How Genie works

1. How Genie works

Now that you've seen Genie in action, let's look under the hood.

2. Genie as a compound AI system

Genie isn't just a simple chatbot—it's what we call a Compound AI System. Think of it as a team. The Brain—the large language model—understands your language and interprets what you're asking. The Map—Unity Catalog—knows where your data lives and how it's organized. The Muscle—the SQL warehouse—does the heavy lifting, running the queries and returning results. By working together, they ensure the answer isn't a guess but a calculated result based on actual data.

3. Unity Catalog integration

The secret sauce is Unity Catalog integration—where Genie gets intelligence about your specific business. Your data team adds table descriptions, column comments, and relationships. Genie reads this metadata to understand what your data means—not just technical column names like "cust_id" or "amt_usd," but the business context behind them. This metadata allows Genie to translate your business questions into accurate SQL.

4. How Genie processes your question

Here's what happens when you ask a question—let's walk through the four steps. First, Genie analyzes your natural language to understand your intent—not just keywords, but what you actually want to know.

5. How Genie processes your question

Second, it looks up relevant tables and columns from Unity Catalog, finding the data sources that can answer your question.

6. How Genie processes your question

Third, it generates SQL with the right joins, filters, and aggregations.

7. How Genie processes your question

Finally, it executes that SQL against your data warehouse and formats results into tables and visualizations. All this happens in about two seconds—what might take you minutes to write, Genie accomplishes almost instantly.

8. When Genie asks for clarification

Genie is designed to be accurate over assertive. It asks for clarification when it encounters semantic ambiguity—a real business choice. If you ask for "Sales" and Genie sees both Gross_Sales and Net_Sales, it won't flip a coin; it will ask which one. "Show me the top products"—top by revenue, quantity sold, or average rating? Each gives a different answer. This is a feature, not a limitation.

9. Demo: Triggering clarification

Let me demonstrate with our Bakehouse data. I'll ask: "Which region performed best?" Genie recognizes that "best" is subjective and asks me to clarify—by revenue, transaction count, or customer satisfaction? I'll select revenue, and I get a specific, accurate answer. The clarification took two seconds but ensures I get exactly what I intended.

10. Let's practice!

Test your understanding of Genie's components. Let's get started.

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.