Snowflake announces preview of Cortex Agent APIs to power enterprise data intelligence

Snowflake on Wednesday announced the public preview of Cortex Agents, a set of APIs built on top of the Snowflake Intelligence platform, a low-code offering that was first launched in November at Build, the company’s annual developer conference.

Asked at a press and analyst briefing held earlier this week how the most recent launch differs from what was introduced at Build, Baris Gultekin, head of AI at Snowflake, responded, “it is not a repeat” and that the agents are what Software Intelligence, which will be in private preview soon, builds on.

Cortex Agents, he said, “plan and orchestrate tasks, use tools such as Cortex Analyst and Cortex Search to execute them, reflect on the results, and improve responses. As part of the planning, they explore options, split into smaller tasks, and overall provide a very highly accurate scalable system.”

The company, he said, believes “that AI agents will soon be essential to the enterprise workforce. They’ll enhance the productivity for many teams such as customer support, analytics, engineering, and they’ll free up employee time to focus on higher value things. Data agents, which is a specialized category of AI agents, will combine data and tools to deliver accurate grounded insights by effectively selecting the right data sources.”

The new agents will be powered by Anthropic’s Large Language Model (LLM), Claude 3.5 Sonnet, selected by the company, according to a blog post, for its “performance across reasoning and coding skills.”

In November, Snowflake and Anthropic announced a multi-year strategic partnership in which the LLM would be available to Snowflake’s users for a number of  its agentic AI products, including Snowflake Intelligence and Snowflake Cortex AI, the company’s managed AI service.

Cortex agents, the blog stated, orchestrate across “structured and unstructured data sources, whether they’re Snowflake tables or PDF files stored in object storage, to deliver insights. They break down complex queries, retrieve relevant data, and generate precise answers, using Cortex Search, Cortex Analyst and LLMs.”

“Agents use Cortex Analyst (structured SQL) and Cortex Search (unstructured data) as tools, along with LLMs to analyze and generate answers,” it added.

Agentic outputs, the blog stated, “are only as good as the quality of the underlying data and the accuracy of the retrieval systems that help ground them. Yet organizations struggle to pave a path to production due to an AI and data mismatch. LLMs excel at unstructured data, but many organizations lack mature preparation practices for this type of data; meanwhile, structured data is better managed, but challenges remain in enabling LLMs to understand rows and columns.” 

Robert Kramer, VP and principal analyst at Moor Insights & Strategy, said that Snowflake initially introduced Cortex Agents in November of last year at its Build developer conference as part of Snowflake Intelligence, “stressing the potential for agentic AI app development and multimodal conversational AI.”

The latest announcement, he said, “expands on this by introducing a public preview of Cortex Agents, focusing on improving the accuracy of multi-agent systems to help users complete complex tasks. One of the critical elements I like about this announcement is that these 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.”

“The jewel of this release could be Snowflake’s partnership with Anthropic’s Claude, which enhances text-to-SQL tasks with Cortex Analyst, allowing users to ask questions in plain language and making data more accessible to non-technical users,” Kramer said.

This partnership, he said “should help enterprises deploy AI applications faster, with better accuracy, and automate complex workflows with Anthropic’s Claude. Snowflake is also introducing Cortex Search for data retrieval, improving data access and analysis.”

Kramer added  “I always come back to processes, change management, and data management, in order to maximize agentic AI capabilities; organizations should ensure their data is well categorized and accessible to allow for structured and unstructured data to be fully leveraged.”

All these features, combined with Snowflake’s built-in governance and security measures, should help organizations manage and utilize data more effectively, he said.

At the briefing, Christian Kleinerman, executive vice president of product at Snowflake, said, “even though we’ve said it many times, that there is no AI strategy without data strategy, it’s been increasingly clear how customer after customer has been validating this sentiment. They say, ‘OK, I have access to a great model. But if I don’t have my data in order, if I don’t know governance, and if I don’t know what data sets I have, it is difficult to get value out of AI.’”

At the end of the day, he said, “what organizations really want is to be able to break down silos, eliminate copies, and be able to get as much value as possible from their data. And a lot of what we have done at Snowflake building the AI Data Cloud is about providing the choices for customers to be able to pursue the data architecture that they want.”

Snowflake’s Gultekin said security played a big role in the decision to partner with Anthropic. “Snowflake prioritizes security and privacy, and Anthropic is dedicated to building safe and reliable AI,” he said. “Claude is now running inside the Snowflake security boundary, so Snowflake customers can build and deploy AI systems while keeping their data governed.”