Snowflake Inc. announced new innovations to help enterprises deliver real business impact with AI, which requires more than high-quality models alone. Snowflake is unveiling Semantic View Autopilot(now generally available), an AI-powered service that automates the creation and governance of semantic views, giving AI agents a shared understanding of business metrics to deliver consistent, trustworthy outcomes. Snowflake is also introducing new capabilities across agent evaluations and observability, end-to-end machine learning (ML), and AI cost governance.
These innovations build on Snowflake?s existing enterprise-grade foundations, ensuring that AI systems such as Snowflake Intelligence are trusted, governed, and ready to operate reliably at scale, all while working directly on organizations? most valuable data. Enterprises are deploying AI agents into environments where business metrics are manually defined and inconsistently governed, leaving them without a shared understanding of business context.
This fragmented approach to building the semantic layer is a bottleneck for AI adoption, producing unreliable outputs and weakening trust in AI. Semantic View Autopilot addresses this challenge by automatically building, optimizing, and maintaining governed semantic views, potentially eliminating the need for manual, error-prone semantic modeling. This builds on Snowflake?s commitment to initiatives like the Open Semantic Interchange (OSI), which establishes an interoperable semantic layer across ecosystem leaders.
While OSI provides the connectivity to share business logic across the ecosystem, Semantic View Autopilot adds the intelligence to create and continuously maintain it, making it the connective layer for trustworthy, scalable AI across all data, wherever it lives. By learning from real user activity and using AI-powered generation, Semantic View Autopilot will help ensure business logic remains accurate and up-to-date across Snowflake data and consumption tools including dbt Labs, Google Cloud?s Looker, Sigma, and ThoughtSpot(generally available soon). Customers can create semantic views using business definitions not only from Snowflake, but also from the business intelligence tools they already rely on.
As a result, enterprises can minimize AI hallucinations while cutting semantic model creation from days to minutes, accelerating time-to-market and delivering a decisive competitive advantage. Leading organizations including eSentire, HiBob, Simon AI, and VTS are already using Semantic View Autopilot to dramatically reduce data-to-insight timelines and free data teams to focus on higher-value AI innovation. To speed up the delivery of powerful ML models, Snowflake is unveiling significant advancements to Snowflake Notebooks(now generally available), a fully-managed Jupyter-powered notebook built for end-to-end data science and ML development on Snowflake data.
Snowflake Notebooks is integrated directly with Cortex Code in Snowsight(generally available soon), a data-native AI coding agent built to automate and accelerate end-to-end enterprise development. This allows users to build and deploy fully-functional ML pipelines using simple natural language prompts, reducing manual effort and speeding up workflows. Experiment Tracking (now generally available) makes it easy for teams to compare training runs, share results, and reproduce the best-performing models from within Snowflake Notebooks, turning experimentation into a repeatable, collaborative process.
When models are ready for production, Snowflake supports real-time use cases with Online Feature Store (now generally available) and Online Model Inference(now generally available), enabling features to be served in milliseconds and predictions delivered at scale. With training, serving, and monitoring all happening within the Snowflake platform, teams can operationalize ML while maintaining consistent governance from data to model to insight.
