Agentic AI is all the rage right now, and for good reasons. Automating tasks or processes can be the difference between cutting costs and going over budget. For Snowflake users, Agentic AI lies in the Snowflake Data Cloud within Cortex. But to build an agentic platform, you need to understand Snowflake’s Agentic AI Framework first. This will ensure proper setup, governance, and usage to give you and your organisation the full benefits of agentic AI.
What does Snowflake’s Agentic AI Framework Consist of?
Snowflakes Agentic AI Framework is made of 4 key components: Snowflake Intelligence Portal, Cortex Agents, Cortex Analyst, and Cortex Search. From an end-user perspective, it all comes together in the Snowflake Data Cloud.
What is Cortex Analyst?
Cortex Analyst is a fully managed, LLM-powered feature that enables applications to answer business questions from structured data using natural language without requiring SQL queries.
Key Capabilities
Semantic Models: Uses YAML files to bridge the gap between business terminology and database schemas
Multi-turn Conversations: Supports follow-up questions that build on previous queries
How It Works
Cortex Analyst is an agentic AI system with a classification agent that rejects ambiguous questions upfront to avoid hallucinations.
It generates text-to-SQL responses by leveraging semantic context
Use Cases
Self-service analytics for business users
Conversational BI dashboards
Automated reporting and insights generation
What is Cortex Search?
Cortex Search is a fully managed text search service that enables low-latency, high-quality fuzzy search over Snowflake data using hybrid vector and keyword search.
Key Capabilities
Automatic Index Refresh: Updates when source data changes without manual intervention
Powered by Arctic Embed M: Uses Snowflake's high-performance, cost-effective embedding model
No Infrastructure Management: Fully integrated into Snowflake with built-in governance
How It Works
Vector search for retrieving semantically similar documents.
Keyword search for retrieving lexically similar documents.
Semantic reranking for reranking the most relevant documents in the result set.
Use Cases
RAG Engine for LLM Chatbots: Provides context for conversational AI applications
Enterprise Search: Backend for high-quality search bars in applications
What are Cortex Agents?
Cortex Agents are agents that orchestrate across both structured and unstructured data sources to deliver insights by planning tasks, using tools, and generating responses.
Key Capabilities
Parse Requests: Understands complex queries and creates execution plans.
Explore Options: Disambiguates ambiguous questions to improve accuracy.
Split into Subtasks: Breaks down complex requests into manageable components.
Route Across Tools: Automatically selects between Cortex Analyst (structured) and Cortex Search (unstructured).
Tool Integration
Cortex Analyst for SQL generation on structured data.
Cortex Search for insights from unstructured sources.
Custom Tools: Supports stored procedures and UDFs for specialized functionality.
Workflow Process
Four-part workflow: understanding requests and planning subtasks, executing via tools, evaluating answer effectiveness, and monitoring/iterating for continuous improvement.
What is Snowflake Intelligence?
Snowflake Intelligence is an agentic platform accessible via ai.snowflake.com that enables users to gain insights and act using natural language, without technical expertise.
Key Capabilities
Create Charts and Get Instant Answers: Discover trends without waiting for custom dashboards.
Access Thousands of Data Sources: Analyse structured and unstructured data together from spreadsheets, documents, images, and databases.
One Interface for Everything: Connects siloed systems.
Tool Integration
Respects Role-Based Access Controls: Maintains data masking policies and governance
Security First: All data remains within Snowflake's secure perimeter
No Code Required: Business users interact directly with data
Conclusion
Snowflake's Agentic AI Framework provides a comprehensive integrated solution for leveraging intelligent automation within the Snowflake Data Cloud. Rather than being a collection of separate tools, the framework operates as a cohesive system, with each component playing a critical role in delivering end-to-end insights. Ultimately, the framework transforms how business users interact with data, providing a single, secure, natural-language pathway to generate analytics and insights from an organization’s entire data estate.
As a Premier Snowflake Partner, Keyrus is equipped with the necessary knowledge and experience to establish your Agentic Snowflake ecosystem. This architectural approach is key to achieving the promised efficiency and automation of agentic AI within a governed enterprise environment.