Skip to main content

Why It's Time to Take AI/BI Seriously

Summary

  • AI/BI is Databricks' analytics suite comprising metric views, dashboards, and Genie Spaces, and this video argues—from the perspective of a Power BI practitioner—that it has matured into a legitimate challenger in the data visualization space.
  • Metric views provide a reusable semantic layer over data tables, and this video demonstrates building one from scratch, using it to create an AI/BI dashboard, scheduling email exports, and configuring a Genie Space for self-service analytics.
  • The presenter identifies where AI/BI thrives—particularly in governed, embedded analytics scenarios—and discusses current gaps and the roadmap toward competing with incumbents in the Gartner Magic Quadrant.

Why It's Time to Take AI/BI Seriously

Watch: Why It's Time to Take AI/BI Seriously
I'm a Power BI developer by trade, and for years I've refused to engage with other data visualization platforms. Every time I did, I just found it was lacking capabilities or had serious deficiencies compared to Power BI. Then 2025 hit and Databricks invested heavily in improving AI/BI, metric views and Genie. Now, AI/BI is ready to be a legitimate challenger in the Data Visualisation space. In this talk I will show you how you can use AI/BI, metric views and Genie to get the most out of your data. We will start from basics, building a metric view, using it to build a dashboard and then setting up a Genie space. I'll then walk you through my top tips on how you can get the most out of each of them. We'll talk about the types of work where AI/BI thrives and discuss where I see it going in its pursuit of being the next big data visualization platform. AI/BI is here and it's ready to be taken seriously.
Talk By: Ashley Warren, Senior Analytics Consultant, Advancing Analytics ;

Chapters

FAQs

What is Databricks AI/BI and what does it include?

Databricks AI/BI is the collective name for Databricks' analytics product suite, which includes metric views, AI/BI dashboards, and Genie Spaces. Metric views define reusable business measures over data tables, dashboards provide fixed-format visualizations, and Genie Spaces enable natural-language self-service querying of the underlying data.

How do metric views work in Databricks AI/BI?

A metric view is a semantic definition layer placed over a data table that defines measures, dimensions, and business logic for reuse across dashboards, Genie Spaces, and SQL queries. This video demonstrates building a metric view from sample data and then using it directly as the source for an AI/BI dashboard, ensuring consistent metric definitions across all analytics surfaces.

Why does the presenter—a Power BI developer—say AI/BI is ready to be taken seriously?

The presenter notes that Databricks invested heavily in AI/BI, metric views, and Genie through 2025, addressing deficiencies that previously made it uncompetitive with Power BI. The result is a platform that appears in the Gartner Magic Quadrant alongside Power BI and Tableau, with particular strengths in governed, embedded analytics scenarios where data already lives in Databricks.

What are the current gaps in Databricks AI/BI compared to Power BI?

This video acknowledges that AI/BI still has gaps relative to mature platforms like Power BI, which the presenter discusses alongside the roadmap for addressing them. The presenter's assessment is framed as personal opinion rather than official commitment, but the overall conclusion is that AI/BI has crossed the threshold of being a viable primary analytics platform.

