You Are Wasting Your Time! – Ep.531

Are you wasting time in Fabric doing things that could be done faster or smarter? This episode tackles the inefficient workflows and outdated practices that might be slowing you down, along with better alternatives you should start using today.

News & Announcements

Microsoft Build is happening next week, and Tommy will be there! While the in-person conference is smaller this year, most content will be available online for everyone to watch. If you’re at Build, stop by and say hello!

Two important Fabric announcements this week:

• Understand your storage with OneLake item-size reporting (Preview) — Workspace admins can now view detailed item-level storage breakdowns in Workspace Settings > OneLake > Storage Report. This shows visible, hidden, and soft-deleted data, helping identify which items are consuming the most storage. The report includes system folders and caches that previously required manual investigation or custom scripts.

• Fabric Jumpstart – Discover what’s possible with Microsoft Fabric — A unified catalog of ready-to-run demos, tutorials, and accelerators helps users go from zero to experiencing Fabric’s full power in minutes. The searchable catalog includes end-to-end accelerators, interactive tutorials, instant demos, and community solutions—all deployable with a single Python command. This addresses a major gap for organizations transitioning from “what is Fabric?” to “how do we do Fabric?”

Main Discussion

Topic: Stop Wasting Time in Fabric

This episode’s central theme is about identifying and eliminating time-wasting activities in your Fabric environment. Mike and Tommy explore common inefficiencies, outdated workflows, and better practices you should adopt to accelerate your Fabric development.

OneLake Storage Visibility

The new OneLake item-size reporting feature is a game-changer for capacity management. Previously, understanding which specific items were driving storage costs required manual investigation with Azure Storage Explorer or custom scripts. Now, workspace admins can:

• Generate detailed storage reports in seconds
• See breakdowns of visible, hidden, and soft-deleted data
• Identify problematic items that are consuming excessive storage
• Understand system folder and cache usage

This eliminates the manual detective work previously needed to track down storage overage charges and helps teams make informed decisions about data management. The feature is particularly valuable as workspace settings continue to expand—what was once a simple menu is now a comprehensive control center for Fabric workspaces.

Fabric Jumpstart: Dashboard in a Day for Fabric

Tommy draws a powerful comparison between Dashboard in a Day and the new Fabric Jumpstart initiative. For Power BI adoption, Dashboard in a Day was transformative—you could download a zip with all artifacts and be training an organization in two hours. Fabric has lacked that “ready-to-run” experience until now.

Fabric Jumpstart provides:

• Accelerators: Fully built, end-to-end solutions for common patterns like real-time billing systems, stateful streaming lakehouses, and platform monitoring
• Tutorials: Guided, hands-on learning experiences built directly in Fabric
• Demos: Lightweight, rapid experiences deployable in minutes for capability exploration

The catalog includes both Microsoft-sponsored core solutions and community-driven innovations, making it easier for teams to explore patterns without stitching together samples or building from scratch. This is especially relevant as organizations move past the “what is Fabric?” phase into practical implementation.

Adoption Curves and Getting Started

The conversation highlights the natural adoption curve for Fabric—approximately three years in, organizations are shifting from understanding the platform to implementing it. This aligns with typical technology adoption patterns where early curiosity gives way to practical deployment questions.

Key insights on the adoption journey:

• Initial phase: “What is Fabric?”—exploration and understanding
• Current phase: “How do we do Fabric?”—implementation and scaling
• Gap to address: Easy, ready-to-run starting points for teams

Fabric Jumpstart helps bridge this gap by providing validated reference implementations and working examples that teams can deploy immediately, accelerating the learning curve and reducing time-to-value.

Looking Forward

Try out the new OneLake storage report in your workspace settings to identify optimization opportunities, and explore the Fabric Jumpstart catalog at aka.ms/FabricJumpstart to find ready-to-run patterns that can accelerate your next Fabric project.

