Launching the Sales Pilot

1. Launching the Sales Pilot

Welcome to the Databricks Genie case study! Let's build a Genie space that a real business team can trust.

2. The data team's firehose

Picture the data team at a fast-growing company. Sales, Marketing, and Ops fire off questions all day, and every one becomes a ticket. The queue never empties. But what if the business could just ask, and trust the answer that comes back? That's exactly what a well-curated Databricks Genie space delivers, and in this course, you'll build one.

3. Before you start

Quick check before we start. This course assumes you've completed Introduction to Databricks Genie, or that you're comfortable with the basics: asking questions and reading the SQL Genie generates. You won't write any SQL here. But you will open Genie's SQL now and then, to check what it's really doing under the hood.

4. Chapter 1: Launch the pilot

So how do you get there? Across the course, you'll take one Genie space on a journey. In Chapter 1, you launch it as a pilot for a small Sales team, getting it to answer reliably from day one.

5. Chapter 2: Go department-wide

In Chapter 2, the pilot's a hit. You widen it to the whole department as Marketing starts asking too, adding more data and locking down the definitions everyone shares.

6. Chapter 3: Earn enterprise trust

And in Chapter 3, you harden it into a resource leadership trust. You'll diagnose what's failing and build checks so it stays reliable as everyone piles on.

7. Your first scenario

So let's start the journey. Your first scenario: the Sales Ops Analyst wants a Genie space where the team can simply ask questions. So you'll spin up an empty space on one table, payments, and ask a basic question with zero curation. Genie answers, but it quietly makes an assumption you never asked for. That's your cue to start curating, and from here, you'll add one lever at a time.

8. Descriptions

Your first lever is descriptions. They tell Genie what each table and column actually means. They take the guesswork out of ambiguous columns, anchor Genie to the right field, and pay off more and more as the space grows. Think of them as the foundation everything else builds on.

9. Synonyms

Next, synonyms. They map your team's vocabulary onto the real column names. Different people ask the same thing in different words, so "earnings," "sales," and "revenue" should all land on the same column. It's insurance that keeps everyone aligned, and it matters more with every new person who joins the space.

10. Default instructions

Then default instructions. These are rules that fire on every single query. Things like "exclude failed payments by default" or "sort breakdowns largest first." This is how you encode the team's unwritten rules, so the logic stays consistent no matter who's asking or how they phrase it.

11. Entity matching

And finally, entity matching. Ask about "successful" payments, and Genie might invent a value that doesn't even exist in your data. Entity matching connects the words people use to the real values in a column, so "successful" resolves cleanly to status equals completed. It's high-leverage on columns with just a few fixed values, like status.

12. How to work through the course

One tip before you dive in. Because every exercise builds on the same space, we recommend doing the full course, or at least a full chapter, in one sitting, so your changes carry forward in the same session. And if something disconnects or breaks along the way, don't worry: each chapter has a Start and a Complete checkpoint space you can jump straight to.

13. Let's practice!

That's the plan: one space, built up lever by lever. Let's launch the Sales pilot!

Create Your Free Account

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