Whatnot, a hyper-growth live-shopping platform processing billions of daily events, shared how they tackled data observability and analytics at scale using Snowflake. They evolved from a centralized dbt setup to a decentralized, modular data stack with infrastructure-as-code, enabling individual business units to manage their own Snowflake warehouses. To democratize data access, they progressed through three phases of AI analytics — from a rigid Slack SQL bot to Snowflake Cortex Agents powering conversational analytics via Hex Threads — achieving 80%+ employee adoption within 90 days. On the monitoring side, Snowflake's next-gen event tables (via Snowflake Trail) made log ingestion 10x faster, and AI-assisted alert creation replaced hundreds of lines of SQL with natural language prompts. Whatnot also enforces strict epistemic hygiene rules for AI agents, prohibiting causal claims and requiring probabilistic language to maintain data trustworthiness.
