At its annual Summit conference in San Francisco, Snowflake unveiled platform updates to help enterprises move AI from pilot projects into operational systems, with new capabilities to ensure AI agents follow governance policies and deliver consistent results across fragmented environments.
Snowflake is positioning itself as the connective layer between enterprise data, AI models and operational controls. Rather than competing to build AI foundation models, the company is building infrastructure to manage AI systems that interact with corporate applications and business data.
“Snowflake has definitively moved beyond its origins as a place where you can park your data,” Brad Shimmin, Futurum VP and Practice Lead for Data Intelligence, Analytics and Infrastructure, told Techstrong.it. “I would place them squarely within the platform provider realm, as we can see with their ongoing push into agentic AI.
“To that end, they are building a very real and comprehensive intelligence layer to ground both everyday analytics and AI agents in truth. That’s ambitious and bold, but it will take some time for Snowflake to realize its vision here.”
In the meantime, to fend off rivals like Databricks and others, “Snowflake has to keep doing what made them famous in the first place, namely the relentless destruction of operational friction for their customers.”
Context as Core Element
An essential theme running through the announcements is context management. As enterprises deploy armies of AI agents, inconsistencies in business definitions produce conflicting answers from the same underlying data. A wide array of variables may be defined differently across dashboards and AI applications, resulting in poor agent performance.
To address that, Snowflake expanded Horizon Context and introduced Cortex Sense. Horizon Context captures business definitions, metadata, lineage information and security policies from numerous systems. Cortex Sense automatically derives additional context from data usage patterns and operational activity.
Together, the technologies are intended to provide AI systems with a shared understanding of business concepts and governance rules. Industry analysts increasingly view this context layer as a critical component of enterprise AI deployments because agents are only as reliable as the data definitions behind them.
“I applaud the company’s efforts to remove friction and complexity for both data professionals and business users,” said Futurum’s Shimmin.
“But the catch here for me is that in their rush to make deployment easy for users, they risk bloating their own portfolio with engineering complexity. You can see this tension in how they’ve built a bit of a bifurcation into the intelligence layer between Horizon Context and Cortex Sense. It makes sense on a whiteboard, but it does ask buyers to piece together a slightly larger puzzle.”
Governance, Security, Open Data Strategy
To support security, Snowflake debuted Agent Identity, which assigns individual identities and permissions to AI agents instead of allowing them to inherit human user credentials. The company also unveiled AI Security Posture Management capabilities and additional governance controls designed to monitor AI systems and reduce risks like unauthorized data access.
Snowflake also expanded its AI development platform. Its coding assistant, formerly known as Cortex Code and now dubbed CoCo, gained integrations with tools including Slack, Excel, VS Code and Anthropic’s Claude Code. New workflow automation capabilities enable organizations to create and share reusable AI-powered processes.
Another new release is Snowflake Datastream, a managed streaming service compatible with Apache Kafka. The service feeds AI applications with continuously updated data without requiring organizations to operate separate Kafka infrastructure.
For business users, Snowflake rebranded Snowflake Intelligence as Snowflake CoWork and expanded its capabilities as a personal AI work assistant. New features include multi-step research capabilities, user memory, automation functions and access to enterprise knowledge through integrations based on MCP.
Snowflake expanded support for Apache Iceberg, the open table format that has gained traction among enterprises seeking greater flexibility across cloud and analytics platforms. Snowflake announced general availability support for Iceberg v3, along with Snowflake-managed storage for Iceberg tables.
Beyond its own platform announcements, the company highlighted growing adoption of Anthropic’s Claude models through Snowflake Cortex AI. Companies including Block, Carvana, Indeed and Notion were highlighted as customers adopting Anthropic’s Claude models through Cortex AI.
