AI Data Cloud company Snowflake has announced a major expansion of its product suite at Snowflake Expedition 2026, positioning its platform as the foundational control and governance layer for enterprise-ready AI agents. The announcements target the primary friction point preventing companies from scaling autonomous systems: enterprise trust.
55% of executives cite a lack of business context as the chief driver of low-quality AI agent insights, while 45% point to fragmented or inaccessible data, according to a recent MIT Technology Review report sponsored by Snowflake. To bridge this gap, Snowflake is combining approved enterprise context, unified governance and autonomous execution into a single, cohesive architecture. Integrated into Snowflake Horizon Catalog, Business Ontology allows organisations to assign accountable owners to core business definitions (e.g., net revenue), link definitions directly to underlying data, and log changes for auditability. Supported by Semantic Views for calculation and Cortex Sense for inferring undefined concepts, it ensures AI agents deliver consistent, approved business metrics rather than unverified approximations. Cortex AI Gateway : This acts as a centralised governance gate for all agent operations across multi-cloud environments, frameworks and AI providers. Key mechanisms include: Policy and identity inheritance: This service connects to Model Context Protocol (MCP) servers, enforcing existing human access permissions, authentication and authorisation rules on agent actions. Cortex AI guardrails and tools: Inspects tool calls against fine-grained authorisation policies and filters for prompt injections or jailbreaks. Security and observability: Records end-to-end execution steps directly into the customer’s Snowflake account for analysis via Agent Trace Intelligence, while Agent Abuse Detection monitors overall agent trajectories for abnormal behaviour. Cost control: Enforces hard spend limits, tracks compute spend at the department or agent level, and uses Dynamic Model Routing to route routine tasks to smaller, cost-optimized models. Snowflake CoCo and CoWork: Agentic apps designed to automate routine business workflows. Snowflake CoWork is a personal productivity and work agent built for knowledge workers While Snowflake CoCo is a dedicated coding agent tailored for data engineering and data teams.
Autonomous bots: Always-on background instances (CoCo Bots and CoWork Bots) that work directly inside day-to-day enterprise tools, escalating to human users only when formal approvals or decisions are required. You use AI every day. Now get your AI Quotient. Take the AIQ test.