How we built a personal data scientist for every Notion employee

This post details the architecture and design choices behind Data Scout, an internal data agent that connects governed data in Snowflake with company knowledge in Notion and Slack. It highlights the challenges of ensuring data meaning and accuracy beyond just query execution, emphasizing the importance of curated context. The system relies on governed agent and data access, shared data context (using Notion databases for definitions, schemas, and business context), and clear instructions for routing and execution. It also discusses the trade-offs in storing context (database properties vs. page bodies) and the governance model involving Notion for workspace access and Snowflake for data permissions.