Companies around the world are turning to AI to drive faster impact and better outcomes for their customers. Yet scaling AI beyond isolated experiments remains a challenge. The organizations that successfully implement AI in production often have one thing in common: They are built on scalable, interoperable and secure AI platforms connected directly to the data.

Snowflake is committed to helping enterprises derive significant value from their AI investments. In addition to new capabilities that accelerate AI readiness by preparing your data foundation and modernizing the developer workflow, we are excited to announce innovations that simplify how enterprises can use and deploy AI at scale. These capabilities can be used to: 

• Provide every user with the ability to answer complex questions in natural language with Snowflake Intelligence
• Develop data agents intuitively and flexibly on top of the enterprise context database, including all your structured and unstructured data, with enhanced reasoning and access to leading model providers out of the box
• Accelerate multimodal AI pipelines using simple SQL, including embedded cost governance controls for easy TCO management
• Accelerate production ML with advanced MLOps capabilities and low-latency serving of ML features for online predictions

Snowflake Intelligence: All your knowledge. One trusted enterprise agent. 

For too long, insights have lived in static dashboards and rigid reports that take time to build, are difficult to navigate and often raise more questions than they answer. Snowflake Intelligence (generally available) introduces a new category of AI: enterprise intelligence agents designed to transform how organizations work, collaborate and innovate by fostering a true data-driven culture. Every employee, regardless of technical depth, can ask complex questions in natural language, uncover the “why” behind every “what,” and take confident action, all within Snowflake’s secure and governed perimeter.

Snowflake Intelligence democratizes data and insights, empowering anyone to reason, decide and act with confidence. Users can analyze, reason and get verified answers from all data, business context and rich semantics in one trusted environment. 

Customers can use Snowflake Intelligence for answers on all data, including unstructured data, with Agentic Document Analytics (private preview soon). Users are no longer limited by traditional RAG, where only tens of documents are sampled per query, but can analyze thousands of documents with a single query. Now you can ask powerful questions on your documents, so you can graduate from “How do I reset my password?” to “Show me a count of weekly mentions by product area in my customer support tickets for the last six months.”

Snowflake Intelligence also puts your semantic layer on autopilot (public preview soon) — adapting as your business evolves and helping you get to insights faster. By learning from live usage data and query history, it automatically surfaces new verified queries for review and continuously refines its understanding of your metrics, definitions and business logic. Snowflake can automatically import context from tools such as Tableau, manually entered SQL and more BI platforms coming soon, jump-starting and continuously improving your semantic models. This results in higher-quality models, faster time to value and insights you can trust from Day 1. 

By delivering all your knowledge with one trusted enterprise agent, Snowflake Intelligence removes the barriers between questions and insights, empowering employees to explore, reason and innovate together. 

Build trusted, context-rich data agents 

Snowflake provides powerful tools for building trusted, production-ready data agents to retrieve all your structured and unstructured data more accurately. These agents can be easily accessed through the prebuilt natural language interface in Snowflake Intelligence or integrated into custom multiagent architectures using the Snowflake Cortex Agents API (generally available).

Cortex Agents orchestrate information across structured and unstructured data sources to deliver insights. They break down complex queries, retrieve relevant data and generate precise answers using Snowflake Cortex Search and Snowflake Cortex Analyst. Customers can select from a choice of out-of-the-box LLMs such as OpenAI GPT and Anthropic Claude for the orchestration. This enables accuracy, efficiency and governance at every step.

Snowflake is making it easier to build agents that provide domain-specific and context-aware insights. With Snowflake Cortex Knowledge Extensions and sharing of semantic views (both generally available), enterprises can enrich agent interactions with context and knowledge from third-party providers and publishers, all while enabling intellectual property protection and proper attribution for content owners. The latest AI-ready data providers and publishers available on Snowflake Marketplace include Alation, Allium, The Associated Press, CARTO, CB Insights, Cotality, Crisp, Crunchbase, Deutsche Börse, Equilar, FactSet, Flipside Crypto, Investopedia, IPinfo, LSEG, MSCI, NASDAQ, Nordot, Truestar and The Washington Post. Additionally, Snowflake provides a range of AI-ready public data sources on Snowflake Marketplace including earnings calls, clinical trials, medical research and Snowflake documentation.

