Genie One Beginner's Guide: Explore Data & Automate Tasks

Summary

• Question Answering: Ask Genie One natural language questions and get instant answers backed by Genie Ontology snippets showing your organization's knowledge sources.
• Multi-Output Creation: Generate charts and documents from a single prompt, with live computation ensuring results reflect your latest data.
• Scheduled Automation: Set up email alerts that run on a schedule, keeping stakeholders informed with fresh insights delivered automatically.

Genie One Beginner's Guide: Explore Data & Automate Tasks

Master the basics of Genie One for exploring company data, building visualizations, creating documents, and automating repetitive tasks.
Learn more about Genie One: https://www.databricks.com/product/genie/one
Chapters:

Chapters

• 0:00 Introduction
• 1:00 Ask a question
• 1:27 Reviewing Genie Ontology snippets
• 4:50 Creating a chart and document from a single prompt
• 5:40 Schedule an ongoing email alert

FAQs

How does Genie Ontology help Genie One?

Genie Ontology maps your organization's knowledge across Databricks, Google Drive, SharePoint, and email. When you ask Genie One a question, it references relevant ontology snippets to provide contextual answers.

Can I create visualizations and documents together?

Yes. This video shows asking Genie One a single prompt and getting both a chart and a document created from the results—with live computation ensuring current data.

How do I set up an email alert?

You define what analysis you want, and Genie One schedules it to run and email results on your cadence—automating repetitive reporting so subject matter experts can focus on strategy.

Does Genie One understand my organization's terms?

Yes. Genie Ontology gives Genie One context about your business terminology and data assets, so when you ask about metrics like revenue or active customers, it knows your organization's definitions.

Full transcript

0:00 Introduction

Hey everyone. I'm a PM at Databricks focused on Genie, and I want to talk about a problem that I think a lot of us have. You're probably the expert in something at your company, and when you're the expert in something, everyone comes to you with questions about it. If it's a quick question, you probably just answer it, but then you do that 12 more times before lunch, and you never actually get to the work that you sat down to do that morning.

So, today I'm going to show you how to use Genie to intervene and free up that time. I can't show you real Genie metrics. I'm going to use our sample data for this. We are a cookie baking franchise with a couple weeks of sales. I'm starting here in Genie 1, and as you can see, it's a pretty simple interface. We have our chat box in the middle. We have these quick start buttons that allow me to create a skill or maybe draft a document.
On the left side here, I have my prior chats, and I can also look at the agents and the assets that are available to me, but more on that in just a second. I'm going to start by asking a question. I'm going to ask about our cookies franchise's sales. Now, I'm going to keep this video real time. There's no tricks. You can see exactly how Genie works and thinks. And Genie has a couple different ways that it can access and think about data.

1:00 Ask a question

For example, it can use MCP connectors for live access, you know, reading from Slack, grabbing tickets from Jira, parsing my calendar, grabbing docs from GDrive. It can also access all the data that I have permission to in the Databricks Lakehouse. Now, in this case, I think I just saw that it found some potentially relevant ontology snippets from the Genie ontology. I'm going to click into that here. Yeah, it's actually found quite a few of them.

1:27 Reviewing Genie Ontology snippets

The Genie ontology is Genie's map of your business and its institutional knowledge, and it's really important because when you ask Genie about something like active customers, it actually knows what your organization means by that. Or when you ask for revenue, Genie can skip over the 40 random tables that mention revenue and go straight to the canonical source that finance actually keeps fresh. It's this institutional knowledge that a generic model often doesn't have.
Now, part of that context probably comes from the information that your data teams have curated deliberately in Databricks. Things like metric definitions, like how to slice and dice revenue, as well as qualitative concepts like what's the difference between incremental revenue and attributed revenue. But, that's really hard to scale to an entire organization. For example, the legal team might be doing something really valuable in the legal team's world for their legal colleagues.
But, the data team may not know or want to manage that. So, Genie also learns from our dashboards, our SQL queries, our Genie agents, and so on. And Genie actually ranks the information it has by how authoritative it is, meaning a certified definition outweighs some random one-off query that I wrote several years ago. So, just rather than just grinding through an agentic loop of trial and error trying to understand our data and eating up a bunch of tokens along the way, Genie just understands our ground truth and goes straight there.
Uh so, in practice, uh looks like I got my answer. Looks like Baked Blissed in in Rome is our is our best-performing franchise, and I can actually see the ontology snippets that Genie used to answer this question. I have the definition up here. I can see the original asset that it was derived from. Uh it looks like it came from a Genie space. It's got a bunch of different experts that are involved in this particular domain, which has to do with sales.
So, in general, this actually really helps for me to evaluate the helpfulness of the snippet and to understand really what Genie used to answer its question. Super cool. Um but, I have a few more questions. I'm going to just follow up wondering why Rome is actually ahead. Now, uh in a lot of models, you're going to end up basically reciting what lives in a document. But, you know, what's happening here is that Genie is also running some SQL.
And that's really critical because something like this, an answer to this kind of quantitative question, it doesn't live in a document yet. That's That's why I'm asking Genie. Uh so, Genie's able to not only recite what lives in documents, but also calculate results on the fly using live operational data. And of course, along the way, as Genie calculates the answer and as it looks through the ontology, it's entirely permissions aware.
It's enforcing the permissions of our data sources through Unity Catalog. So, for example, if we are pulling some information from a Teams channel, it's certainly not showing me information that I don't actually see in Teams. Uh so, do you look at that? It looks like it's actually not anything to do with the number of transactions. Rome's kind of in the same ballpark as the other franchises. It has to do with the order size.
Rome has a super high order size compared to the other regions. So, maybe we're running like a catering business there or something. Uh that looks super promising. Uh maybe I want to now get a chart of this, and maybe I'll throw it into a document as well so that it can be, you know, a nice shareable format for my teammates. And once Genie produces this document, I'm probably going to post it to maybe our team's Slack channel or or maybe I'll file a Jira ticket uh if something looks like it's going really wrong.

4:50 Creating a chart and document from a single prompt

I'm just going to skip to a faster answer because I really trust Genie here. Awesome. Now I have a document. This is pretty much ready for me to send out. Pretty helpful. Um in fact, it's so helpful that why don't I have Genie actually check on this for me every single day. I don't necessarily want an email every single day if nothing is wrong. Maybe instead I want this to run only when there's something I actually care about.
For example, email me when sales orders drop materially. So, I'm going to have it do that. >>

Learn more about the Databricks Data and AI platform.