Snowflake announces new AI development tools and interoperability for data sources. The solutions should help companies bring AI applications into production faster and more reliably.
Snowflake announces several new features at its Build conference in London. “We are keeping the pace of innovation and announcements high,” says Christian Kleinerman, EVP for Product, on the sidelines of the news.
That’s right, although not everything Snowflake is highlighting today is brand new. Some things have already been announced, but are generally available in their final form as of today. The common thread according to Kleinerman: “Snowflake distinguishes itself through simplicity and trust. Everything in the Snowflake platform is connected and must be manageable and scalable.” He adds that customers also immediately receive information about the financial side of their (AI) plans.
The first announcement already took place yesterday. Snowflake announced a $200 million collaboration with OpenAI. The other important news from Build relates more to its own product development.
Cortex Code (CLI)
Snowflake has made Cortex Code generally available. Cortex Code is an AI-driven coding assistant that is integrated into the Snowflake platform. The AI assistant is designed to help developers and non-technical users build data products, AI applications, and workflows through natural language.
The tool understands the data structures, security settings, and operational context of companies, allowing them to go from idea to implementation faster. Users can use Cortex Code in Snowsight (available soon) or in popular development environments such as VS Code or Cursor via the CLI, which is now available. Cortex Code was previously announced, but the Cortex Code CLI variant is completely new.
Semantic View Autopilot
Snowflake is also launching Semantic View Autopilot in general availability. This is a service that automates the creation and maintenance of semantic views. These are important for AI systems to use consistent business logic and to limit the risk of hallucinating output.
By using usage data, the tool automatically builds semantic models that can be integrated with tools such as Looker, Sigma and ThoughtSpot. This significantly speeds up the journey from data to insight.
Expansion Snowflake Postgres
Snowflake is also expanding its support for Postgres. Snowflake Postgres, which will soon be generally available, will then run entirely within the Snowflake AI Data Cloud. This allows companies to combine transactional, analytical and AI applications on one platform. The solution is compatible with open-source Postgres and requires no code changes for migrations. This makes it possible to integrate operational and analytical workloads without complex data flows.
Snowflake is also introducing pg_lake, an extension that allows Postgres environments to communicate directly with Apache Iceberg tables, without intermediate steps. This should help customers work more efficiently with open data formats.
AI management and data accessibility
With the introduction of Cortex Agent Evaluations, companies can analyze and improve the behavior of AI agents before they go into production. This evaluation tool provides insight into reasoning, error analysis and cost efficiency. Cortex Agent Evaluations should be generally available soon.
Snowflake also announced improved features for data management and interoperability. The Horizon Catalog makes it possible to apply governance to data that is consulted via other engines.
“Customers don’t want two catalogs,” says Kleinerman. “All Iceberg and REST interfaces now end up in the Catalog.”
Snowflake now also supports open data formats such as Apache Iceberg and Delta Lake via the new Open Format Data Sharing model. Furthermore, an integration with Microsoft OneLake is available, which enables bidirectional access to Iceberg data between both platforms.
Backups
Finally, Snowflake announces that Snowflake Backups will be generally available. This built-in data protection service offers protection against data loss due to failures or ransomware, and makes it impossible to delete data after creation. The backups are immutable and should certainly offer added value in highly regulated sectors.