Snowflake expands agentic AI platform with business context and governance tools

Snowflake Inc. unveiled an extensive array of artificial intelligence capabilities on Monday, aimed at integrating business context, governance, and autonomous development within a unified platform. This announcement came during the company's Expedition 2026 virtual conference. The new tools include a Business Ontology framework, enhanced agent governance through Cortex AI Gateway, and autonomous data engineering features enabling AI agents to construct, supervise, and troubleshoot data pipelines with minimal human involvement.

Snowflake's strategy hinges on the notion that more developers and engineers will leverage AI throughout the software development lifecycle as the pace of creation accelerates. Additionally, back-end departments are expected to adopt AI agents for routine tasks. For instance, sales operations are now creating agents capable of assessing sales risks and preparing customer outreach without any coding.

During the keynote, Chief Executive Sridhar Ramaswamy emphasized that AI is transforming software creation, stating that what once took months can now be accomplished in hours, days, or even minutes. He noted that the economic landscape of software development is shifting, as AI agents replace specialized developers, thereby granting business users the role of developers. This shift in roles is altering the competitive landscape, as proprietary knowledge becomes more valuable than specialized expertise.

To address the challenge of ensuring that AI models comprehend and contextualize business knowledge, Snowflake introduced Business Ontology. This extension of the company's existing semantic views enables organizations to establish approved definitions that are consistently applied across the enterprise. The Business Ontology aids in preventing discrepancies in agent responses to business-related queries, even when both agents possess accurate underlying data. Currently in private preview, this feature will transition to public preview shortly.

Cortex AI Gateway, which facilitates single access for agents within an organization, was introduced in July 2026 as a centralized governance and monitoring solution. Today's announcement expands its functionalities, including agent abuse detection, observability and tracing, Cortex sandboxes, AI cost governance, dynamic model routing, and tool integration. These enhancements are made possible through technology from Snowflake's recent Natoma acquisition.

Ramaswamy stressed that an AI agent interacting with a business must be trusted and possess knowledge of the data's meaning, the reliability of the data, and the permissions it holds. The platform ensures that existing security mechanisms, such as data masking and access restrictions, are applied to agent-driven operations while adding capabilities for monitoring activity, detecting potential abuse, and recording agent actions.

IT and security teams can also implement cost controls, select models based on efficiency or accuracy, and restrict the external tools available to agents.

The company's emphasis on governance, control, and transparency aligns with the goal of granting agents the same level of observation and autonomy typically afforded to employees. This approach addresses the need for control and accountability in enterprise AI agent operations.

Snowflake's initiative comes as other industry leaders are also racing to establish governance layers for enterprise AI agents. SAP is expanding its Autonomous Enterprise architecture and governed Joule agents, while Salesforce has introduced Enterprise AI Harness and AI Control Plane to secure autonomous workflows. By approaching the issue from the data platform perspective, Snowflake aims to serve as the foundation that enables agents to access trusted business information and operate within enterprise policies.

In addition to the announced tools, Snowflake unveiled the public preview of Cortex Training. This service allows customers to fine-tune open-weight AI models with their proprietary information without requiring extensive processing power. By allowing the training of specialized models on their own hardware, Snowflake enables enterprises to achieve faster response times and cost savings in token usage.

This move is particularly significant for enterprises that require stringent control over their AI models due to increasing restrictions on foreign open-weight AI or regulations governing sensitive data handling.

Furthermore, Snowflake announced an expanded partnership with Oracle Corp., allowing Snowflake's platform to be offered as a first-party service on Oracle Cloud Infrastructure. This integration provides customers with seamless access to Snowflake's analytics and AI capabilities while enabling direct interaction with Oracle databases and applications. The collaboration aims to enhance the overall user experience and streamline the deployment of Snowflake's AI solutions within Oracle's ecosystem.