersonal and agentic Assistant that can execute multi-step work.
 Personalized writingBetaAssistant adapts to user tone and vocabulary using 3–5 writing profile clusters based on person's writing samples.
 Assistant routes requests to agents automaticallyBetaAssistant can route requests to specialized agents in natural conversation flow.
 Glean CanvasComing SoonDedicated collaborative writing and editing surface with the Assistant.
 LLM model choice in AssistantComing SoonLets users pick their preferred model within Assistant, matching model to task.
Glean Agents & AutomationScheduled agentsGenerally Available*Agents can run on set schedules (daily, weekly, monthly). *Subject to usage-based pricing.
 Wait for user input stepGenerally AvailableFeature allowing mid-run agent check-ins with users for additional input.
 Featured agentsGenerally AvailableAdmins can spotlight evergreen or seasonal agents for discovery.
 Conversational agent builder (Vibe Coding)BetaUsers can define and iterate agent steps through conversation, no technical expertise needed.
 100+ new actionsComing SoonNative integrations across platforms (Slack, Teams, Google, Jira, GitHub, Salesforce).
Developer Platform & InteroperabilityEnhanced Chat API (Agentic Engine 2)Generally Available*Enterprise-aware chat API with secure context, citations and built-in tool use. *Subject to usage-based pricing.
 Remote MCP serversBetaDevelopers can connect Glean context with external tools via OAuth MCP servers.
 Agent toolkitBetaAPIs, SDKs and MCP servers for building, running and observing agents with preferred frameworks.
 MCP directoryComing SoonDirectory of 20+ enterprise-grade MCP servers (Asana, Canva, PagerDuty, Notion, GitHub).
Learning Opportunities

Glean:LIVE Reveals Glean's Ambitions 

Glean launched in 2019 as a modern approach to enterprise search. With last week's announcements, the company clearly has set its sites on becoming much more than a search engine.

The success of the Enterprise Graph — and the new features built on it — will depend on the breadth and depth of integrations Glean can sustain. The Salesforce scare in May showed how integral these relationships are. The MCP directory is a good starting point, but with mid-sized companies often running 200-300 active apps, it's only the beginning.

Because Glean started as an internal search engine, it already has a head start: many integrations are already in place. The hundred plus new actions will be key to move existing read-only integrations toward more useful, action-oriented automations.   

fa-regular fa-lightbulb Have a tip to share with our editorial team? Drop us a line:

About the Author
Siobhan Fagan

Siobhan Fagan is the editor in chief of Reworked and host of the Apex Award-winning Get Reworked podcast and Reworked's TV show, Three Dots. Connect with Siobhan Fagan:

Main image: Glean
Featured Research