Snowflake Inc. Unveils Comprehensive AI Capabilities at Expedition 2026
Snowflake Inc. has announced a broad suite of artificial intelligence capabilities aimed at integrating business context, governance, and autonomous development within a unified environment. This initiative responds to the growing trend of enterprises delegating software engineering and operational tasks to AI agents.
During the company’s Expedition 2026 virtual conference, Snowflake introduced a new Business Ontology framework, enhanced agent governance through Cortex AI Gateway, and autonomous data engineering functionalities. These features empower AI agents to construct, monitor, and troubleshoot data pipelines with minimal human intervention.
The company anticipates that an increasing number of developers will incorporate AI throughout the software development lifecycle as the pace of creation accelerates. Moreover, back-end departments are beginning to deploy AI agents in their daily operations. For instance, sales teams are developing agents capable of analyzing sales risks and preparing customer outreach without requiring code.
“AI is industrializing software creation. Actually, it’s industrializing intelligence,” stated Chief Executive Officer Sridhar Ramaswamy. “What once took months can now take hours, days, or even minutes.” Ramaswamy emphasized that, as AI transforms the economics of software development, traditional tasks typically performed by specialized developers are being assumed by business users, thus shifting the competitive landscape. The competitive advantage now arises from proprietary knowledge rather than specialized skill sets.
Snowflake is focused on enhancing the capability of standard AI models to comprehend and contextualize business knowledge effectively. The Business Ontology extends the company’s existing semantic views, enabling organizations to establish approved definitions used consistently across the enterprise. This standardization prevents discrepancies in responses from different AI agents, even when leveraging accurate underlying data.
Utilizing data from Business Ontology, Snowflake CoCo, the coding agent designed for data teams, and Snowflake CoWork, the personal work agent for knowledge workers, are built to provide grounded responses based on collectively established business concepts. This feature is currently in private preview but is expected to transition to public preview soon.
Governance Framework for AI Agents
The newly introduced Cortex AI Gateway serves as a centralized access point for managing enterprise AI agents. Although it was initially launched in July 2026 to provide centralized governance and monitoring capabilities, today’s announcements expand its functionalities. Enhancements include agent abuse detection, observability and tracing, Cortex sandboxes, AI cost governance, dynamic model routing, and integrations utilizing technology from Snowflake’s recent Natoma acquisition.
“An agent that acts on your business must be trusted, and it has to understand three critical things: what the data signifies, which data to trust, and what it is permitted to do,” Ramaswamy explained. This governance framework applies existing security measures, such as data masking and access restrictions, to agent-driven operations while introducing additional monitoring capabilities, abuse detection, and activity tracking. IT and security teams can implement cost controls, select models based on efficiency or accuracy, and finely tune agent access to external tools.
This approach emphasizes the importance of governance, control, and transparency, positioning agents with a level of oversight akin to that of employees and affording them comparable autonomy.
Snowflake’s advancements occur alongside a wider industry push to create a control layer for enterprise AI agents. Competing companies, such as SAP and Salesforce, have similarly been strengthening their frameworks for autonomous workflows and agent governance. Snowflake aims to serve as the foundational platform through which agents can access vetted business information and operate within established enterprise policies.
Tailoring AI Models for Specialized Applications
In addition to its governance initiatives, Snowflake announced the public preview of Cortex Training, enabling customers to customize open-weight AI models with proprietary data. This service simplifies the process of creating specialized models using post-training techniques without necessitating extensive graphic processing unit resources. Customers can train models effortlessly without having to manage the underlying infrastructure.
“Specialized, smaller models deliver more efficient and quicker responses, often with cost benefits concerning token usage,” said Christian Kleinerman, Executive Vice President of Product. With growing restrictions on foreign open-weight AI and increasing regulations surrounding sensitive data handling, many enterprises seek sovereign control over their AI models. This method allows for easier auditing of what traverses AI models hosted on personal hardware as opposed to third-party vendor clouds.
Furthermore, Snowflake announced an expanded partnership with Oracle Corp., making its platform available as a first-party service on Oracle Cloud Infrastructure. This integration will facilitate customer access to Snowflake’s analytics and AI capabilities while allowing seamless operation with Oracle databases and applications.