Case Study: Notion builds agent orchestration and cuts AI costs by 90% with Claude

Notion creates more intelligent workspaces with Claude

Notion, a collaborative AI workspace used by millions, faced the dual challenge of making its vast amounts of user-generated knowledge easily findable and then expanding from simple search to enabling teams to delegate and collaborate on complex, real work done by AI agents. This required moving beyond one-to-one agent interactions to a system with full visibility and approval workflows.

To solve this, Notion integrated Claude Managed Agents, a suite of APIs for building cloud-hosted agents. The solution leveraged Claude's high-quality reasoning and instruction-following to handle tasks from coding to generating client deliverables. The results were significant, including a 90% cost reduction and up to 85% lower latency through prompt caching. Teams can now concurrently run 30+ agent tasks from a single board, transforming solo work into a collaborative process and saving companies like dbt Labs over $35k annually.

Notion

Simon Last

co-founder