As you build your agents and applications in Snowflake, we want to make sure they are interoperable and connected to your other systems. MCP has become the standard way for agent tooling to connect and integrate. The Snowflake managed MCP server (generally available) allows you to securely connect to Cortex Agents, Cortex Analyst and Cortex Search using the MCP protocol. This makes it possible for your other systems and multiagent architectures to retrieve high-quality business context through Snowflake AI, without your data ever needing to leave Snowflake’s secure governance boundary.

To support high-performance, production-grade AI workloads, Snowflake Interactive Tables and Warehouses (generally available soon) deliver the sub-second serving layer required to power agent-driven experiences with high concurrency and consistent speed. 

Build multimodal AI pipelines with faster performance 

Snowflake is reimagining data analysis with Cortex AISQL (generally available), enabling the creation of AI pipelines that work with text, documents, images, audio and video, all using familiar SQL, without the need for separate AI services. From production-ready functions for multimodal data analysis at scale, such as AI_COMPLETE, AI_TRANSCRIBE and AI_CLASSIFY, to document-specific functions, such as AI_PARSE_DOC and AI_EXTRACT, developers can tackle complex use cases at any scale. For a no-code experience, users can upload files directly into the intuitive Document AI Playground (in public preview soon)​​ to extract or parse information and generate corresponding Cortex AISQL code automatically. With Dynamic Table support for Cortex AISQL, AI-driven insights stay continuously up to date through incremental updates.

Cortex AISQL makes AI processing accessible to every SQL user while empowering them to build AI pipelines with ease, streamlining the development process for faster delivery and improved developer experience. Additionally, it delivers high-performance AI pipelines. Based on proof-of-concept code using similarly sized models, Cortex AISQL runtime results have shown three to seven times faster query runtime than those in manual pipelines, as a result of the performance optimizations built directly into Snowflake’s engine. This performance gain translates to lower computational costs and faster time to value. 

To simplify total cost of ownership, users can take advantage of embedded cost governance controls, including helper function to count AI tokens (public preview), feature- and model-level role-based access control (public preview) and cost guardrails (generally available soon). These capabilities make it easier to estimate, monitor and manage cost while maintaining enterprise-grade governance. 

Additional enhancements include major query performance optimizations, expanded data type support and new SQL constructs and functions for simplified data transformation. Built-in performance observability enables pipelines to remain efficient and cost effective at scale. 

Customers are using Cortex AISQL to build powerful multimodal AI pipelines, achieving faster performance and lower cost. This includes customers such as Allegis Group, a global leader in workforce and business solutions, and Eaton, a power management company doing business in more than 160 countries. 

Build and deploy custom AI/ML models in production

Snowflake ML is a fully integrated set of development, inference and operations capabilities built directly on top of your governed enterprise data. Our latest enhancements make it even easier for data scientists and ML engineers to build, deploy and manage production-ready models on a single platform without any data egress. Customers such as Coinbase, IGS Energy and Scene+ are building faster, more scalable ML workflows in Snowflake.

For model development, ML teams can now more easily build high-quality models by identifying, sharing and reproducing some of the best-performing models across training runs through natively integrated experiment tracking (public preview). This simplifies collaboration, improves reproducibility and accelerates model iterations across the enterprise. 

With full support for modular ML workflows, Snowflake ML enables scalable inference for models built anywhere, including external ML platforms or model hubs such as Hugging Face, all while maintaining Snowflake’s governance and security controls.

For inference, users gain instant access to top pretrained Hugging Face models with one-click deployment (public preview) from a simple UI and serve features for low-latency predictions under 50 ms with Online Feature Store (public preview). This supports critical use cases such as personalized recommendations, fraud detection, pricing optimizations and anomaly detection, enabling your models to make decisions based on the freshest data available.

Learn more and try now

There is no AI strategy without a data strategy. With Snowflake AI’s latest advancements, we’re making it simpler and faster for every enterprise to build, deploy and scale AI in production, driving measurable impact and better business outcomes. 

Get started with building with AI in Snowflake today by following along with these step-by-step quickstarts: 

• Build an enterprise intelligence agent with Snowflake Intelligence
• Process unstructured data at scale with Cortex AISQL
• Productionize an end-to-end model in Snowflake ML

Forward-looking statements

This article contains forward-looking statements, including about our future product offerings, and are not commitments to deliver any product offerings. Actual results and offerings may differ and are subject to known and unknown risk and uncertainties. See our latest 10-Q for more information.