Snowflake ships Business Ontology and agent governance
Snowflake announced a suite of agentic AI platform updates on November 3, 2026, during its Expedition virtual conference, designed to move enterprises from agent pilots to production deployments with governance guardrails built in.
Business Ontology: Approved Definitions at Scale
The centerpiece is Business Ontology, a governance layer within Snowflake Horizon Context and Snowflake Horizon Catalog Snowflake. The system records approved business-term definitions, the person or entity that approved them, and the data asset used to compute each term. These definitions can then be linked to semantic views, tables, and other data structures, ensuring all agents reference the same authoritative source of truth.
Business Ontology is available in public preview, according to Snowflake's official description Snowflake Investors.
Grounding Agents in Trusted Context
Approved definitions flow directly to Snowflake's named agents. Snowflake CoCo, described as a coding agent for data teams, and Snowflake CoWork, positioned as a personal work agent for knowledge workers, receive these definitions to ensure their responses use only authoritative business language and calculations Silicon Angle. This grounding mechanism aims to reduce hallucination and ensure compliance with organizational definitions.
Cortex AI Gateway: Agent Controls and Observability
Snowflake expanded its Cortex AI Gateway to add governance across models, tools, and agent actions Enterprise Times. The gateway now includes agent-abuse detection, observability and tracing, sandboxes, AI-cost governance, dynamic model routing, and tool integration capabilities. These controls allow security and operations teams to monitor agent behavior, limit costs, and isolate risky actions.
Autonomous Data Engineering
The announcement also included capabilities intended to let AI agents build, monitor, and troubleshoot data pipelines with less human intervention, expanding the scope of autonomous work beyond analytics and customer-facing tasks VenSearch. This positions agents deeper into the data infrastructure layer.
Why It Matters
The release reflects growing enterprise demand for guardrails around agentic systems. By bundling governance, grounding, and autonomous capabilities, Snowflake is attempting to address three pain points: ensuring agents use correct business logic, keeping costs and behavior visible, and automating infrastructure work that traditionally requires human engineers. The mix of developer tooling and compliance controls suggests the vendor is targeting both data teams and enterprise governance functions.
