Snowflake Inc. today announced the general availability of Snowflake Intelligence, the centerpiece of the company’s latest wave of artificial intelligence-driven products aimed at allowing employees across skill levels to derive insights from enterprise data.

Announced in June at Snowflake Summit 2025, Snowflake Intelligence enables users to ask complex business questions in natural language, leveraging both structured and unstructured data from Snowflake and third-party data sources, such as Salesforce Inc.’s Data 360. Snowflake said the platform is already in use by more than 1,000 customers in the test phase and has been used to deploy more than 15,000 AI agents, a type of software that can interact with its environment, make decisions and perform tasks.

Users can query data using a natural language interface without needing to write code. The product automatically interprets business semantics, generates and executes SQL queries. Intelligence integrates with Snowflake’s Horizon catalog to ensure access control across clouds, regions and formats.

Snowflake said its underlying model integrates AI services from partners such as Anthropic PBC and OpenAI LLC, and can synthesize multimodal data, such as logistics records and internal communications, to identify trends, root causes and recommend actions. Responses are supported by verified queries and semantic views, which are marked by a “green shield” icon indicating certified data sources were used.

Understanding the ‘why’

Intelligence is positioned as “a strategic component of how [organizations] understand what’s happening,” said Jeff Hollan, the company’s head of Cortex AI Agents. “It’s not just to get basic information but to understand why a trend is happening.”

To enhance reliability, Snowflake’s AI researchers introduced a new evaluation framework called Agent Goal, Plan, Action that reportedly catches up to 95% of errors during testing. The company also claims that text-to-SQL performance is now up to three times faster than before.

Hollan said agentic capabilities extend beyond answering queries to performing actions. “My agent understands things like how to integrate with my other systems,” he said, demonstrating a scenario in which a company’s logistics provider is automatically changed and a team is notified via Slack.

Existing data governance rules are automatically enforced, and users are only shown data they are authorized to access, said Christian Kleinerman, executive vice president of product. “We focus every day on making sure Snowflake is trusted and that our customers can entrust us with their data,” he said.

Agentic development tools

The company is also using its Build conference this week to announce a set of developer-focused enhancements designed to support enterprise-grade AI development. They include the general availability of Cortex Agents, which allow developers to define custom data agents, and a Model Context Protocol Server, which facilitates secure communication between Snowflake and external AI tools.

Developers can build and test AI pipelines within Snowflake using Dynamic Tables and Cortex AISQL. A forthcoming feature, AI Redact, will help identify and remove sensitive information from unstructured data.

The company is also introducing Cortex Code, an AI assistant that integrates directly into the Snowflake interface to help with tasks such as query optimization and system navigation. Cortex Code helps users understand their Snowflake usage, optimize complex queries and fine-tune results to minimize costs.

To support version control and collaboration, Workspaces now include Git and Visual Studio Code integrations. Snowflake is also extending support for open-source tools, such as data build tool, enabling teams to manage analytics workflows within the Snowflake environment. The new dbt Projects on Snowflake, now generally available, enables developers to use dbt to test, deploy and monitor their data transformations directly within their Snowflake environment.

Expanded lakehouse support

Snowflake is expanding its support for data lakehouses – which combine the features of a data lake and a data warehouse — with updates to the Horizon Catalog and the Snowflake OpenFlow data integration service. These enable organizations to ingest, govern and share data across systems using open-source elements, such as Apache Iceberg and the Apache Polaris Catalog.

The goal is to enable AI agents to securely access all enterprise data, regardless of format or location, Kleinerman said. “Organizations struggle with AI readiness due to fragmented governance and siloed data systems,” he said. “This couldn’t be more timely.”

Interactive Tables and Warehouses, now generally available, provide sub-second query response times to power real-time dashboards and applications. A new near-real-time streaming analytics feature is in private preview.

“Think of [Interactive Tables and Warehouses] as being either for applications or dashboards that need very low latency,” Kleinerman said. “At the end of the day, we want to provide subsecond analytics.”

A vote for PostgreSQL

In another significant step into the open-source ecosystem, Snowflake has also introduced a fully managed version of the PostgreSQL database engine, allowing organizations to run transactional workloads alongside analytical tasks on a single platform. Postgres developers can use a new extension called pg_lake to read and write directly to Iceberg tables from Postgres, eliminating the need for an extract, transform, load process. Snowflake developed pg_lake and has released it to open source.

“We have a very long list of customers asking for access to the preview and interested in getting the Postgres capabilities,” Kleinerman said. The pg_lake extension is for anyone who wants to turn Postgres into an interface to manage an open lakehouse.”

Kleinerman said that Snowflake has shortened its product release cycles and is moving faster than ever. “Many of the technologies that we’re saying are [generally available] now were introduced for the first time six months ago or even less,” he said.

Snowflake plans to continue addressing adoption barriers, which often stem from fragmented data access and governance rather than the AI models themselves. The company aims to do that by embedding AI into its native data environment and enforcing existing governance rules across all new capabilities.

Image: Snowflake