Snowflake is expanding its enterprise artificial-intelligence platform with tools intended to give AI agents a consistent understanding of corporate data while allowing companies to control how those systems use models, tools and business information.
The company announced the capabilities Nov. 3 at Snowflake Expedition 2026, presenting them as part of an effort to make security, context and governance central to the operation of autonomous software agents rather than features added after deployment. Snowflake said its platform connects data across catalogs and cloud environments and is designed to apply governance controls across that broader estate.
Approved definitions for AI systems
One of the new capabilities, called Business Ontology, is being added to Horizon Context, which Snowflake describes as the location for governed business context within its Horizon Catalog. The tool is intended to give organizations a centralized place to record the official meaning of business terms used across their data.
That distinction matters because an AI agent can produce an answer that appears plausible while relying on an incorrect or inconsistent interpretation of a company term. Under Snowflake's approach, agents would use definitions approved by the organization. The company said the capability is designed to provide a shared understanding of core business concepts, helping agents produce answers grounded in the terminology used by the business.
“Trusted and reliable agents are hard to build.”
Snowflake attributed that difficulty less to the underlying models than to the way corporate information is distributed among systems. The company cited a recent MIT Technology Review report sponsored by Snowflake in which 55% of executives identified insufficient business context as the main reason for poor-quality insights from AI agents. The source also pointed to data that is fragmented or difficult to access as part of the challenge.
Controls for models, tools and actions
A second offering, Cortex AI Gateway, is designed to extend Snowflake's security and governance capabilities to AI agents. The gateway governs how agents access models, call tools and take actions across enterprise systems, according to the company. Snowflake also said it can help organizations select appropriately sized models as they manage their budgets.
The company said the combined capabilities are intended to keep decisions and operations secure, auditable and connected to a single set of policies. Snowflake emphasized that this approach is meant to apply even when data is stored across different clouds, catalogs or computing engines.
| Capability | Business purpose described by Snowflake |
|---|---|
| Business Ontology | Maintains approved definitions for business terms and supplies that context to agents. |
| Cortex AI Gateway | Governs model access, tool use and agent actions across enterprise systems. |
Early corporate users
Snowflake said industry companies including Glean and Kargo rely on Cortex AI Gateway to govern AI agents across their businesses. It also said dentsu and Merkle use Business Ontology to ground agents in definitions approved by their organizations. The announcement did not provide deployment volumes, financial results or specific performance measurements for those customers.
The broader business issue is the tension between expanding AI use and preserving oversight. Companies want agents to work with information held in multiple systems, but inconsistent terminology and uncontrolled access can make automated recommendations harder to verify. Snowflake's announcement focuses on addressing those issues through centralized definitions and policy controls. Its proposed framework places the emphasis on making an agent's operating context and permissions visible to the organization responsible for it.
- Context: Agents can draw on business terms approved by the organization.
- Governance: Policies are intended to cover model access, tools and actions.
- Accountability: Snowflake says agent decisions and operations can remain auditable.
Snowflake's move reflects the growing importance of infrastructure around AI agents, not just the capabilities of the models themselves. As companies connect agents to more data and enterprise applications, the ability to define terms consistently and monitor automated actions becomes part of the deployment question. The company presented Business Ontology and Cortex AI Gateway as new layers for that oversight across an organization's data environment.
