Notion says users built more than one million Custom Agents in the two months after their launch, a striking measure of how the workspace is shifting from storing work to delegating it. The agents can run in the background across team pages, databases and connected tools, with companies including Ramp and Clay already using them at scale.
Watch Desk analysis
What happened
AI Magazine reports that Ramp built more than 300 agents and Clay rolled out more than 80 company-wide within a week. Notion’s Developer Platform also lets teams assign work to outside agents, including Claude Code, Cursor and Decagon, from inside the workspace. Its Workers platform can run custom code and sync data from systems such as Salesforce, Zendesk and Postgres.
The commercial model is changing too. Custom Agents moved out of beta in May with usage-based pricing of US$10 per 1,000 credits, giving Notion a consumption-based revenue line alongside its subscriptions. The AI Magazine report also says more than half of Notion’s year-end 2025 annual recurring revenue came from AI-enabled customers, citing an estimate from Sacra.
Why it matters
The practical bet is that teams will want one place to assign and track work whether a person or an AI agent does it. If agents can draw on live business data and hand work between tools, the workspace could become more than a document hub: it could be the control room for delegated tasks.
The million-agent figure measures what users built, not how often those agents run or how much useful work they complete. Still, the combination of user-built agents, outside integrations and usage pricing shows Notion is building both a platform and a business around delegation.
Our read
This is a meaningful product and business shift, not just a new button labelled “AI”. Notion is trying to make agents part of ordinary team workflows, while charging for usage rather than simply adding another seat. The test is whether those agents reliably save people time once they leave the launch-partner stage and meet the messy reality of company data.
What to watch
- Whether Notion shares figures on active agents and the work they complete.
- How usage-based credits affect customer costs as teams scale up.
- Whether integrations and external agents become dependable parts of everyday workflows.
Discussion spark: Should workplaces organise AI agents inside the same workspace as staff, or does that make it too easy to delegate work before anyone has proved the agents are useful?
Sources and evidence
- Notion: AI Agents for a Workspace Used by 100 Million People – AI Magazine (6 November 2026, 09:02 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.