Snowflake has announced the public preview of Cortex Agents, a set of APIs built on top of the Snowflake Intelligence platform, which was first launched in November at the company's annual developer conference, Build. According to Baris Gultekin, head of AI at Snowflake, the new agents are a significant expansion of the platform, enabling users to plan and orchestrate tasks, execute them using tools such as Cortex Analyst and Cortex Search, and reflect on the results to improve responses.

The Cortex Agents are powered by Anthropic's Large Language Model (LLM), Claude 3.5 Sonnet, which was selected for its exceptional performance across reasoning and coding skills. This partnership is a result of the multi-year strategic partnership between Snowflake and Anthropic, announced in November, which aims to make the LLM available to Snowflake's users for various agentic AI products, including Snowflake Intelligence and Snowflake Cortex AI.

The Cortex Agents are designed to orchestrate across structured and unstructured data sources, delivering insights by breaking down complex queries, retrieving relevant data, and generating precise answers. According to Snowflake's blog post, the agents use Cortex Analyst (structured SQL) and Cortex Search (unstructured data) as tools, along with LLMs to analyze and generate answers. This enables users to ask questions in plain language, making data more accessible to non-technical users.

Robert Kramer, VP and principal analyst at Moor Insights & Strategy, believes that the partnership with Anthropic's Claude is a significant aspect of this release, as it enhances text-to-SQL tasks with Cortex Analyst. He notes that the agents have the potential to process large volumes of structured and unstructured data, benefiting data teams and business analysts involved in planning, reasoning, and collaboration.

Kramer emphasizes the importance of processes, change management, and data management to maximize agentic AI capabilities. He advises organizations to ensure their data is well-categorized and accessible to allow for structured and unstructured data to be fully leveraged. Snowflake's built-in governance and security measures, combined with these features, should help organizations manage and utilize data more effectively.

Christian Kleinerman, executive vice president of product at Snowflake, reiterated the importance of data strategy in AI deployment, stating that customers need to have their data in order, with proper governance and knowledge of data sets, to get value out of AI. Snowflake's AI Data Cloud aims to provide customers with the choices to pursue the data architecture that they want, breaking down silos and eliminating data copies.

Security played a significant role in Snowflake's decision to partner with Anthropic, with Gultekin highlighting that Snowflake prioritizes security and privacy, and Anthropic is dedicated to building safe and reliable AI. As a result, Claude is now running inside the Snowflake security boundary, enabling customers to build and deploy AI systems while keeping their data governed.

In conclusion, Snowflake's Cortex Agents, powered by Anthropic's Large Language Model, mark a significant milestone in the company's AI strategy. By providing users with enhanced AI capabilities, Snowflake aims to help organizations deploy AI applications faster, with better accuracy, and automate complex workflows. As the AI landscape continues to evolve, partnerships like this one will play a crucial role in shaping the future of AI adoption.