The Snowflake Summit 2025 is held this week in San Francisco. Data Platform notably reveals Snowflake Intelligence and Data Science Agent.
Snowflake was not yet positioned on the front of AI agents. It is now done. On the occasion of its global event which is held from June 2 to 5 in San Francisco, the group lifted the veil on two new features centered on so -called agent artificial intelligence. Called Snowflake Intelligence and Data Science Agent respectively, they aim to make uses of AI and machine learning more accessible and better integrated with existing business environments.
First novelty, Snowflake Intelligence offers a conversational interface cut to formulate requests in natural language based on several agents. Translated on the fly in SQL language, the latter make it possible to generate textual reports in a few seconds compiling indicators and graphics from structured and unstructured data. Usable without writing a line of code, they aim to put complex requirement within the reach of data analysts and other business managers without knowledge in development.
The tool operates directly within the Snowflake environment. It automatically applies the safety and governance rules configured within the solution. Thanks to Snowflake OpenFlow (which is also a novelty), he can also question data from external sources like Google Drive, Slack, Workday or Zendesk.
Data visualization without code
Snowflake Intelligence will also open access to content available on the web via Cortex Knowledge Extensions. The latter will enrich the answers by mobilizing sources like CB Insights, Stack Overflow or The Associated Press. Their integration in Snowflake, which is expected in the final version “very soon”, aims to improve the relevance of the analyzes, always without requiring manual intervention.
Via Snowflake Intelligence, the user may, for example, ask a question about commercial indicators scattered within one or more databases or even seek marketing information within a large documentation. He will obtain a synthetic, contextualized and visualizable response crossing different sources. In the end, he will even be able to generate an email to his team aimed at correcting any reasons raised. Behind the scenes, Snowflake Intelligence relies on language models provided by Openai and Anthropic.
“Snowflake Intelligence allows business teams to consult the data without systematically depending on data scientists”
For Baris Gultekin, head of AI at Snowflake, “AI agents must be part of a coherent environment, secure and open to the various company data formats. Snowflake Intelligence constitutes a response to these needs by offering a means of interacting directly with all the information of organizations. And this, keeping control of uses”, explains Baris Gultekin.
For some of his early adopting, including the American Whoop, Snowflake Intelligence brings an evolution in the structuring of work around data. “This tool allows business teams to consult the information without systematically depending on data scientists,” confirms Matt Luch, senior director of analytics at Whoop. “It frees them time for more strategic projects.”
Data Science Agent, for its part, is aimed at technical teams. This agent automates the development stages of machine learning models. Based on the Claude d’Anthropic language model, it analyzes a problem, offers a structuring of the machine learning flow (data preparation, variable engineering, training …), then generates a functional pipeline to be run in a notebook type environment. “The idea is to help data scientists save time on implementation tasks without replacing their expertise on methodological choices”, argues Baris Gultekin.
A Data Platform oriented AI
According to Snowflake, more than 5,200 customers already use its AI AI Cortex capabilities, with use cases ranging from generation of reports to documents. For the rest, new tools are being deployed, such as Cortex Aisql to question the SQL AI models, or large -scale documents processing functions and semantic research.
With these announcements, Snowflake continues its historical AI integration strategy into data management practices by offering tools oriented use and designed to adapt to the constraints of the company. Snowflake Intelligence and Data Science Agent will be launched in the coming weeks, respectively in public beta for the first and private beta for the second.
Other announcements from the Summit 2025 snowflake:
- Snowflake OpenFlow: a multimodal data ingestion service drawn to connect to any source,
- Agentic Snowflake Native Apps: agents who will appear on the Snowflake Marketplace,
- Snowconvert AI: A module that simplifies data migration to the Data Platform of Snowflake,
- Second generation Standart Warehouse: version 2 of the protocol is now supported by Snowflake,
- Computers -based adaptive and adaptive Warehouse: new services cut to adapt the machine resources for data processing.
