Introducing Cortex Sense: Grounded Context for the Data You Never Modeled
Snowflake, Tuesday, June 30th, 2026
Snowflake unveils Cortex Sense, which auto-builds contextual understanding of enterprise data without manual documentation.
Cortex Sense operates alongside semantic views to help AI agents understand undocumented data by automatically learning from existing business signals like past queries, transformation models, and BI metrics.
Rather than relying solely on hand-curated metadata, it uses a self-correcting loop that continually identifies gaps and conflicts to improve accuracy over time. Internal testing showed it increased query accuracy from roughly 25% to over 86%, while reducing per-query costs by two-thirds compared to frontier models.
It targets enterprise AI agents that need reliable, grounded context.