Snowflake Inc. and Anthropic on Wednesday said they have teamed to embed artificial intelligence (AI) agents into enterprise operations, allowing them to process and analyze mountains of data without compromising security.
The public preview of Cortex Agents, APIs built on top of the Snowflake Intelligence platform, integrates Anthropic’s Claude 3.5 Sonnet large language model (LLM).
“We believe that AI agents will soon be essential to the enterprise workforce,” Baris Gultekin, head of AI at Snowflake, said during a press briefing. “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.”
Snowflake’s partnership with Anthropic is designed to address a challenge that has bedeviled enterprises adopting AI: Deploying powerful AI models at scale. Claude runs within Snowflake’s security boundary, greatly mitigating the transfer of sensitive data to external AI services.
Indeed, Snowflake said it achieved 90% accuracy on text-to-SQL tasks using Claude to aggregate and interpret results, outperforming competing enterprise search stacks on retrieval accuracy with OpenAI embedding models by at least 12% in a series of benchmarks.
Cortex agents, according to Snowflake, work in concert 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.”
Snowflake also announced upgrades to Cortex Analyst, Search and Observability. A advanced JOIN validation reduces instances such as JOIN hallucinations and double-counting, and it lets Analyst support multitable queries without compromising precision.
Wednesday’s news builds upon a multi-year strategic partnership announced between Snowflake and Anthropic in November.
