Snowflake used its annual Summit conference in San Francisco to unveil a broad set of AI updates aimed at making enterprise adoption easier while tightening controls around data and security. Among the announcements was Adaptive Compute, a new capability in Horizon Catalog designed to automatically optimize infrastructure usage in real time.

The company said the new feature is part of a larger effort to help businesses run AI applications with less operational overhead and fewer security concerns. Snowflake framed the updates as a way to reduce the friction that often slows enterprise AI projects, especially when companies need to balance access to data with governance requirements.

## A push to make AI easier to use

One of the main products highlighted at the event was Snowflake CoWork, the renamed version of Snowflake Intelligence. The company described it as a personal AI agent for knowledge workers that brings together data, context and tools in one place. New capabilities include Cortex Sense, which combines business definitions, operational knowledge and data; User Memory, which lets people schedule actions through tools such as Gmail and Slack; and a Skill Catalog for discovering and reusing agent skills.

Snowflake also introduced updates to its Horizon Catalog, its governance and security hub. Horizon Context is meant to ensure that employees, tools and AI agents work from the same business definitions. Adaptive Compute sits inside this framework and is intended to reduce wasted resources by adjusting compute dynamically as demand changes.

The company said the catalog now includes additional security protections as well, such as verified agent identities, continuous monitoring of security posture, and automated defenses against jailbreak attempts and zero-day vulnerabilities.

## New automation for developers and data access

Snowflake also unveiled Snowflake CoCo, the renamed version of its coding assistant previously known as Cortex Code. The assistant can now operate more autonomously, reducing the need for users to remain active on screen while tasks are completed.

Another announcement, Datastream, is aimed at simplifying data pipelines for AI applications. Snowflake said the service removes the need for brokers, connectors or separate streaming infrastructure, allowing applications to keep access to fresh data without additional plumbing.

At a private media roundtable, Christian Kleinerman, Snowflake’s executive vice president of product, said the company’s approach is built around helping organizations become more productive while maintaining security, compliance and governance.

## Anthropic partnership adds momentum

Snowflake also highlighted progress in its partnership with Anthropic, saying Claude models are gaining traction on Snowflake Cortex AI, the company’s AI product suite. In the conference keynote, Anthropic President Daniela Amodei said trust is a key reason the company focuses on enterprise customers and partners with Snowflake.

The partnership appears intended to strengthen both companies’ positions in the enterprise AI market. For Snowflake, it adds credibility to its AI lineup. For Anthropic, it offers another path into businesses that want access to frontier models without sacrificing controls.

Snowflake’s latest announcements point to a broader strategy: make AI tools easier to deploy, easier to govern and less dependent on custom infrastructure. The company is betting that enterprises will adopt AI more quickly if it can be delivered in a way that fits existing data systems and security expectations.