Notion has launched a new AI agent, developed over two years, that transforms the user experience from a document/database interface to a conversational, chat-based one, significantly lowering the barrier to entry for new users.
The agent's capabilities include creating and managing Notion content, performing web searches, and integrating with external data sources like Slack and Google Drive to provide rich context for its actions.
Notion's strategy is to be an "applied AI company," leveraging the best available third-party models (like GPT-4) rather than training their own, focusing instead on the interface and data representation layer.
A key technical breakthrough was representing Notion pages as Markdown for the LLM, creating a clean "Agent Computer Interface" (ACI) that improves reliability and performance.
12 quotes
Concerns Raised
High token costs for custom agents are forcing a re-evaluation of the current seat-based pricing model.
Initial attempts to build agents on custom-trained models failed, highlighting the difficulty of competing with state-of-the-art frontier models.
Opportunities Identified
The new chat-based interface significantly lowers the barrier to adoption for Notion.
Custom, autonomous agents have the potential to unlock massive productivity gains for teams.
Integrating more enterprise data sources like Salesforce will increase the agent's utility and context-awareness.