Published: November 4, 2026 | News development: November 3, 2026
Snowflake has announced new enterprise AI capabilities designed to make autonomous AI agents more reliable, secure and easier to govern. The updates, unveiled at Snowflake Expedition 2026 on November 3, focus on two persistent enterprise problems: agents that misunderstand business data and agents that can access systems without consistent oversight.
Business Ontology aims to give AI agents a shared definition of business data
Snowflake’s Business Ontology is designed to provide approved definitions for important business terms and connect those definitions to the underlying data. That matters when different departments calculate or interpret metrics such as net revenue, customer segments or operational performance in different ways.
By linking approved definitions to tables and views, the company says its agents can use a more consistent business context instead of relying only on what they infer from disconnected data sources. Snowflake says the capability is being introduced through Horizon Context in Snowflake Horizon Catalog. The announcement describes Business Ontology as a preview-stage capability, so organizations should confirm availability and eligibility with Snowflake before planning a production rollout.
Cortex AI Gateway adds centralized controls for enterprise AI agents
Snowflake also highlighted Cortex AI Gateway as a control point for agent connections across models, tools, clouds and enterprise systems. The gateway is intended to apply authentication, permissions and policy checks when agents access company data or call tools through interfaces such as Model Context Protocol (MCP) servers.
Other capabilities described in the announcement include agent activity observability, cost attribution, budget limits, model routing and detection of potentially abusive agent behaviour. Availability differs by feature: Snowflake’s release notes distinguish generally available capabilities from public-preview, private-preview and in-development features. Buyers should check the current product documentation rather than assume every announced capability is generally available.
Why this matters to enterprise AI buyers
For companies evaluating enterprise AI platforms, the announcement reflects a shift from testing chatbots to operating AI agents inside real business workflows. Access to data alone is not enough. Teams also need consistent definitions, least-privilege access, audit trails, cost controls and a way to investigate what an agent did.
- Data and analytics teams: governed definitions can reduce disagreement between dashboards, analysts and AI-generated answers.
- IT and security leaders: centralized policy enforcement can make it easier to review agent access and tool usage.
- Finance and procurement teams: usage attribution and model routing can help track and manage AI software spending.
What organizations should evaluate before deployment
Before selecting an AI agent platform, buyers should test it against their own data definitions, identity provider, cloud architecture and compliance requirements. Ask which controls are generally available, how agent actions are logged, whether policies apply consistently across third-party tools, and how usage costs are measured.
For more technology coverage, explore NewsTree24 for updates on AI tools, enterprise software and data analytics.
Source
Snowflake: Business Ontology and Cortex AI Gateway announcement, November 3, 2026.
Featured image: illustrative enterprise analytics image by Luke Chesser via Unsplash.