Answer business questions with SQL grounded in a semantic model
Define tables, metrics and synonyms once in a semantic model, and let business users ask questions that turn into checked SQL.
The use case
Text-to-SQL fails when the model guesses what 'active customer' or 'net revenue' means. Snowflake Cortex Analyst answers natural-language questions against a semantic model the data team writes, listing tables, joins, metric definitions, synonyms and verified example queries. Answers come back with the SQL so analysts can audit them. The data team owns the semantic model and adds verified queries for common questions. Users should treat answers to new kinds of questions as drafts, and analysts review the SQL before numbers go into board decks or financial reports.
Example workflow
- Write a semantic model for one subject area
- Add verified example queries for common questions
- Expose it through a chat app or API
- Users ask questions and see the SQL with results
- Data team reviews logs and adds missing definitions
Tools
- Snowflake Cortex Analyst
- Snowflake
Risks and guardrails
- Undefined metrics get guessed; cover key terms in the model
- Row-level security must be enforced in the warehouse
- Numbers for financial reporting need analyst review
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Snowflake docs: Cortex Analyst: docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-analyst
- Snowflake MCP server (official page): docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-agents-mcp (checked 2026-11-09)