Horizon Context: Put Business Knowledge to Work for AI

Snowflake’s new Horizon Context addresses a common AI challenge: agents need clear, consistent definitions of business terms like customer retention to deliver accurate answers. Without shared context, teams waste time reassembling scattered knowledge from reports, documentation, and expert experience for each new AI use case. Horizon Context aggregates metadata from 18 sources—including Power BI, Tableau, and dbt—into a governed layer, enriching it with business meaning to make data usable for people, BI tools, and AI agents.

The redesigned Horizon Catalog Explorer now provides a single interface to discover, enrich, and reuse data knowledge across Snowflake and connected systems. Business analysts can view data definitions, tags, and contacts, while AI workflows like Snowflake CoCo leverage lineage to trace data origins and assess change impacts. Metadata visibility allows exploration without granting access to underlying data, and new permissions like APPLY COMMENT enable stewards to curate descriptions without broad privileges.

Column description propagation automates the reuse of business definitions, carrying descriptions through unchanged transformations to downstream columns. For example, a customer identifier’s description can follow into a renamed view column, reducing repetitive maintenance while preserving owner control. Propagation stops when values are transformed, ensuring descriptions remain accurate and context-specific.

Horizon Context integrates three pillars: evidence from data systems, expert knowledge, and tools like CoCo and semantic views. This foundation supports consistent interpretation of AI-generated answers, enabling teams to investigate issues like retention declines by tracing data lineage and verifying governed metric definitions. The goal is to shift focus from reconstructing context to applying it to business decisions.