Snowflake Intelligence: A Beginner's Guide

What is Snowflake Intelligence?

Snowflake Intelligence is an AI assistant that lives inside Snowflake. Think of it as a colleague who knows your data and can answer complex questions in natural language. Instead of writing SQL or building reports yourself, you ask things like “Why did sales of Fitness Wear grow so much in July?” or “What issues are people reporting with jackets in support tickets?” and get clear, data-backed answers.

It goes beyond just showing numbers (“what” happened) and helps explain the “why”—so you can make better decisions faster. It’s built for enterprises, so it respects your existing security and governance while scaling with your data.

Why Does It Matter?

• Trust and traceability: You can trace every answer back to its source and define “golden” questions that always get verified, consistent answers.
• One place for data: It can reason across structured data (tables, numbers) and unstructured data (support tickets, documents) in one place.
• Self-service: More people can explore data and get insights without waiting on a data team.

Who Uses It? (Use Cases)

How It Works: The Main Pieces

Snowflake Intelligence is built around agents—AI assistants configured with specific “tools” and instructions. When you ask a question, the agent chooses the right tool and returns an answer. Two core tools are:

You can also add custom tools (e.g. “send an email”) so the agent can both answer questions and take actions—like emailing a summary to the right people.

Getting Started (High-Level)

You’ll need a Snowflake account with the right role and a region that supports Snowflake Intelligence. Then, in broad steps:

1. Set up data: Create a database/schema, load your data (e.g. from S3), and run the setup scripts (Snowflake provides a sample repo for this).
2. Configure Cortex Analyst: Upload a semantic model (YAML) that describes your business concepts and how they map to tables and columns. This is what lets the AI “understand” your data.
3. Configure Cortex Search (optional): Point it at tables with text (e.g. support cases), pick the columns to search and to display (e.g. title, product, transcript), and create a search service.
4. Create an agent: Give it a name, add instructions and example questions, then attach the tools (Cortex Analyst with your semantic model, Cortex Search services, and any custom tools like email).
5. Use Snowflake Intelligence: Open the Snowflake Intelligence UI, select your agent, and start asking questions. You can ask for trends, reasons, support insights, and even “send a summary email.”

Snowflake’s official guide walks through a concrete example (e.g. “Sales//AI” agent with marketing and support data) so you can try real questions like “Show me the trend of sales by product category between June and August” or “What issues are reported with jackets recently?”

Bottom Line

Snowflake Intelligence turns your Snowflake account into a place where anyone can ask complex questions in natural language and get trusted, traceable answers—combining structured and unstructured data, with optional actions like sending emails. Once you’ve set up an agent with the right tools and semantic model, you’re ready to explore your data without writing SQL.