Full transcript

[00:07] Thank you for coming. I hope everybody's having a great summit. Thank you for choosing to spend some of it here with me. Um especially thank you if you're an English football fan and giving up the second half of the World Cup match to be here. Oh, that was a frustrating moment
[00:20] when I realized that was the scheduled time. Um um this is the standard forward-looking statement. I'm sure you've seen all of this. Um, anything that I say in here is my opinions and is not a commitment from data bricks. Um, but yeah, we're going
[00:33] to get started. I'm Ash um or Ashley. I'm a senior analytics consultant with Advancing Analytics. Don't know if you can tell that from the from the football shirt. Um, we're a data and AI consultancy based in the UK. We're a
[00:45] data bricks gold partner, one of the smallest in the world. Um, and we help customers with all sorts of things, largely data bricks, but I'm a PowerBI person by trade originally. and I've come to data bricks sort of later in my
[00:57] career. Um so using a lot of PowerBI, a lot of data bricks. Um I'm a regular speaker at events um across mostly across the Europe. This is my first time speaking in the states. So thank you for having me. Um here's the agenda for
[01:11] today. Um I'm going to start by sort of giving you a bit of an introduction to AIBI. Um what it is and the different elements of it. We're going to build some of the first assets that you can you can start um using and then we're
[01:24] going to talk about where it thrives. And we're doing all of that to sort of set the scene and get everybody on the same page so that we can answer the main question that we're all here for, which is why it's time to take AIB seriously.
[01:36] Um I'm going to make sure we're all on the same point when I answer that question. And then I'm going to talk at the end about where I think AIBI is going to go. or speculation um at that point, but we'll we'll talk about where
[01:48] I think the gaps are and where I think we're going to see it move in the next year or two. Um so what is AIBI? Um it's the collective name for the data bricks analytics products. Um so there's a few
[02:02] other things as well as these, but these are the three main ones in my eyes. The metric views, the dashboards, and the genie spaces. Um so we're going to start with some metric views. A metric view is just a way of sort of defining measures
[02:16] and definitions over the top of a table. Um so if you're wanting to sort of reuse your um your your definitions across multiple different um dashboards or
[02:30] genie spaces or or even other assets outside of data bricks if you're using other visualization tools. You can do that with a metric view. gives you a standard um a standard format for that.
[02:42] You define it in YAML. Um but there is a a UI element which we're going to see in a minute. And then you use it to power everything else to make it quicker to build your dashboards, quicker to build your genie spaces. Um dashboards, I'm sure anybody who's here has seen a
[02:55] dashboard um at some point in their career. It's just data bricks version of that. Um the data behind it is a SQL um is SQL based um and we can then use that
[03:07] to make charts, make graphs. Um lots of different things there. And then the genie spaces, I'm sure everybody was in the keynote and heard about Genie over and over again. It's amazing. I love Genie. Um talk to your data. And the
[03:20] Genie spaces are basically just a contained version of that for a very specific area of your data. Um, so you load them with tables, you add context, and then they your users ask questions to it, and it writes SQL to answer those
[03:33] questions. So it's not just making up something off the LLM's context. It's writing some SQL, which it then runs against your tables to get you an answer. So demo, straight in with a demo. Um, let
[03:46] me just uh swap to this for a second whilst I switch over my screen. And hopefully data bricks hasn't logged me out. It hasn't. Awesome. So, I'm in data bricks. Um I'm going to use the Wonder
[03:59] Bricks um sample data and we're going to build we're going to build a metric view, a dashboard, and a Genie space. And I'm going to try and do all of that in 15 minutes. Um just to show you how
[04:11] quick and easy it is to build this stuff, right? And I'm not going to use Genie code to do it either, which makes it even faster. Okay, so um we're going to start from the bookings table. So, Wonder Bricks is kind of Airbnb. um
[04:24] that's that's what it's kind of modeled on. Um it's not their actual data, but it's a standard sample data set that's available in Unity Catalog. We're going to build from the bookings data for this one, which is a fact table of all of the bookings that they've had um across a
[04:38] given time period. And to to start building a metric view, we go up to the create at the top and then there's a metric view. Um I don't know if you can see that. It's quite small, but um it's an option in the create setting and we can create start creating our D our
[04:50] metric view. It asks you where you want to put it. Um, and I'm going to put that in a catalog I've built specifically for this. It's probably worth noting that this is databicks free. You've probably
[05:02] seen that at the top. So, you should have access to this in your environments wherever you are. It's just setting that up. Um, and what it immediately does is create that first table um with all of the all of the the fields in it and it
[05:15] adds a little count metric over the top of it. Um, my compute is just starting up, but once that has started up, you'll also get a little graph. Um, and I'm going to start adding in a join. So, I know bookings joins to users. Um, so I'm
[05:30] going to add a user. And it suggests what it thinks the join might be. Um, in this case, it's not quite got it right. It always seems to want to add created date on as well. It's just a user ID um that we're joining on here. Create the
[05:43] join. And hopefully now the graph will have load loaded in. Um I select the fields that I'm interested in. So I'm going to pick just a couple of them. User type, country. Um uh actually no, I'm not going to go country. I'm going
[05:55] to go company name. Um and we can see we're starting to get this now. So we've got the source here, which is our fact table. We've got our users, which is our dimensions. We can add other joins. You
[06:07] could add joins off of user as well. Um, I'm going to go with properties here, um, because I know that one's got, um, I know that one's got some something we can snowflake off of. Um, but I'm going to bring in maximum guests and the
[06:19] property type. And then I'm also going to join off of properties to destinations. So, you can see I'm just adding in these tables really quickly. Um, helps that I know what the data set looks like. Um, but so
[06:32] far we're at four minutes. Um, so you can see we're already get starting to get a an idea of what our think what our metric view might look like. We've got this little uh it's a bit small. Let me zoom in. Uh oh, it's not the right size.
[06:44] Great. Thank you, data bricks. That's uh the zoom isn't great. Um but um this is this is the UI preview. This is still in preview. Um but it's it's something I'm still using. Um you can go into
[06:57] dimensions. We can drop out anything we're not interested in. So I'm not interested in these IDs. they don't mean anything to me. Um, I'm not interested in the the the updated date, for example, and I can just delete those out. Those were the ones that were
[07:09] pre-made. Um, and then you can start adding your metrics in. So, it's automatically added me a count. Um, I am going to add in some new ones. Clicking the add button. Um, total revenue. It's suggesting some options here. Um, let me
[07:22] see if I can make that a little bit bigger. Um, okay. I'll have to zoom in myself. It's decided to create a total revenue. Um, I clicked on the total revenue and it's it's it's written some SQL for me based
[07:35] on the table that um we've got. We can start adding in some comments, some synonyms. So, um, we called it total revenue, but maybe we some people just call it revenue, some people just call it booking revenue. Um, you can also add
[07:47] in formats and tags. You can do all sorts um to do that. And you can preview it to sort of see how that will look out. And that's just running a very quick um aggregation over to all of the
[07:59] data for that SQL expression. So you can check that it works. And we can go and add I don't know if I actually clicked add there. Oh, I did. Cool. Um so we could add in a bunch more. So you can do anything anything you can write in SQL
[08:11] um as your metric you can put into the expressions down here. And then when we click save, that's created the metric view which will look in Unity catalog like a view. Um, it's got a slightly
[08:23] different logo to a view, but it's tiny, so you're probably not going to be able to see that unless you really like zoom in. Um, but it's what's actually happening is it's creating this YAML spec under the hood. So, that you that
[08:35] UI is for anybody that um I prefer using the UI personally, but when they first came out, it was all it was all YAML. So, you had to write this this specific setup here. There's a lot more that you can put into the YAML. Um there's
[08:48] additional fields and settings and things that you can cut into there that we haven't done in this um in this so far. But this is then you can then control this through data brick asset bundles. Um the automation bundles.
[09:00] Sorry, they've they've changed the names. Um so you can you can version control these because they're just text effectively that that's that's creating the the spec. That's a metric view. Um
[09:12] we're going to go over to a dashboard now. We're going to create a dashboard. How am I doing on time? Yes, we're on track. Um, so this has already started creating a dashboard. As I said earlier, you could at this point use Genie. Um,
[09:24] so if uh I could go and put my prompt in here and it would start building out a dashboard to do that. I'm not going to do that. I've been building dashboards for a long time. I want I know and I know what I want for this one. So I'm going to make this my booking
[09:37] dashboard. Oh, my typing has gotten terrible. Everybody watching. um you go to the data tab and here you can add in the data for that dashboard. So this could be all any tables you want that
[09:49] are in Unity catalog um and or it could be a SQL script that becomes the source of the of the data. So you could write a select star from or or wherever to to build that out. Um in the all you've got this would be your access to Unity
[10:01] catalog. I don't have a huge amount in my my Unity catalog at the moment because this is just the free version. Or you can go to the metric views and I've got a bunch of metric views in here. I'm going to use the one that I created earlier rather than the one that we did now. They are entire pretty much
[10:15] the same. It's just I know this one works and I've not got enough time to debug it if it doesn't. Um, but it should be entirely the same and it's run and you can see here it's it's got the fields that I've chosen. It's got the measures and it's by this point it's
[10:28] taken away and it's abstracted all of that sort of join condition and where it came from. But if I went back to Unity catalog and looked at the definition of this metric view, you'd be able to see that there. Um, so this is my data set
[10:41] now for my dashboard and I can just start building. Um, so I'm going to start with a I'll start with a little text box at the top just because then I can get the typing out of the way. Oh, brilliant. Thank you. Hit the wrong
[10:55] button. Uh, select all. Make this a big header at the top. Um, and then I can start putting in other things. So I can put in a filter if I wanted here. So there's this little button down here for filters. There's also text boxes. and
[11:07] there's visuals. Um, and then there's the genie code button as well. But for filters, you could then go over to the right hand side and it's got all of the fields that we've got in our in our metric view. I'm going to go with booking month. Um, which is just a this
[11:19] didn't exist in the data set. This is something I added in specifically which just takes the month um of the the check-in date um for the um for the the record. Um, we can then start adding in
[11:33] visuals as well. Um, so I'm going to click the visual tab and then I can choose my visual type here. Let's go with I I hate myself for doing this. Let's go with a pie chart to start with. The first visual I'm making in AIBI is a