Episode Transcript

0:01 Dance to the [music] day to laugh in the mix. Fabric and A. I get your fix. Explicit measures. Drop the beat now. Pumpkins [music] feel the crowd. Explicit measures. Tommy, good morning. How are you doing today? today? Happy Memorial Day, dude. Welcome back to the Explicit Measures podcast with Tommy and Mike. Very happy to have everyone back here. excited to see

0:32 everyone back here. excited to see everyone and welcome back to the show. All right, jumping in today. Our main topic is around stop wasting your time. So, what things are you doing today currently in fabric that is taking a long time? What should you stop doing and maybe what you should maybe what you should start doing? So, we’re going to go through that and unpack where we think there maybe are better opportunities for you to speed along and move quickly through your fabric environment. Okay, that’ll be our main topic. Tommy, let’s go over to some

1:02 topic. Tommy, let’s go over to some news first. Do you have any news that you want to bring to us today? We do. We have two fabric updates as right just to go through right before build, right? Yeah, Yeah, build’s next week. Build is next week. I’ll be there at Build. So, if you are at Build, make sure you come by and say hello. It’s supposed to be a pretty big event online and there’s not a lot of people at the conference. It’s like a smaller like localized conference, but a lot of the content will be published and online. So, you’ll be able to watch a lot of the speaking sessions and communication that’s coming through the website. So, make sure you go check

1:32 the website. So, make sure you go check out Build, go sign up if you are wanting to learn about all the new things coming out for Microsoft for yours going to be online. Sadly, mine will not, but I’m going to see if I’m at the conference. I’m going to see if I can sneak around to a spot and have like a little video and and try and we’ll see how well it turns out. It get these conferences get extremely busy. I have all these dreams and hopes of like what I want to be doing when I get there and sometimes it just doesn’t turn out the way I think. So, the goal would be to be able to show up and be able to do my session again

2:02 and be able to do my session again like in a smaller form and then do it like online like almost like a podcast thing. we’ll see if I actually have time and and can actually do it. So, we’ll we’ll see. Yeah, I we did one of our podcasts live right at at Ignite or did we didn’t we record one or two? Yeah, the Ignite conference that we went to had a little like I guess podcast area. It was like a little area for creators. Had like a table and some chairs. chairs. and we had a camera running and we just gave our our initial impressions. What was Ignite all about? Ignite at the time was all about

2:32 Ignite at the time was all about co-pilot. It was co-pilot everything. Everything you looked at was,, I think we were joking at the time like,, when does when does paint get co-pilot and like literally like a couple weeks later, paint with like co-pilot in it. Yeah. So, that that was fun. But yeah, I will be checking up and wishing you the best of luck, my friend. The but we do have some news articles here. So, the first one is understanding your storage with one lake item size reporting. So, a little different here. When we think about when we normally think of billing, Mike, we’re usually thinking it at the

3:03 Mike, we’re usually thinking it at the capacity metrics app. There’s a storage explorer and things in Azure back and forth and it shows you the totals or if you want to dive into actually like what specific item it’s hard,, hard to do off the top and this new report fills that gap which allows you to see the latest storage report to see the broken down into visible, hidden, and soft deleted data. So just because you delete it, you still got to pay for it in the past. you’re able to see in by the large or items anything that’s

3:34 large or items anything that’s problematic. You can see anything that has a cache result or to avoid repeated scans if it’s scanning those items. And again, it’s a onelay calculate item sizes when you trigger a refresh, the runtime or any CU usage. So you can see the entire items, all the system folders, hidden system folders, too. So pretty cool. And it’s available in workspace settings one lake storage report which Mike finally this has been a little bit hidden I think for a while Tommy and I’m I’m very pleased that they’re exposing

4:04 I’m very pleased that they’re exposing this a bit more. The other thing that is also a bit hidden here too is when you look at your storage account like which tables in your lakehouse are taking up the most amount of room. I’m not sure if we’re going to get there eventually yet but it would be very helpful to see that as well. But yes totally agree with this one. This is definitely the storage report and the workspace settings. So a fabric workspace settings has been gaining a lot more menu items. Have you noticed this Tommy? Like there’s now it used to be like a handful of items and now it’s like get integration. Now

4:35 and now it’s like get integration. Now we have a one lake item. [snorts] Now there’s data factory and data engineering, data factory, data warehouse. Like all these new settings are now appearing at the workspace level which I think is a in general a nice place to centralize all the workspace settings there as well. Yeah. And I this is so important too because we’ve always struggled with this unless you use the API again I want just like any report you want a brief 3 second 30 seconds and three minute view of your data u to see you like and in

5:05 of your data u to see you like and in this case where am I spending the most money where can we go back and it wasn’t super easy to do so in the capacity metric it’s good it’s really good but obviously there’s no such thing as a perfect report as will continue to say and [laughter] well it’s like it it serves a very specific purpose to what it was there for like it gives you like a you’re looking for cus you’re looking for cus by item in a workspace like that’s where the capacity usage something went

