Skip to content

Snowflake Expands AI Agent Governance With Business Context and Controls

Snowflake expands AI agent governance with business context and controls news banner
Less than 1 minute Minutes
Less than 1 minute Minutes

Snowflake has announced new capabilities to help enterprises deploy AI agents with clearer business context, stronger governance, and greater visibility into costs. Unveiled at Snowflake Expedition on November 3, 2026, the updates extend Snowflake Horizon Catalog through Business Ontology and Cortex AI Gateway, addressing practical obstacles that can keep agentic AI projects from reaching production.

The announcement highlights a persistent challenge: agents need reliable business meaning alongside access to data. According to a Snowflake-sponsored MIT Technology Review Insights survey of 400 senior technology executives, 55% identified insufficient business context as the leading cause of poor agent insights, while 45% pointed to fragmented or inaccessible data.

Trusted Business Context for AI Agents

Business Ontology is designed to give organizations a governed place to define business concepts, assign owners, and approve their meaning. Teams can connect definitions to the data and metrics that implement them, helping agents interpret questions consistently. For example, an agent preparing a revenue report would use the approved finance definition instead of selecting among conflicting interpretations.

The capability sits within Horizon Context and works alongside Semantic Views, which calculate metrics, and Cortex Sense, which infers context. Human-approved definitions take precedence when inferred information conflicts with them. Existing glossaries can also provide a starting point, with imported definitions submitted for owner approval. Recorded changes also help teams trace how definitions evolved and who approved them at a particular time. This supports investigation when reports or agent responses are challenged.

Centralized Governance and Cost Control

Cortex AI Gateway addresses how agents interact with models and enterprise tools. It centralizes access controls and monitoring, allowing teams to apply established permissions to AI activity. Cost governance attributes spending to users, teams, or applications and supports enforceable limits, while observability records activity within the customer’s Snowflake account.

Tool governance adds authorization checks before connected tools execute. Dynamic Model Routing is designed to direct requests to approved models based on factors including cost, accuracy, and latency. These controls aim to help enterprises understand which agents are acting, what they are using, and how much their work costs.

Governed Automation for Enterprise Workflows

Snowflake also connects this foundation to CoCo, its coding agent for data teams, and CoWork, its personal work agent for knowledge workers. CoWork supports scheduled workflows and actions across connected business applications while retaining user permissions, bringing governed AI into recurring operational work.

Availability varies across the announcement. Business Ontology is scheduled for public preview, Cortex AI Gateway Tools is in public preview, and Dynamic Model Routing is in private preview. Deployment plans should reflect the maturity and availability of each capability.

For enterprise leaders, the practical implication is that scaling agents requires coordinated ownership of business definitions, access policies, and operating budgets. A focused implementation can begin with one workflow, clear success measures, and agreed human approval points before expanding adoption.

Partner With Prolifics for Enterprise AI

Partner with Prolifics to prepare your Snowflake environment for enterprise AI. Prolifics combines data engineering, governance, and Snowflake implementation expertise to help organizations build trusted data foundations. Connect with our team to assess readiness, prioritize use cases, and plan an implementation aligned with your business goals.