[11:46] pie chart. I would suggest you don't use them unless you get unless there's a specific reason to. Um, but I'm going to choose uh I'm going to choose here there's two different options. Uh, so lots of different options along the side, but everything that you can amend has got a little plus. So I'm going to
[11:58] choose angle, which is the the size of the pi effectively. And I'm going to do that on total revenue. Um, it's going to load. It's going to take and that's really helpful that it's just an empty blank, but I'm going to turn my labels
[12:10] on and I'm going to switch my labels to field. Now, we can see it's that 40 million. But if I want to split it by different areas, I can start putting into this color section. And let let's go with booking status. Um,
[12:24] awesome. It's now added in um added in those sort of uh statuses. And you can change the colors down here as well. One of the things that's worth noting with AIBI is that you can you can add in a C like a template a brand template to set
[12:37] your colors. Um, but you do still if you want to always have if if I was to create something else that used booking status with colors and I always wanted my cancelled to be I don't know why I'd want my cancelled to be green, but it is
[12:50] in this case. If I always wanted it to be green, um, then you can set that. Um, you have to set that individually on each visual, but we can add in a bunch more here as well. So maybe I'm going to put in a counter and I'm going to put in the date as the booking month and I'm
[13:02] going to put in the value as the the total booking. You just then grab slice change the size. It's it's automatically like snapping to a grid. So underneath here
[13:16] there is a grid. Um and it's just going to snap the data to that. We can add in a bunch more things however you want to lay this out. Um you can go and chuck them in here and you can start adding in tables. you could start adding in any of
[13:29] the fields that you want. So maybe I want to see property type by country, total number of guests, and that will slowly load in. Um obviously hopefully um when uh when the Raiden engine
[13:41] becomes available, as you saw in the keynote earlier, that'll be even faster. Um so you that will just snap. But this is off of my compute which I'm setting up here. So this is on my serverless. Um but if you have specific compute you
[13:54] wanted to use this to, you can change that. When I'm happy with this, uh, which I probably should change the page name. Um, you can add you can add other pages as well as the note on this. Um, you can
[14:06] then hit publish and you've got two options here. The one on the left is that it's going to use my permissions for Unity catalog. So, anybody I share this to gets to see this as though they're me or the one on the right is
[14:18] that anybody who sees this sees the data as though they're them. So if they don't have access to that data or you've got some masking happening uh in Unity catalog or you've got some sort of row filters that the that are on there on
[14:30] those tables based on their role then that will um they will inherit that based on that um setting. Um and then it's published. So there's a little view published at the top now. So we can swap
[14:42] over to the published dashboard. Um unit uh dashboards they they refresh when you tell them to. Um, so there's a little one here, six, seven seconds. It's counting up. Um, I could just hit
[14:54] refresh and that would refresh and bring the data through. It's running off of your compute. Um, and it's going back to Unity Catalog. So, if you're using streaming data, um, when you hit refresh, it brings all of that. It just reloads all of that data. And if your
[15:06] queries are super quick, that can come in really quickly as well. And then you can schedule it. So, I can set a schedule of I want this to run at, you can see my time. It's currently 10 10 PM back home um, for me. Um but uh you can
[15:19] set that to run whenever you want. You could you can also schedule that to using cron. So you can have that setting so that it refreshes or whatever cadence you want it to refresh at. You can choose which like SQL warehouse you want it to use to do this. And you can also
[15:34] um set it up so that it sends an email to a person. And on that email you can even include attachments like a a like a p like a PNG of the um of the page. So like a little visual image version of
[15:48] the dashboard or even if you wanted to you could take including the data. I didn't name these widgets. I probably should have but running a little behind already but you could click I want to add this one in and it would also send a
[16:00] CSV file with all of the data related to that pie chart in this case to the person. So that's that's a quick overview of dashboards. We're going to do a very similar thing for genie spaces. So I'm going over to genie. Um
[16:12] we can create a new genie space and again I can click to do that from a metric view. Um you can also build genie spaces from individual tables in Unity catalog. Um but it immediately from when
[16:25] you use a metric view it already has pretty much everything it needs or it should have most things that it needs to be able to start using um using the data. Um you can go you can and you
[16:37] should go into configure and add more context. So genie spaces need context. um they need to know about your business at the moment at least. Maybe ontology will change that. Maybe it will automatically get that. We'll see where that goes uh over the next six months,
[16:51] year as they as they roll that out. But we can add in a description of the space for people. And again, the Genie is just this is Genie code adding in a little bit of a description and you it can suggest based on what data is in the
[17:04] metric view what questions might be interesting. So I'm going to add I'm going to add this one. I'm also going to trigger it to start to ask that question by clicking on it. Um, but I'm going to add some other common questions whilst we're at it. Um, I'm doing that because
[17:18] it takes a little while for these to run sometimes. Um, but I'm but but focus on the right for now. I'm going to just move that away. So, uh, it's just going to be there so that it's running that query whilst we whilst we look in. This
[17:30] is our metric view. Um, we could go into that and if there's anything in the metric view that you want to change for this ve for this specific genie space because your metric views might be used across loads of of different spaces or
[17:42] dashboards, you can. Maybe you don't want this genie space to have average lead time. So you just hide it. Or maybe in this specific context, maybe this is one department and they call total revenue something out like sales amount
[17:57] and you can add in an extra synonym and it will learn then that when they say sales amount they mean to use total revenue. Um you can add in general instructions. Um I'm not going to do that to start with. I'm going to go and
[18:10] have a look at some of the other tabs. Um, because the the general text is the least effective of all of the different things that you can all the ways you can give genie context, the general text is the least effective. So, it's the last
[18:22] one that you should use. Um, joins. If you had multiple tables or multiple metric views, you could tell it how to join those metric views together or how to join those tables together. And it's very similar to what you saw in the
[18:34] metric view. You just choose a table on the left, table on the right, and the join condition. Or you can write that as SQL as well. And then there's SQL expressions. So maybe in this case I want um in this case I want everyone to
[18:48] know like or the genie space to know how to define the United Kingdom. And then I'm just going to write some uh write some SQL to do that. And it will be oh where is it? Country is equal to
[19:03] United Kingdom. And again I could then add some synonyms. And it's not just the United Kingdom. Maybe it's Great Britain. Um or maybe it's the UK. And there's lots of different words you could put in there for for for the
[19:15] country that I come from. But um whatever the field is, if you've got different definitions, make sure you put in the synonyms so that Genie knows um what it's doing. And you can even give instructions if someone wants to know
[19:29] how many of bookings are related to the United Kingdom. So, Genie now knows what the United Kingdom is or this Genie space does at
[19:42] least. You can also add in SQL queries. So, if you have specific questions um that that you know you need a specific SQL answer to, you can add that in. You can hit add, write your what is what is the question and what is the answer
[19:54] effectively. Um Genie will over time as it gets used. Um it can sort of suggest what questions it thinks should go in there. Um, the best way to do that is for you as the person developing it to
[20:06] just ask a bunch of questions to the genie space that you think your users will ask and eventually it will start saying, "Oh, this this question seems to be asked a lot. Should I should I use the SQL? Are you happy with this SQL that that I've been answering it with?"
[20:19] Um, you could also add in your general instructions. So, maybe in here you've got a two different types of revenue. Um, one of them is the f financially recognized revenue and one of them is the sales revenue. basically around
[20:32] accounting differences. Um, and you want to tell it when always use the always use or the version A or version B. Maybe it's local currency versus U US dollars version of the um of the of the revenue
[20:46] and you want to tell it always use the US dollars one. Um, you can put that in here. But as I said, this is the least effective version. Whilst we've been talking, Jeie had gone off and asked answered the question. So it went and uh it went and answered the question of
[20:59] what distribution of booking sales values we've got. So how many different uh bookings were across each of these statuses and you can go in and see the code that it wrote. Um this is not
[21:11] always going to be perfect but that's that is the way with Genie. Sometimes sometimes good enough is good enough and this has kind of got the answer but I could be having a look in there and just checking that that works. If this is more complicated you might need to do
[21:23] that. Um, and it gives you a little summary at the bottom and slowly you'll build up more context with these over time. I am slightly behind schedule so I'm going to move on from there. If anybody has any questions about anything I've talked about in this then I'll be
[21:36] around afterwards and you're more than welcome to come and chat. Um, I I would love questions but let's go back to the slides. So where do we use AIB? Um, so each of
[21:49] these three things you saw there in oh that ended up being slightly more than uh than 15 minutes. I think that was 17 minutes. Um, we've got we built a metric view, we built a dashboard and a genie space. Where do we use them? Okay, so
[22:02] metric views they're really good if you've got a single fact table in your mod semantic model or in your model um your data model. That is not always going to be the case. But if you do have a really simple model, genome metric views really good. Or even if you've got
[22:16] lots of a model that's got lots of different fact tables in it, which aren't super interlin, metric views are brilliant. Um, if you need to use them um across multiple different assets, so if it's a oneanddone metric view, um
[22:29] it's just going into one dashboard, then that's maybe going to take more time to build your metric view than than to just build the dashboard. But if you're going to reuse that fact across lots of different areas. So sales for example,
[22:41] um sales data gets used in all sorts of different domains, all sorts of dashboards. Build it once, reuse it everywhere, and then you've always got a consistent definition of what total revenue is. Similarly, if you've got a
[22:54] really complex data estate, sometimes it's easier to to just build the metric views, even if you're not intending to use it everywhere, just having the metric views all lined up. And then you can have your users just using metric views instead of using the the
[23:07] underlying tables. And they get the met they get the measures for free. Dashboards. Um my personal favorite use of dashboards is for reporting on anything that happens within data bricks. So u my colleagues build some
[23:20] amazing data pipelines. Um and they want to track costs. Well, build a dashboard. If they want to track the data quality of the the data that's going through that pipeline, dashboards really good for that. Just sort of a they've run the thing. They don't even have to jump out
[23:32] of data bricks. They just go to a different tab, see the dashboard. Oh, look. It's great dash the the pipelines worked. We use this all the time. Um, it's really helpful. Anything low latency, especially now
[23:44] with Ryden, um, the the new engine that's coming, um, data is going to be able to process really quickly. If you're using stream data, having to take it out of data bricks and putting it in another visualization tool is uh is
[23:58] going to take longer than just doing it in data bricks, especially in the future as we go forward. And then anything that gets routinely asked to a genie space if if users are using the genie space and they always want the exact same flow of
[24:11] questions, you're probably going to be better to just build a dashboard for them um so that they can routinely get that. If they're always asking what's my top five customers, what's my bottom five customers and they just need to know the names, put it in a dashboard because the genie space takes a little
[24:24] bit of time to answer the questions. Genie spaces, they're really good if for self-service, right? So, especially for the nontechnical users, we've seen that with with uh with Genie one and all of the announcements that have been going
[24:36] on in the keynote, the intention is that Genie is is there for the nontechnical people to be able to answer their questions. It's also there for all of us as technical people as well, but it's especially good for for self-service.
[24:48] Um, it's really good for explo exploring things. If you don't know what flow you want to take through your data to to work out what's going on, Genie is really good for that. Especially now with agent mode, it can do some really complicated reasoning, can do a lot of
[25:01] really, really clever things. Um, I've seen it um with one of my clients where it was able to infer things about their business um from from looking at some of the tables that I didn't know as a
[25:14] consultant. It was able to work out the link between certain products to then um to answer certain sort of sales questions related to those products. And it's also really good over the top of dashboards. So, I didn't show you there,
[25:26] but when you go to a published dashboard, you can now pop up a little genie um there and it'll it's basically creating a genie space over your dashboards for you for free. So, if somebody is uh somebody is maybe use the
[25:38] dashboard and then they've got follow-up questions that the dashboard's not built for. There's a little chat box that can pop up in the little in the bottom right now um that that they can ask those questions to. So, they don't have to come back to to me to build more
[25:50] dashboards for them. It's just there ready to go. Genie one. Um this has had the app on the right here has had three names in the space of three months. Uh so went
[26:03] from data bricks one to data bricks genie to data bicks genie one. Um it's but it is the future. Um you heard about it in the keynote. Um the idea of getting data into the hands of your users. So this is kind it's not AIBI
[26:17] specifically but it is kind of very closely related and it's a way of getting into users into AIBI. So they can go and ask their question but then they can also pop out and go to the
[26:29] dashboards and see all of their dashboards and they can do it on their mobile. Now you can do it online um in the UI as well but the intention is for mobile. you saw Ally on the stage with his mobile at multiple times like clicking buttons um on that and that
[26:42] that's the that's the idea and Genie's general mission I took this directly from the launch um the launch announcement of Genie 1 um back back when it was first announced a few months
[26:55] back was to make data AI and insights accessible to everyone and that bottom bullet point I wrote about a month ago um I I think you'll have seen from the keynotes this week that they're definitely delivering ing on this idea
[27:07] of uh of developing this further. It is a key thing for data bricks in my eyes to make sure that they the users are getting access to their data. So some tips for AIBI
[27:25] dashboards. Pick the right visualization type for the information. Doesn't always need to be the fancy one. If it's if the bar chart is the right thing to do and you've got 12 bar charts on your page, do that. Don't just make it a pie chart for the sake of it. I did in the demo,
[27:37] but that was a terrible idea, but that was more just to give you an idea. I hate pie charts. Um, yeah. Um, follow a logical layout. People look top left to bottom right. So, put the important stuff in the top left. And also make it
[27:50] bigger. Um, if you do that, people will look at it and for longer. Um, don't add anything to the page that that doesn't help you tell a story, right? So, if the point of the dashboard is to be um about sales, don't add some random data about
[28:03] something else, right? Or don't add stuff that isn't going to give you directly u an answer to a question that you think a user will need. Use colors if they convey meaning only if they convey meaning. There's no point just putting loads of colors on there and
[28:16] making it confusing for no reason. And try and be consistent with color. So, if you have your status and um cancelled is always is red, then make it red everywhere, right? Don't have it red one graph and green the next graph. It's
[28:29] really confusing for everyone. And generally for dashboards, and this applies for both AI, BI or anything else, there's this concept of the 330 second rule. I borrowed this from a group called SQLBI. Um they're a PowerBI
[28:42] um consultancy, but um they they go with this idea that you should be able to tell what a dashboard is going to tell you within 3 seconds of loading into it. Not very long, so it's got to be super obvious. within 30 they should be able
[28:55] to filter and get it down to whatever detail they need to and within 300 seconds which is five minutes they should be logged out of that dashboard. Right? If they're not it's probably too complicated. Right? Dashboards are not meant to be for stupidly complicated um
