Genie Deep Research: Open-Ended Business Analysis

0:04Thank you, Ken. All right, let's take a look at all this in action. Huh. So, we're going to start by taking a look at some of the fun stuff we've done for dashboard authors. Then we're going to zip over to take a look at Databricks one for consumers and then we're going to close out with Genie Deep research. I know Ken gave you a little bit of a peek at that, but we're going to look at it a lot more closely. So one of the big updates in the last

0:26year is our own internal marketing team has adopted uh AIBI wholesale both dashboards and Genie. So we're going to look at everything today through the lens of their data anonymized of course. So let's start by popping over to the author view. And the first thing we can see is that Ken was telling the truth. We now have custom themes. I can change every color, bell, whistle, font, bevel edge to match

0:49exactly what my business expects to match the rest of my brand. We also have a special uh visualization type that allows me to do AI forecasting with just a click of a button. This is using a profit model under the hood. The same AI functions that you're used to using elsewhere in Databricks. We just make it a lot more accessible. Now, these two things are great, but you know, another element that's just so

1:11critical to every BI author's tool chest is the ability to author custom calculations. So, let's take a look at that. Pop this open. You can see this is just some of my regular SQL. I'll make it a little bit bigger. You know, all I'm doing here is selecting business unit account vertical. Remember, we're looking at marketing data. Some DBU dollars, monthto date. Um, but if I come peek over here, I've actually done a

1:33couple custom calculations on this data. So, I have added a dimension that is just bu by vertical. You know, maybe it's another slice I want to take a look at things through. Um, and then I've got four measures. So, if I pop those open, what I'm doing here is just calculating what percentage of dollar DBUs each fiscal quarter are coming to us from each region. Super simple, but nice to be able to do those calculations right

1:55in place. I get all the benefits and interactivity by doing that on the same data set instead of exploding them all over the place. But what's so important and what is so critical and part of what Ken just announced is this isn't just dashboard local anymore, right? I can create a metrics view directly from within my dashboard and oh, the warehouse is starting. We're in test serverless. Uh, and you can see that I have my source is exactly that SQL

2:17query. I also have those regular dimensions I selected in the SQL as well as the custom calculation I added on top. And then of course, all of my measures are there. I'm going to say, yeah, let's add this metric view to the dashboard because we want to use it there. And let's why don't we pick where we already have the rest of our data just to keep it all organized.

2:40And we'll go ahead and hit create. So what this is going to do is create a new metrics view in Unity catalog so that now everybody else in my organization can leverage that same logic and apply it in their own artifacts. So everything down to the syntax comes through to the comments. It's now a normal UC securable. I can govern like any other data object in UC

3:02and everybody who uses it is going to reap the performance benefits as well as the single source of truth reusability pieces. And if I pop back over, we can see it got automatically added to my dashboard. Bob's your uncle. I'm off and running. But this is all great stuff for authors, right? Uh we're Databricks. We love doing things for people who love to code. You know, everybody who dreams in Python and R, those are the people we've

3:24always done stuff for. But Ken just showed us Databricks one. And this is a complete reimagining of what Databricks can be for business users in your organization. So when you come into data bicks one, the first thing you're going to see is that this is not the traditional datab bricks workspace that your data scientists, data engineers know and love, I'm presented immediately with this easy search box. If I want to search for any assets I have,

3:47importantly, I can flip over to ask Genie. If there's something I want to do an analysis right on that page, I don't have to go search around, find the right spot. I can just get right into asking and answering my questions. You know, one of the things that we heard from so many of you that is challenging in BI is finding the needle in the haststack. So rather than expansive lists, we've actually really leaned into a couple curation mechanisms to help with that signal to noise

4:10problem. The first one is this for you page. And you can see that it's really a curated list of objects that are most critical to me. They're maybe from my teammates. They're recently shared. I've recently accessed them. We've also leaned heavily into this concept of domains which you learned about earlier in the Unity catalog talk. Now domains allow me to organize groups of things by semantic, you know, re relevance. So I'm

4:33no longer going to some list of 40,000 dashboards, some list of 40,000 genie spaces, and then trying to find an app URL somewhere else. I have it all organized for me centrally. This is of course a simple one for our demo, but I can find everything that's related to each other right in the same place. I'm not digging around everywhere anymore. So let's open the dashboard. You know, one of the things we really want to make sure is that anybody using AIBI

4:56dashboards reporting sees all the classic BI capabilities that they know and love, right? So last year we took a look at some of this awesome cross filtering. That didn't go away. It's still there. You know, of course, one of the most important things is offering drillable filters. So we have that as well. And you know, what is a dashboard if you can't drill? So you can now add drill to your dashboard. carry through

5:19that kind of setting from that original viz. I know. Me, too. I'm excited. Yes, that's a big one. We've been waiting for it. Um, and you can see I've got my quarter set to Q2 and and APAC and I can run around. And you know, of course, this is great. We expect all this from a BI tool. Like, we're glad data bicks is providing it. We know data bicks gives us genie, but you know, Ken told us this

5:43year was all about kind of getting to a new class of questions, right? Last year, we're all about what happened. Make it easier for me to understand kind of different slices and dices. But this year, what we're actually able to do is start to ask why of our data. So, you know, if I'm looking at this little marketing impressions chart, I can see there's kind of a uptick in the slope right here. Normally, I'd want to know why. I'd dial my favorite data scientist, you know, Katherine Maggie,