5:36 capacity usage something went over on overage charges where do I look for that how do I find that information like all that stuff is really rich like that’s that’s the stuff that you need to go start figuring out where this stuff lives so anyways that’s really I think extremely helpful in that in that regard. regard. All right, so that is our first item. The second item we have, Mike, is it’s not really a new product feature, but I think for a lot of organizations I’ve been seeing more and more, I think just like we’ve had this whole agentic explosion, Mike, I’ve seen in the last

6:08 explosion, Mike, I’ve seen in the last six months, organizations really not having as many questions about what is fabric, but finally getting to that transition of how do we do fabric? And yes, yes, I think but and to me that makes perfect sense when you think about any new technology that adoption curve, right? No one’s going to just immediately buy something new. And what how old are we into with fabric now? About three years on the dot when build came out when we did that live stream. One of the best things I’ve ever done when we did

6:38 best things I’ve ever done when we did that live stream. That was fun. yeah, that was hilarious. but so this makes a lot of sense that you’re actually seeing the organizations one they don’t have a choice because you can’t get premium but more it’s just more comfortable. Well with that in mind we have something called fabric jumpstart and it’s a ready to run catalog of tutorials demos and accelerators to help the user experience in Microsoft fabric. So it’s almost like the five minutes to wow but in fabric. Yep. you don’t have to worry about stitching together samples or building

7:08 stitching together samples or building everything from scratch. Symbol single destination for high quality,, endto-end solutions. And it’s designed to help users learn fabric, explore fabric, and start building things in fabric quickly. This one is something that I have been wanting for a while. So last week I did a aentic thinking podcast around the tool that Alex Powers had built. yeah, yeah. Task flow assistant basically what it was.

7:38 was. So well, we interviewed him on the podcast. podcast. Yes. Yes. Yeah. Yeah. Th this feels similar in its nature, but it’s a bit more structured, right?, this is the fabric jump start program basically is here’s a bunch of patterns that we have built. Like here’s here’s general patterns that you look at. And I I saw this website come out very early and I was waiting for some public announcements around this one. It’s pretty really really cool solution. It is a GitHub repo. So there is a a fabric GitHub repo as well. So I’ll put

8:08 fabric GitHub repo as well. So I’ll put these links in in the description here as well. as well. But this feels a lot more a lot less like, hey, I need an agent to go build me something. This is a lot more of I’m going to run a Python script and it will deploy everything for me. And so one of the some of the things that came out I guess a while ago for deployment pieces was everything deployment was like a a cobble together mix of everything, right? It was some PowerShell, some Python, some C. Yeah, C. Yeah, C# scripts. Like, it doesn’t need to be

8:39 C# scripts. Like, it doesn’t need to be all that mess. So, I think this when I look at this one, this is actually a really nice blended way of being able to like let you look through specific purpose-built items. Hey, here’s how to do a real time grid intelligence, event grid, right? here’s Spark stateful streaming, right? here is monitoring workspace events, right? So these are patterns that are evolving that are now being centralized. Now I think

9:11 are now being centralized. Now I think the the this will be good if there is more items being added, right? So so if if there is if this is a a novelty where it’s just like a one and done and no one can add any templates to this., one of the thing I I’ve noticed here is it’s it feels very very Microsoft driven right now. Like all the contributors are pretty much from Microsoft from what I’m aware of. of. And And I’m not seeing there’s some releases

9:41 I’m not seeing there’s some releases here, but it’s been releasing around more around the website side of things. So, as long as the catalog of things you can build continually increases, I think we’ve got a good pattern here. I like this. I This is what I’ve been wanting for a while. like give me a place where I can land a single pattern and have it run. Well, here’s the thing. I I think about this and I what I equate this to what it’s trying to do is dashboard in a day. And I think the big adoption with PowerBI for me, Mike, in terms of

10:11 with PowerBI for me, Mike, in terms of wide organization, not just me as an individual, is if someone wanted to do a dashboard in day, tomorrow I get a call, we can set it up, have them ready to go in two hours, right? Because you don’t have to worry about the tenant. You don’t have to worry about the artifacts. It’s literally here’s my training resources repo with all the data files and a zip. Just download PowerBI and we’re on our we’re off to the races, but we don’t have anything really that e we don’t have anything close to that easy dead simple easy to

10:42 close to that easy dead simple easy to get people,, a fabric in a day. They do have those trainings, right? you need a tenant, you need the login, you need the artifacts set up,, so it’s very different because we have everything ready for us without any, you everything ready for us without any,, sign in for PowerBI, know, sign in for PowerBI, but we’ve never really had that for fabric yet. What’s your thoughts? What’s your take on the training or just getting that starting point? Yeah, fabric is interesting, Tommy, because it in PowerBI, we had a lot of like five

11:12 in PowerBI, we had a lot of like five minutes to wow or workshops that were getting started quickly and getting value out immediately. There’s a lot more bells and whistles with fabric and wiring things up and getting them together and yes, most of it works really well and you can do those fabric in a day experiences, but to your point,, it helps to have a good data set. It helps to have all those organizations. So there’s there’s definitely things fabric in a day workshops that are out there.

