A practical guide to building agentic BI applications on Snowflake using Cortex Agents, Cortex Analyst, and Semantic Views. Covers the full development workflow from data preparation and semantic view design to agent configuration, monitoring, cost tracking, and security. Key recommendations include starting with 'golden questions,' keeping agent scope narrow, investing in verified queries, using Cortex Search for high-cardinality fields, and enforcing governance through Snowflake's existing RBAC and data masking rather than prompt-level controls. Includes concrete SQL examples for querying AI observability events and usage history.
Table of contents
Key Snowflake Building Blocks for Agentic BI on Structured DataCore Design Principles for Cortex Agents: Plan AheadAgent Development WorkflowCore Design Principles for Semantic ViewsGet Daria Rostovtseva’s stories in your inboxMonitoring agent executionMonitoring Agent CostSecurity and governance for Cortex AgentsConclusionSort: