Snowflake announced a set of new AI capabilities at its Expedition 2026 virtual conference on November 3, targeting a problem common in enterprise AI rollouts: AI agents that give conflicting answers because they are working from different, unapproved definitions of core business terms.
The centerpiece is Business Ontology, a framework inside Snowflake's Horizon Catalog that lets organizations create a single, governed definition for concepts such as revenue or customer churn and link those definitions to the underlying data tables. Snowflake's CoCo coding agent and CoWork knowledge-worker agent will draw on those approved definitions when generating responses. The company also expanded Cortex AI Gateway, which applies security and governance policies to AI agents — including controls over which tools an agent can call, spending limits that can be set by team or agent, and an audit log of agent actions. Cortex AI Gateway also collects metadata from common business intelligence tools including Tableau, Power BI, dbt, SQL Server, and PostgreSQL. Business Ontology is currently in private preview, with public preview expected soon; some Cortex AI Gateway features remain in private preview or development.
For organizations already running Snowflake as their data platform, these tools address a governance gap that tends to widen as agent deployments scale: without a shared definition layer, different agents can produce incompatible outputs from the same underlying data. Snowflake users evaluating agentic AI pilots may want to monitor the public preview release date for Business Ontology, review the Cortex AI Gateway's budget-enforcement controls, and assess whether their existing business glossaries can be imported to seed the ontology.
Sources
Does this affect your business?
Our advisors can help you figure out the right move, at no cost.