[29:09] use cases. They're meant to be easy to use. And if that's not what your dashboard is for, maybe it's too complicated and you need to tone it down. Genie spaces. Um my thoughts on
[29:21] Genie spaces, they are amazing. They're an incredible technology, but almost all of the problems I've experienced with them are related to the people and the processes around them, not the actual technology. Um, so there's always talk
[29:33] about accuracy, but really the problem there is that you've not got somebody actually looking after the genie space and making sure it's accurate. It's not the tech like the tech is uh the tech is intelligent enough as Alli was saying
[29:45] earlier. It's the context that it's not getting right and you need to make sure you've got a governance strategy that's going to enable it to get right to get it right. So who owns and is accountable for the genie space positives and negatives right? they've got to be able
[29:57] to take on when it does get something wrong that okay yes apologies it got something wrong but uh we we learned from that we fixed it and it's going to be good next time but look at all of these amazing things it's helped us
[30:09] deliver right accountability right you cannot just take have somebody who's there to say this is amazing who doesn't take on when it goes wrong and take responsibility for that similarly you need some you need to know who is going
[30:22] to be responsible for updating the space who is going to verify its accuracy and who is going to support your users. They need to know what they can and can't ask of the space. They need to know what they can and can't do with it.
[30:34] Um, so you've got to be making sure that if you're going to roll a genie space out, you've got answers to all of these questions. If your users don't trust the genie space, they won't use it. And all of that development time that you put into it is useless. And if they don't
[30:47] have the answer to these questions, if they can't easily find the answer to them, then they will probably come to stop trusting the genie space and stop using it. So we should all be on a level pegging
[31:00] hopefully. We've got nine minutes. Um so why it's time to take AIB seriously. The point of this talk, right? Um so this is the Gartner Magic Quadrant for 2025. Um,
[31:14] the 2026 version is due to come out any day um at all. I was hoping it would either come out beforehand or I'd have got some like whispering of what it is, but I haven't. This is from just over a year ago, so keep an eye out for the
[31:27] next few days. It normally comes out in June. Um, so Microsoft have been in the top right of this graph every year since 2018. Number one products. PowerBI number one product in the space. Um,
[31:40] Tableau and Click who are in that top group. um neck just behind Microsoft. They've been up there competing with them since 2018. Um AWS, Google, and ThSpot have all massively improved uh
[31:54] since 2018. You can see these red lines are from 2018 for Google and spot or 2021 for AWS, which is when it made its debut on the chart. And they've accelerated up uh into into good
[32:08] positions in the in or around the leader space. Um those are the six tools I'm kind of going to focus on for the rest of this um as the main competitors. But where would data bricks isn't on the quadrant. It wasn't on the quadrant in
[32:21] 2025 despite almost all of AIBI being released before June 2025. Um Gartner had obviously pre-committed everything and not had a chance to to rate it yet. But what I would be placing data bricks
[32:34] personally somewhere in that green square. Now, that doesn't sound doesn't look that impressive, but that would be the best ever debut for any tool onto any version of this magic quadrant going back to 2008.
[32:46] I think I went back to um nothing has ever debuted higher than that other than a rename. That's where I'm personally am putting data bricks and I don't foresee it regressing from there. I think it's only going to get it's only going to go
[32:59] further up and up into that corner. So, it's not perfect yet. Um but it's definitely to me that would be an amazing start for for for this and would project onwards. Um again speculation
[33:12] some of my colleagues have put it in slightly different places based on where they think but that's personally where I think there's an amazing ability to execute everything. Almost everything I've ever seen data bricks do they've done really well. whether they've been ambitious enough um at different stages
[33:26] something is a different question but um yeah but why is that good why do we take that seriously why do I think that's improving across those six tools that we saw especially these three red ones that
[33:40] I've pointed out that have improved and Microsoft who've stayed in the top right and the whole thing is sort of set around um there's two key trends across all of those tools that have got better and
[33:52] this. So, one of them is the is the analytics being close to engineering and it should say AI as well there. I don't know why I dropped that off, but obviously we all know data bricks is amazing at an amazing engineering tool. It's an amazing AI tool. It's been one
[34:05] of the leaders in both of those categories in the the the equivalent Gartner quadrants for a very long time for both of those and having all of that next to your analytics is really useful.
[34:17] Um, and the other thing is that this is being built with AI functionality from the start. Um, I'm a BI person. I'm not an AI person, but the what you can do with these um, dashboards with with
[34:30] Genie spaces and how quickly you can build it with things like Genieode, how well integrated the Genie spaces, for example, are with the dashboards, the ability to to get answers from your questions. That's not something that
[34:43] every other BI tool has natively built in. Data bricks have built that from the start. They've always been about that. Um, so that in my opinion though, data
[34:55] bricks are the only one on that list. You can maybe argue Microsoft as well that are in both of these camps where their tools were built for AI and where their platform is an excellent
[35:07] engineering and AI platform. So to me that immediately shouts that this is going to get better and it's going to probably accelerate towards the top of that leader chart as they as data bricks invest more and more time and energy and
