By James Aspinwall, co-written by Alfred Pennyworth (my trusted AI) — March 7, 2026, 12:10
Two Sides of the Same Coin
Glean owns the knowledge layer. It indexes every app, every document, every Slack thread in your company and builds a living graph of who knows what, who works with whom, and what matters right now. It answers questions, surfaces context, and respects permissions down to the individual user.
WorkingAgents owns the orchestration layer. It connects AI models to tools, APIs, and workflows through the Model Context Protocol (MCP), turning natural language into action. It doesn’t store your knowledge — it moves through it, coordinating agents that do real work.
Together, they close the loop between knowing and doing.
The Gap Each One Fills for the Other
Glean’s strength is retrieval. Its limitation is execution. Glean can tell you the company’s PTO policy, find the Q3 revenue deck, or summarize a Jira epic. But it won’t file the PTO request, update the deck with new numbers, or reassign the Jira tickets. It finds answers. It doesn’t take action on them.
WorkingAgents’ strength is execution. Its limitation is context. WorkingAgents can orchestrate a multi-step workflow — pull data from a CRM, draft an email, update a spreadsheet, notify a team. But it needs to know which CRM record, what to say in the email, who to notify. It acts. It doesn’t inherently know the organizational context behind the action.
Put them together and you get an AI system that understands your company deeply and can act on that understanding.
Partnership Areas
1. Glean as the Knowledge Backend for WorkingAgents
WorkingAgents agents could query Glean’s API as a tool — the same way they already call CRMs, databases, and external APIs. When an agent needs organizational context (“Who owns the Acme account?” or “What’s our standard NDA template?”), it asks Glean instead of requiring the user to provide that context manually.
This turns every WorkingAgents workflow into a context-aware workflow. The agent doesn’t just follow instructions — it understands the company well enough to fill in the gaps.
2. WorkingAgents as an Action Layer for Glean
Glean already surfaces answers and generates content. The natural next step is acting on those answers. A Glean connector built on WorkingAgents’ MCP infrastructure would let Glean users go from “find the answer” to “do the thing” without leaving Glean’s interface.
Example: A user asks Glean, “What’s the status of the Morrison deal?” Glean finds the answer across Salesforce, email, and Slack. Then a WorkingAgents agent — triggered from within Glean — drafts a follow-up email to the client, updates the CRM stage, and schedules the next check-in. One question, full resolution.
3. Permission-Aware Agent Orchestration
Glean’s real-time permission model is one of its strongest features. Every answer respects the user’s actual access rights — no data leakage across roles or departments.
WorkingAgents could inherit Glean’s permission graph when executing workflows. Instead of building a separate permission system for every tool an agent touches, the orchestration layer defers to Glean’s understanding of who can see and do what. This is critical for enterprise adoption where a sales agent shouldn’t access HR data, even if the underlying systems are technically connected.
4. Enterprise Graph-Informed Routing
Glean’s enterprise graph knows organizational structure — teams, reporting lines, expertise areas, project ownership. WorkingAgents could use this graph to make smarter routing decisions:
- Approval workflows: Route a purchase request to the right budget owner based on Glean’s org graph, not a hardcoded list.
- Expertise matching: When an agent needs human input, route to the person Glean identifies as the subject matter expert.
- Escalation paths: If an automated workflow hits an edge case, escalate to the right team based on Glean’s understanding of ownership.
5. Unified Personal Assistant Experience
Glean has a personal assistant. WorkingAgents orchestrates task workflows. A combined experience gives users a single interface where they can ask questions and take action, with the system knowing when to retrieve and when to execute.
“Prepare me for my 2pm meeting” becomes: Glean retrieves the attendees, recent communications, and relevant documents. WorkingAgents compiles the brief, pulls the latest numbers from the data warehouse, and drops the summary into the user’s notes app — all before 1:55pm.
What a Joint Customer Gets
| Capability | Glean Alone | WorkingAgents Alone | Together |
|---|---|---|---|
| Find information | Yes | No | Yes |
| Execute workflows | No | Yes | Yes |
| Org-aware context | Yes | No | Yes |
| Multi-tool orchestration | No | Yes | Yes |
| Permission-safe actions | Partial (read) | Configurable | Full (read + write) |
| Personal assistant | Search-focused | Action-focused | Complete |
The Integration Path
The cleanest integration is MCP-native. WorkingAgents already speaks MCP. Glean exposes APIs for search, retrieval, and graph queries. Wrapping Glean’s API as an MCP tool server means any WorkingAgents agent can query Glean as naturally as it queries a database or sends an email.
Going the other direction, Glean’s extensibility framework could host WorkingAgents connectors, letting Glean users trigger orchestrated workflows from within Glean’s interface.
Neither company needs to change its core architecture. The integration is at the protocol level — clean, maintainable, and respecting each platform’s strengths.
The Bottom Line
Glean makes your company’s knowledge accessible. WorkingAgents makes it actionable. The partnership thesis is simple: enterprise AI needs both understanding and execution, and building one well is hard enough. Better to combine two platforms that each do their part exceptionally than to build a monolith that does both parts adequately.
The companies that win with AI won’t just be the ones who find answers faster. They’ll be the ones who act on those answers automatically, safely, and in context. That’s the synergy.