The Easy Button for Context-Rich AI Agents: Snowflake Intelligence & Cortex Knowledge Extensions

The promise is bold: AI agents will transform how enterprises operate, delivering enhanced efficiency, deeper insights and accelerated decision-making. But there's a critical bottleneck holding them back: These powerful AI innovations are only as good as the information they can access. While they excel at processing internal company data, they are not able to tap into the wealth of knowledge that exists beyond organizational walls, in copyrighted publications, paywalled research and real-time news sources that could make the difference between stale insights and impactful ones.

To address these challenges, Snowflake launched Snowflake Cortex Knowledge Extensions, now generally available, which create a clear pathway for enterprises to easily connect their AI solutions to proprietary external content in a fair and transparent manner. Cortex Knowledge Extensions are available on Snowflake Marketplace from top-tier providers such as The Associated Press, The Washington Post, Gannett | USA TODAY Network, Stack Overflow, Packt Publishing and PubMed (published by Snowflake).

At Snowflake, we know that to fully realize this promise, enterprises need AI systems that can seamlessly blend internal data with external expertise. Snowflake Intelligence, now in public preview, represents a transformative step forward, enabling business users to converse naturally with their enterprise data while easily incorporating trusted external sources. This agentic experience allows organizations to unlock deeper insights and accelerate decision-making across their entire data ecosystem.

The challenge: Bridging the gap between AI and external expertise

Foundation models are powerful tools, but they lack knowledge of current events because they are trained on data that’s months or even years old. Data engineers often resort to setting up and maintaining complex, less reliable scraping practices with other platforms that don’t provide an out-of-the-box solution. 

Additionally, enterprises have compliance and governance concerns, as enterprise subscriptions typically contain AI usage license restrictions put in place by the publishers. Enterprise IT security may put restrictions on sending prompts or context outside of the security perimeter as these are often considered intellectual property.  

On the flip side, content owners and publishers are concerned about their intellectual property being scraped and used for AI training without proper compensation or attribution.

The rapid growth of reasoning-based agentic AI assistants has led to publishers actively blocking their content from web search tools because there is no monetization path. This creates a critical gap: Enterprises want to enrich their AI with external expertise and context from publishers they trust and are relevant to their business, while content providers need a pathway to monetization and assurances that their work is protected.

Snowflake's solution: Cortex Knowledge Extensions

Snowflake addresses this challenge head-on with Cortex Knowledge Extensions on Snowflake Marketplace. Cortex Knowledge Extensions are designed to benefit both sides, enabling enterprises to easily enrich their agents and AI applications, such as Snowflake Intelligence (in public preview), with proprietary context and knowledge from third-party providers and publishers, all while enabling intellectual property protection and proper attribution for content owners. 

Enterprise data and AI teams benefit from Cortex Knowledge Extensions in meaningful ways:   

• Avoid complex, less reliable scraping practices: Quickly access high-quality, AI-ready data sources from premium publishers without the need for scraping, chunking, inferencing or model training.

• Easily enhance AI context, without training: Not only provide agents access to all the structured and unstructured data held within an organization but connect up-to-date third-party content and knowledge with retrieval-augmented generation (RAG). Responses are clearly cited and link back to relevant sources for quick verification.

• Build confidence in enterprise security and compliance: Prompts are never shared with providers, and appropriate licensing is in place for the content to be used in AI and agentic use cases. 

How Cortex Knowledge Extensions work

Cortex Knowledge Extensions facilitate secure and governed use of external data in AI apps and systems using RAG and Snowflake Cortex Search, Snowflake’s low-latency, high-quality semantic search engine that finds relevant documents across the provider’s content for a user or agent prompt. This leverages Snowflake's zero-ETL Secure Data Sharing functionality, which allows content owners to revoke access to their content at any time. Crucially, outputs from Cortex Knowledge Extensions include citations and links to the original content for increased reliability and accuracy. Because this follows a RAG architecture, the AI output is grounded in real content sources.

• Provider setup: Content owners and publishers set up a Cortex Search service on their data and publish it on Snowflake Marketplace or share it privately. Providers can include up-to-date, live data refreshes as well as archival content in their listings, depending on the needs of the enterprise customer. 

