Snowflake Inc. is using its Summit 2026 conference today in San Francisco to present a vision of what it calls the “agentic enterprise,” unveiling a broad set of products and enhancements that it says help organizations build, govern and operate artificial intelligence systems on top of trusted enterprise data.

The announcements span data interoperability, AI governance, security, application development and knowledge-worker productivity, all tied together by a growing portfolio of AI agents.

AI’s long-term value depends less on foundation models and more on the ability to connect those models to enterprise data, business context and operational systems, Christian Kleinerman, Snowflake’s executive vice president of product, said in a briefing,

“Models keep changing, and capabilities keep advancing, but the data is constant,” he said. “We are helping our customers combine their data with whatever model is leading in the industry and provide governance for both.”

Iceberg v3 support

Snowflake is introducing a new framework for interoperable data and AI that aims to eliminate the silos and duplication that often complicate enterprise AI projects. A major element is expanded support for Apache Iceberg, including general availability support for Iceberg v3 and new Snowflake-managed storage for Iceberg tables.

The company is also extending Horizon Catalog, a governance, security and discovery service built into the Snowflake AI Data Cloud. Horizon Catalog integrates the Apache Polaris open-source catalog for Iceberg tables and extends it with bidirectional interoperability between Snowflake-managed data and external engines. Snowflake said the goal is to allow organizations to work from a single governed copy of data across its own Data Cloud, data lakes and third-party platforms without moving or replicating information.

Kleinerman said the enhancements recognize the reality that openness and interoperability have become strategic priorities for customers concerned about lock-in. “Snowflake is fully committed to interoperability, and we want to make sure that customers can mix and match if they choose to do so,” he said.

Focus on governance

Governance is another major theme. Snowflake said that consistent business context and centralized controls are essential as enterprises look to deploy autonomous agents at large scale. Expanded Horizon Catalog capabilities capture semantic information, business definitions and metadata from multiple systems and make that information available to AI applications.

The enhancements leverage technology acquired through Snowflake’s purchase of Select Star Inc. and are intended to help AI systems reason from consistent business definitions rather than fragmented data sources.

Horizon Context, which is the metadata, business definitions, security policies and data lineage that Horizon associates with data, “is the part of the Horizon Catalog that brings metadata and context into Snowflake to help AI produce better results,” Kleinerman said.

Citing recent McKinsey & Co. research that found that security and risk concerns are the top barrier to scaling agentic AI, the company is unveiling several new security capabilities aimed specifically at AI agents. Agent Identity gives each AI agent a distinct identity and permissions model. AI Security Posture Management, part of the Snowflake Trust Center, helps continuously monitor the security posture of AI systems. New centralized governance features prevent data exfiltration and enhance ransomware protection.

The expanded security features recognize that traditional security models were designed for human users, not autonomous software agents capable of accessing data and taking actions independently, Snowflake said. Its approach applies zero-trust principles to AI environments while maintaining centralized auditing and governance.

Agent Identity requires every agent to have a verified identity before it can access enterprise data or take action. “If I create an agent, by default the agent runs with my credentials and my permissions,” Kleinerman said. “What we’re doing with Agent Identity is enabling that separate identity to be constrained and restricted.”

New for developers

Snowflake is also expanding its AI development platform with significant enhancements to Cortex Code, a coding agent for developers, data engineers and technical users. Renamed CoCo, the agent now supports desktop and mobile applications, integration with Salesforce Inc.’s Slack, Excel, VS Code and Anthropic PBC’s Claude Code. Autonomous workflow execution is now supported through native AI services called Cloud Agents and Automations, and a new Skill Catalog allows for sharing of reusable workflows.

A significant addition is Snowflake Datastream, a fully managed Apache Kafka-compatible streaming service integrated into the Snowflake platform. Datastream enables organizations to stream data into Snowflake without maintaining separate Kafka infrastructure while preserving Snowflake’s governance, security and management capabilities. It supports real-time data pipelines that feed AI applications and agents with continuously updated information.

Kleinerman described Datastream as a response to customer frustration with operational complexity. They have often asked Snowflake to provide “a Kafka-type of streaming solution, but in the Snowflake way,” he said. That means “managed infrastructure, with automatic scaling, governance and security embedded deeply in it.”

Personal agents

For business users, Snowflake has renamed Snowflake Intelligence “Snowflake CoWork” and expanded its capabilities to function as a personal AI work agent. CoWork is designed for nontechnical knowledge workers and incorporates personalized skills, memory, automation and integrations with enterprise applications through the open-source Model Context Protocol.

New CoWork capabilities include Deep Research, which can search structured and unstructured enterprise data using multistep reasoning. Artifacts enable interactive dashboards and reusable insights. User Memory personalizes responses based on user behavior. Cortex Sense is a new shared context layer that provides AI agents with access to business definitions, operational knowledge and enterprise context.

Kleinerman characterized CoWork as part of Snowflake’s broader effort to become “the operating system for the enterprise intelligence,” combining data, AI models and governance into a single platform.

Finally, Snowflake announced Cortex Training, which extends its Cortex AI portfolio to include model customization and training. Cortex Training provides managed graphics processing unit infrastructure that allows organizations to fine-tune open-source models such as Qwen and Mistral using proprietary data without leaving the Snowflake environment.

The service is intended to simplify model development, reduce infrastructure complexity and improve governance. Customers can train domain-specific models on their data while avoiding the need to manage large GPU clusters.

Snowflake said the announcements further its evolution from a cloud data warehouse provider to a broader AI platform. Rather than competing directly to build foundation models, the company is positioning itself as the layer that connects models, enterprise data, business context and governance.

The company said its message at Summit 2026 is that successful enterprise AI depends less on which model organizations choose and more on whether they can provide those models with trusted data, consistent context and secure operational controls. It’s betting that as AI agents become increasingly autonomous, governance, interoperability and enterprise context will be critical differentiators.

Image: Robert Hof/SiliconANGLE