Back office ops · Production

Notion builds collaborative agent orchestration using Claude Managed Agents

The problem

Deploying agents at scale requires solving team collaboration — approval flows, shared context, and parallel work — but agents currently feel one-to-one, and customers resist complex multi-agent chains just to accomplish simple tasks.

Workflow diagram · grounded in source
1
Task creation triggers Claude session
trigger
“You create a task, move it to "ready to start," and it invokes a Claude session”
2
Claude gathers connected context
ai_action
“Claude picks up context from connected pages, our design system, API docs, and product requirements documents”
3
Parallel agent execution
ai_action
“they can kick off a ton of jobs in Notion, 30 or 40 at the same time”
4
Route tasks for approval
routing
“our platform routes them to the right person for approvals”
5
Human review of agent output
human_review
“People now see Claude show up in their Notion task doing work, and they give a thumbs up or thumbs down”
6
Skill auto-update feedback loop
feedback_loop
“Claude identifies the lessons the agent can learn and feeds them back into the skills as updates. A lot of these skills are basically auto-maintained”
Reported outcome

Notion integrated Claude Managed Agents into their task board, collapsing 12 hours of prototyping into about 20 minutes, enabling 30 or 40 parallel agent jobs with automatic approval routing, and creating auto-improving skills from user feedback.

Reported metrics
Prototyping time12 hours of prototyping work collapse into about 20 minutes
Parallel jobs capacity30 or 40 at the same time
Skill maintenanceauto-maintained
Reported stack
Claude Managed AgentsNotion
Source
https://www.anthropic.com/customers/notion-qa
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Notion integrated Claude Managed Agents into their task board, collapsing 12 hours of prototyping into about 20 minutes, enabling 30 or 40 parallel agent jobs with automatic approval routing, and creating auto-improvi…

What tools did this team use?

Claude Managed Agents, Notion.

What results were reported?

Prototyping time: 12 hours of prototyping work collapse into about 20 minutes; Parallel jobs capacity: 30 or 40 at the same time; Skill maintenance: auto-maintained (source-reported, not independently verified).

How is this back office ops AI workflow structured?

Task creation triggers Claude session → Claude gathers connected context → Parallel agent execution → Route tasks for approval → Human review of agent output → Skill auto-update feedback loop.

Related back office ops cases