An interview with Notion CEO Ivan Zhao on Custom Notion AI agents launching in the coming week, over 50% of Notion databases now being built by agents, and more
CEO Ivan Zhao on the company's big custom agents launch, changing its business model for AI, hiring a 16-year-old engineer, and more.
Context & Ripple Effects
Notion’s custom-agent launch extends its earlier move into customizable background agents that could create documents and take actions. It also builds on Zhao’s prior positioning of Notion as an AI workspace, including discussion of usage-based AI pricing.
The reported share of databases built by agents suggests the product is moving beyond AI assistance inside existing pages toward agent-led creation of the workspace itself. That makes the accompanying business-model change consequential: AI usage becomes more central to how Notion packages and monetizes work.
First-order effects
- Notion customers will be able to configure AI agents for their own workflows, while database creation increasingly shifts from manual setup to agent-generated structures.
- Notion must align its AI business model with agent usage, as custom agents can create recurring, potentially more variable demand than one-off writing assistance.
Second-order effects
- Rival work-management and collaboration platforms face pressure to offer configurable agents that act across customer-specific workflows, not merely embedded chat or drafting tools.
- Teams may revisit how they standardize databases and internal processes: reusable agent configurations can become a new layer of workflow design and administration.
Third-order effects
- If agent-built workspaces become routine, collaboration software may compete less on blank-canvas authoring and more on the reliability, control, and monetization of agents that maintain operating workflows.
- The shift raises the importance of governance around agent permissions and outputs, particularly as agents move from helping users create content to creating shared organizational structures.
The trend: This is part of the shift from AI features inside productivity software to configurable agents that construct and operate the work surface itself.