Snowflake launches AI agents to ease enterprise data access

Snowflake has introduced new agentic AI features and expanded its enterprise-grade AI capabilities, aiming to enhance data analysis and machine learning (ML) workflows for businesses in Canada and worldwide.

Snowflake Intelligence, set to enter public preview soon, provides business users and data professionals with a unified conversational interface driven by intelligent data agents. This development enables users to pose natural language questions and quickly access actionable insights from both structured and unstructured data.

The company has also announced Data Science Agent, currently in private preview, which acts as an agentic companion designed to assist data scientists by automating routine ML model development tasks. These additions are intended to streamline AI and ML workflows, widen access to data within enterprises, and remove the technical barriers that traditionally slow business decision-making through natural language interactions within Snowflake.

"AI agents are a major leap from traditional automation or chatbots, but in order to deploy them at scale, businesses need an AI-ready information ecosystem. This means enterprises must be able to unite data silos, maintain enterprise-grade security and compliance, and have easy ways to adopt and build agents. Snowflake Intelligence breaks down these barriers by democratizing the ability to extract meaningful intelligence from an organization's entire enterprise data estate — structured and unstructured data alike. This isn't just about accessing data, it's about empowering every employee to make faster, smarter decisions with all of their business context at their fingertips," Baris Gultekin, Head of AI at Snowflake, said, commenting on the evolution of AI agents.

Organisations frequently face difficulties in decision-making due to fragmented data governance, separate data formats, and a lack of technical analysts. Snowflake Intelligence addresses these issues by enabling business teams and non-technical users to interact conversationally with their enterprise data, all without needing to write code.

Snowflake Intelligence operates within the user's existing Snowflake environment, inheriting all established security controls, data masking, and governance policies. It consolidates data from multiple sources, including Snowflake, Box, Google Drive, Workday, and Zendesk, via Snowflake Openflow, allowing users to retrieve insights from spreadsheets, documents, images, and databases simultaneously. Data agents can create visualisations and help users act on insights through natural language prompts. The platform also provides access to external knowledge via Cortex Knowledge Extensions available on Snowflake Marketplace, with content provided by sources such as CB Insights, Packt, Stack Overflow, The Associated Press, and USA TODAY, to add further depth and context to responses.

The system is powered by large language models from Anthropic and OpenAI and is built on Cortex Agents, currently in public preview. All are presented through a no-code interface that seeks to ensure transparency and explainability in the use of AI.

"By integrating Claude's reasoning capabilities directly into Snowflake's platform, we're further eliminating the traditional barriers between data and insights. Business users can now have natural conversations with their enterprise data, while data scientists can automate complex ML workflows — all through simple natural language interactions. This demonstrates how Claude's advanced reasoning can democratize AI while maintaining the enterprise-grade security and governance that organizations require," Michael Gerstenhaber, VP of Product Management at Anthropic, said, highlighting the integration's potential.

Snowflake Intelligence is aimed at moving organisations away from reliance on analytics teams for insights, enabling broader employee access to data.

"At WHOOP, our mission is to unlock human performance and healthspan, and data is central to everything we do. Snowflake Intelligence marks a big step forward in our ability to be a data-first organisation, ensuring that all employees can access insights without relying on analytics teams as the intermediary. By eliminating the technical barriers to gleaning the insights we need for decision-making, our analytics teams can now shift from manual data retrieval tasks to more strategic, predictive, and value-generating work," Matt Luizzi, Sr. Director of Business Analytics at WHOOP, said.

To support data scientists, Snowflake's Data Science Agent automates time-consuming tasks linked to ML workflows. The agent, also using Anthropic's Claude, segments ML workflow challenges into separate steps such as data analysis, preparation, feature engineering, and training. It leverages advanced reasoning, contextual understanding, and action execution to generate validated ML pipelines that can be run from a Snowflake Notebook. Users can iterate with suggested improvements or follow-ups, helping to reduce time spent on experimentation or debugging.

Currently, more than 5,200 customers, including companies such as BlackRock, Luminate, and Penske Logistics, are using Snowflake Cortex AI as part of their business operations. Snowflake is introducing several new AI features, such as enhanced document processing, batch semantic search, and the new Cortex AISQL, now available in public preview, aiming to facilitate analysis of multi-modal data at scale and assist teams that may lack extensive AI engineering skills.