6:06shout out, and I would ask, you know, what what's going on here? Can you tell me what's contributing to this? But now, I can just go ahead and ask Genie to explain this increase. It's going to immediately d drop drop into what we call a top driver summary. And what it's going to do is kick off all of the questions that I would normally ask Maggie and Katherine to help me out with. It's going to say, okay, what is

6:28the kind of change? What comprises this different change? And pretty immediately, I can see, okay, Apac marketing was all over things. Like under the hood, there's actually a little bit of a decline there that's kind of obscured by the topline increase. But overall, this is all being driven by APAC, which is super helpful to know. And if I want to kind of switch by execution team, audience, engagement group, and kind of look at the before

6:50and after, I can see that. Now, this is all great. We love top drivers. We love anchoring kind of questions and what you can see in a dashboard. But what about when, you know, I'm just kind of asking and answering questions as part of my normal analysis. I'll make this a little bigger for our friends in the back. We've seats up front, too. Um, and let's take a look at if I was just using Genie

7:12kind of on its own. Now, this is our marketing engagement genie. Internally, we affectionately call it Marge for any of you Simpsons fans out there. And immediately, we can see that it is the genie we know and love, right? Lots of jargon, YouTube's net new names, excitement, MPDs, you know, any sort of abbreviation you can think of, our team's using it. And what I always like

7:34to do is just start by testing the genie is, you know, still working the same way we all know and love. So, why don't we ask you to, it's about marketing engagements. Let's ask about campaigns. Just make sure everything's working the way we want. So, what were my most and least effective campaigns in America? Genie is going to go do its thing. And well, okay, so this is good. It's still

7:56doing what we saw last year. If I say, you know, effective, it's not going to go scour the internet and tell me effective for some other company. I'm going to, you know, say, okay, how show me both in terms of attributed U2s, and I'm going to spell it correctly, and MQLs. So, U2s, MQLs, marketing qualified leads. I didn't give Genie any

8:18kid gloves. I threw all of the abbreviations right at it, and I expected to come back with something useful. So, yep, we're glad to see it. I can see, okay, this campaign awesome at MQLs. This one did a great job in U2's attributed and this one was a dud. Maybe we actually never ran it and I could kind of keep asking questions along there until I get the kind of conclusion I want. But what Ken promised

8:40us and what we just saw is that we're going to be able to start answering a whole new class of questions this year, right? Not just what happened, not just well what break it down by this category. Oh, slice it that way. Now, we can ask questions that are very open-ended and ask Genie to help us guide us in what the right plan is and how to tackle it. So, now I'm going to throw a question at Genie. I'm going to leave it how it was the whole last year

9:02and we're going to see what the experience was. So, how could we optimize our funnel? Now, the genie of last year is not going to be super helpful here, right? And yeah, question is too vague. It's good. It didn't try to make something up, but it's telling us, you know, it needs to kind of be led down the yellow brick road. you know, I have to say answer this, then answer that, then answer that. And this this is

9:26not super helpful or super efficient. Like, yes, I'm still not writing SQL, but we can do better. So, let's ask it again. So, how could we optimize, not improve, we're optimizing. Optimize our funnel. Okay. But this time, I'm going to put it in deep research mode. And we're going to kick it off. So, we're in purple now. So, we're in deep research mode. Much smarter. And Jeanie is going to actually

9:48come up with a research plan that we can review and modify if we want. So let's see what we've got here. Okay. All right. So based on our question, Jeie has the following plan. We're going to look at conversion rates at each stage of the funnel. We are going to look at the impact of account engagement status. Uh we're going to look at some product adoption and impact on funnel progression. And you know, I'm not our our marketing data expert, but this plan

10:11looks pretty good to me. I'm sure somebody else would have notes, but I'm going to say let's approve it and run. And notice we're still in kind of deep research mode. We've got the purple roll in there. And Genie is actually going to be kicking off all of the like I'm not telling it to do anything. All of these separate substeps of its research plan in parallel. So luckily I've got my trusty SQL warehouse. Okay, we're getting that one back. Um and as this

10:33analysis completes, it's going to actually be merging it and giving me a summary. Awesome. Okay, let's go. Miranda didn't have to do an analysis. Let's go. All right. So, it's telling us right away, you know, to optimize your funnel, focus on this. You know, it's giving me all sorts of details, which, you know, again, if this was on my shoulder, I'd be really in the details. But, you know,

10:56the most important thing is that it answered this question. It analyzed kind of where we need to focus it our efforts. It's giving us some suggestions for those efforts, but it's also giving us citations. So, if I click into any of these statements, I can trace back exactly what Genie did at that stage. I can look at exactly kind of what the research summary was for that stage, what the data was for that stage, and I

11:19can either have high confidence in the conclusion Genie's drawing because I can like see if I I agree with that data interpretation, or I can go back and say, "Hey, I actually think this is off. Let's rerun the plan." So, thank you for letting me come join you and share a little bit more about what we've been working on in AIB land. We are thrilled to have you take a look at Databricks 1 Genie Deep research as well as some of the other BI goodies. All right, with

11:41that, I'm going to have Ken come back. Thank you guys.