Snowflake Launches Business Ontology for AI Agents, Adds 3 Governance Tools

3 articles · Updated · InfoWorld · Nov 3

Snowflake said Business Ontology will enter public preview soon, giving enterprises a way to formally define terms such as net revenue, assign owners and link those definitions to underlying tables and views.
The tool extends Horizon Catalog and June's Horizon Context so AI agents can use approved business meanings instead of inferred metadata, aiming to reduce conflicting interpretations across systems and departments.
For developers, centrally maintained definitions should cut the work of hard-coding business logic into each agent and make it easier to reuse context across multiple applications as requirements change.
Snowflake also expanded Cortex AI Gateway with Cortex Sandboxes, Agent Trace Intelligence and Agent Abuse Detection—three controls now in private preview to isolate agent-written code, inspect execution traces and flag permission overreach.
Analysts said the approach could help CIOs scale agents more safely, but keeping definitions current and reconciling different departmental meanings of terms like customer or revenue will still require ongoing governance.