Semantic views put business metrics and entities in the database

Semantic Views hold business concepts in the database as schema-level metadata. Teams define metrics and model entities and relationships so physical columns gain shared business meaning. Views can be created with Semantic Studio, SQL such as CREATE SEMANTIC VIEW, or a Snowsight wizard; they can be queried in SELECT, used with Cortex Agents, and shared in private, Marketplace, or organizational listings.

The piece frames the problem as mismatched business language versus storage names—for example gross revenue stored as amt_ttl_pre_dsc—and inconsistent formulas for concepts like net revenue. A semantic view can fix net revenue as SUM(gross_revenue * (1 - discount)) with correct aggregation when users ask for net revenue by region. Without that layer, reports and apps often multiply conflicting calculations and wrong aggregation grain.

For data, content, and AI work, the abstraction is meant to improve BI consistency and AI accuracy by pairing LLM reasoning with rule-based definitions; Cortex Agents read the view definition and generate SQL against physical tables. That is a platform-local semantic layer: one place for business logic that agents and analysts can share instead of re-encoding meaning in each tool.