Snowflake launches Horizon Context to give AI business meaning

The announcement addresses a familiar gap in enterprise AI: a model can produce a fluent answer, but if it starts from the wrong definition or data set, the answer is not useful. Snowflake's premise is that much of the needed knowledge already exists in reports, transformation logic, documentation, and team experience, yet stays scattered. Horizon Context is meant to gather that evidence and attach business meaning to it once, rather than reassembling it for every new AI use case.

The building blocks are metadata connectors for 18 sources, a redesigned Horizon Catalog Explorer, and targeted permissions so business experts can curate descriptions without gaining access to the underlying data. Column description propagation carries descriptions to supported downstream columns when values pass through unchanged, and stops when values are transformed, combined, or aggregated. Snowflake CoCo can then use catalog context for tasks such as data discovery and impact analysis.

The business value Snowflake emphasizes is reuse: teams build on shared knowledge rather than recreating it in each dashboard, application, or prompt. That foundation supports more consistent interpretation and gives people evidence to examine when reviewing an AI-generated answer, though the extent to which it reduces inconsistency across agents is not quantified in the announcement.