[35:19] money into making this better. Um it's in a really really strong place in my opinion to to kick on. It's a good start. You saw how quick and easy it was to build, right? That's that's an amazing baseline. Um there's some things
[35:32] we're going to need to do to make it better. Um and then that's where I'm kind of going to go with this. This is my personal wish list. Um, this was all written about a month ago. Um, so I think metric views need to start getting
[35:44] the the multiple fact tables, the ability to to do cross fact table joins for the measures. You need to have some ability to have custom field sorting. So sometimes when you've got dates, um, it sorts alphabetically. If your dates are
[35:57] like months, I want to be able to sort like you can in PowerBI for example, where you can say that January is one, February is two, and so it actually gets them in the right order. Um, premputee. So this is something that's sort of
[36:10] roughly been coming for a while. Um, so with a metric view being able to say, well, this field I always aggregate. So we can premputee that um that we need to know the sum of that field, for example, and it can it can then not just have to
[36:22] do that at runtime. um that might become less important with Raiden. Who knows? Um interactivity. So there other BI tools. This is the thing that normally
[36:34] people try and put AIB down for. Um the interactivity isn't great at times. It's getting better. Definitely getting better. So drill through bookmarks, being able to have custom UIs and things like that within your dashboards. Um
[36:48] more options for filter for formatting. Um better UX for discovering. So, um, if once you if you've got hundreds of dashboards, it can be a bit of a difficult thing to sometimes tell which one's which and how do how do I go into
[37:00] it, for example, for nontechnical people. Um, I think Genie does a good job of benchmarking and monitoring, so letting you track the accuracy of it um, already, but it I think it could do more
[37:13] to help people automate that and do better with that. Um, I didn't show that in the Genie spaces, but hopefully that's something that's coming. And then um there are some throughput limitations to Genie at the moment. So at massive
[37:25] scale um sometimes the you you can only send a certain number of questions to a genie space. I can't remember the number off the top of my head in minutes. Um but sometimes that's a bit low for massive massive companies. But I've
[37:39] added these little stars because these are things that have been already announced since I wrote this slide a month ago, right? And I know that as well that there's probably some other things that are in the works. The the
[37:51] they are investing a lot of time and energy into this this this formatting. There's custom formatting. They announced last week, end of last week. You can now build your own completely custom visuals inside um inside AIBI
[38:04] dashboards using Vega, which is uh Vega light, sorry, which is something that PowerBI developers have been doing for a for a while themselves. We've got premputee coming. It's It was actually there. If you were spotting it earlier
[38:17] on when I was building the metric views, there was a little tab for it. It's not perfect yet. There's still some kinks to work out for that. So, uh, creases, sorry. Um, I don't know if that's a word that translates over to the US. Um, but, uh, yeah, there's other things in here
[38:30] as well that I think are going to be coming soon. Uh, so I've got one and a half minutes. This is my summary slide of why it's um, why it's ready to be taken seriously. So easy to build, right? We built a metric view, a
[38:42] dashboard, and a Genie space. They weren't perfect. They weren't very good. But I did it in 15 minutes on stage in front of everybody, right? Like without a huge amount of prep for that. You can build these things really quickly. It takes minutes, hours rather than days,
[38:55] weeks to start building if you've already got your data in your tables. It's also um existing next to your engineering and AI. So there's no tool silos. You don't need to have people jumping from one product to another. you don't need to worry about um
[39:08] integrations of pulling data out of data bricks into PowerBI or into Looker or wherever your your um visualization tool of choice is. It's easy to govern. Metric views let you um let you set your
[39:21] metrics once or your measures once and then just reuse them across everything. You saw we reuse the same metrics in a dashboard and a Genie space. I didn't have to write them twice. And Genie spaces have I didn't cover this in the
[39:34] talk, but in my opinion for all of the other tools that are out there and that do similar to this that I've used, Genie Space has the best for sort of ongoing maintenance and keeping on top of them. I I really rate Genie Spaces routinely
[39:48] having arguments with my uh my Microsoft focus friends around Genie Spaces versus the equivalent tooling in Microsoft. And they're really easy to use. Um, so, uh, yeah, I'm running out of time, so I'm not going to cover everything on that
[40:00] slide, but you can do everything from your mobile. Um, so thank you for listening. Um, I post stuff on LinkedIn every Monday. Um, please feel free to add me on LinkedIn. If you want the the slides from this talk, um, the the one
[40:14] on the right will do that. It'll we'll ask you um to click accept that you're going to share the information to my marketing guy. He's not going to message you about it. It's just so that he can track how much traction we're getting with this. Um, and remember, please
[40:27] finish your surveys. Uh, complete your surveys. Uh, I really appreciate if you told me if you like the session or not. Um, if there's things you think I missed out. And yeah, thank you everybody for coming. I'm going to go down to our
[40:40] booth. Um, so Advancing have a booth down in the expo, um, which is next to the 3. If anybody's got any questions any point today, I'll be there the rest of the day. Please do come and ask me them. I'll also hang around here for a
[40:52] few minutes as well. So, thank you very much.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.