Snowflake scores AI readiness across 43,000 tables

Snowflake’s new AI Readiness Score audited roughly 43,000 tables in 30–40 minutes on a Medium warehouse in an internal test. Snowflake’s engineering post says the Cortex Code skill scores demand coverage and semantic-view completeness; in another internal schema, adding views over six of seven consumption-ready tables moved the score from 39 to 82. For teams adopting Snowflake as an agent control plane, the useful implication is that permissions alone are insufficient: agents also need explicit keys, relationships, metrics, descriptions, and verified queries before natural-language access becomes dependable.