• Content acquisition: Enterprises can discover, try and purchase AI-ready content offerings directly on Snowflake Marketplace. Once installed, a shared Cortex Search Service appears in the customer’s Snowflake account. It enables users to “speak to” the content without ever being able to see all of it — crucial for preventing model training on the publisher’s content. For example, a user can ask “What risks are we facing in our exposure to steel price fluctuations?” Using multistep agentic reasoning, Snowflake Cortex AI is able to execute SQL queries through Snowflake Cortex Analyst to assess current investments into steel commodities and use Cortex Search to sift through Cortex Knowledge Extension news sources to discover looming geopolitical risks. 

• AI integration: The sources can be easily connected to agents in Snowflake Intelligence, which allows teams to talk to their enterprise data as well as the licensed content to unlock real business insights. To integrate the licensed content into a system or application, users can prompt Cortex AI APIs with their choice of large language models (LLMs) as well as through MCP servers on Snowflake (public preview soon). 

Enter Snowflake Intelligence

Snowflake Intelligence serves as the unified interface where these powerful capabilities come together. Now in public preview, this agentic experience transforms how business users interact with data by enabling natural-language conversations across both structured and unstructured information. 

Unlike traditional AI solutions that primarily focus on document retrieval, Snowflake Intelligence leverages established technologies such as Cortex Analyst to generate more accurate SQL queries directly against enterprise data while simultaneously accessing external knowledge sources through Cortex Knowledge Extensions. This dual capability means users can ask complex questions such as "How do our Q3 sales compare to industry trends, and what external factors might be influencing performance?" and receive comprehensive answers that blend internal metrics with relevant market intelligence.

What sets Snowflake Intelligence apart is its enterprise-grade foundation built on Snowflake's robust security and governance framework. All existing role-based access controls, data masking policies and governance rules are automatically honored in every conversation, helping ensure users only receive insights based on data they're authorized to access.

This seamless security inheritance eliminates the need to redefine permissions at the application layer while providing transparency into how insights are generated. Enterprises can start using Snowflake Intelligence today to democratize data access, empower business users with self-service insights and elevate their data teams from reactive support to strategic enablement, all while maintaining robust data protection and compliance controls.

Use cases for premium content for agents

By integrating premium publication sources available on Snowflake Marketplace with AI assistants and agentic systems, enterprises can unlock transformative capabilities across business functions and industries:

• Financial services: Enhance risk analysis and investment strategies. An AI assistant powered by AP Business News can answer prompts such as "What’s the risk profile of steel commodities this quarter?" providing real-time insights into market volatility and geopolitical factors affecting commodities. This enables more informed and agile financial decision-making.

• Technology: Improve developer efficiency and streamline problem-solving. By incorporating Stack Overflow content or Packt Publishing textbooks, an LLM can provide targeted answers to programming challenges. For example, a developer could ask "I’m getting TypeError: unhashable type: 'list'. What should I do?" and receive fast, relevant guidance directly within their internal development environment.

• Market research: Conduct more nuanced and accurate sentiment analysis and competitive intelligence. Leveraging expert research from CB Insights allows AI assistants to analyze consumer trends, competitive landscapes and emerging market opportunities with greater depth, leading to more impactful business strategies.

• Life sciences: Accelerate drug discovery research and clinical analysis. Integrating research papers and journal articles from PubMed Biomedical Research Corpus into an LLM empowers researchers to ask complex questions such as "In adults with obesity without diabetes, how does Tirzepatide compare with semaglutide for percent weight loss at ~68-72 weeks and discontinuations due to GI adverse events?” dramatically reducing the time spent on literature reviews and aiding in the identification of novel therapeutic pathways.

• General business operations and strategic planning: Beyond specific industry applications, the ability to contextualize business questions with current news and trends from sources such as The Associated Press, USA TODAY Network and The Washington Post offers broad value. This allows leadership to quickly get briefings on specific topics, monitor brand reputation in real time, understand the context around customer accounts and gain a competitive edge by staying ahead of industry shifts, compliance/regulation and public perception.

By leveraging Snowflake Intelligence and Cortex Knowledge Extensions on Snowflake Marketplace, enterprises can seamlessly blend their proprietary data with high-value external knowledge and context. This not only enhances the accuracy and utility of their AI applications but also fosters a more informed, agile and competitive organization by putting the trusted information directly at the fingertips of their AI-powered workforce.

Get started with agents in Snowflake today

• Gain insights from structured and unstructured data with Snowflake Intelligence via this quickstart.

• Explore Cortex Knowledge Extensions by bringing Snowflake Documentation into Snowflake Intelligence via this quickstart. 

• Watch this Snowflake Intelligence demo with content from the USA TODAY Network and Cortex Knowledge Extension demo to chat with Snowflake Documentation.