11:39 day workshops that are out there. I think it’s a bit overwhelming for someone to sit down and say I’m going to learn real time streaming. I’m going to learn batch processing. I’m going to learn Spark notebooks. I’m going to learn lakehouse and data warehouse like all in the same day. Yeah. Yeah. And I’m struggling with Excel. Yeah. So good luck. Right here. Here Excel user look at all these other data engineering tools you now have at your disposal. And now you have all this like I think this in general that’s just too much in one day. and some of the the fabric in a days

12:09 and some of the the fabric in a days also felt a little bit in my opinion somewhat contrived like they’re contrived examples. So I really I do think like the tools that or fabricated days that are actually toolbased building solutions like hey we’re going to turn on workspace monitoring and we’re going to watch the monitoring. We’re going to build monitoring around workspace real-time events. That that that seems like a really good use case because a lot of people can then go through the workshop, build the solution, and then have the answer at the end end of this. This feels like a shortcut to some of

12:40 This feels like a shortcut to some of these. Right. Here’s a here’s a off-the-shelf solution. You go here, open the solution up, and you just run import, and then you install this jump start project. Jumpstart. install. and you name the cost to the item and then it will build out all the items here. So, in some ways, Tommy, this does a really good job of like the documentation on this one. It’s got the full chart of all the things you’re building. It has a nice little analysis of what workloads is it

13:10 analysis of what workloads is it using, PowerBI, data engineering, data factory, data science, or real-time intelligence. It shows you what pieces it’s touching from this ecosystem. This one even has a video from YouTube on it. Like here’s how to use this one from YouTube. So I think these in places where there’s good documentation around this stuff, you can it’s almost like advertising. Hey, here’s a solution you may want to use. I’ve documented how it works. That’s how this should be done. Like community should be able to build these things and be like there’s a single place we put all of our really

13:41 single place we put all of our really rich things. and then you’re able to then understand how they work. And so I I think this is where where they’re doing a good job with this solution. solution. You hope they keep up the momentum here and the progress because I think the road they’re doing is right. I don’t think it’s the destination yet. Yes. And I think that’s my point mainly around this one is okay, I understand that Microsoft has put this in. There’s there’s people probably on the CAT team that are motivated by doing this. It helps get the the word out

14:12 this. It helps get the the word out there, but if the community doesn’t stand behind it, if there’s not an easy way for the community to start building more of these things, okay, what happens in six months from now when things move on, architecture slightly change or we have more options or there’s new things, are those going to be continually added? So, right, right, I really do want to encourage like I think this is the cat team or the Microsoft team doing this one. If you’re going to be doing fabric jump start, continually think about adding one a month or once or twice a month. Like figure out new patterns, add another

14:42 figure out new patterns, add another one. Even if it’s simple simple examples would be great. I well we we’ll get to the main topic but Mike I I I caution with that because I just don’t want random things built like please please let’s build out the same way dashboard in a day’s format of first half power query second half we’re going to learn data visualization then we’re going to do this that has the structure here because I would love with all my clients and all the people I’ve already done training with PowerBI to have that

15:13 done training with PowerBI to have that same easy. Again, and I’m not saying things are not easy now, but a dead simple way to get people started. So, if you’re going to build out a random part of the fabric jump start, I want I would rather build on top of each other like actually have courses or road maps for each rather, right? Because you want to be a data engineer. Yeah. So, I think we’re saying the same thing too. I think you can I think you could potentially leverage the technology of what Jumpstart is doing. And and the reason why I say that, Tommy, is

15:43 reason why I say that, Tommy, is Yeah. Yeah. Yeah. So jump start is is just running like fabric jump start and then it’s calling something from the git repo where it’s saying, “Hey, go use these series of scripts and install all of them.” So one of the things I think that was a little bit tricky earlier, Tommy, is in Fabric Jumpstart, they have like a how do you get started with it, right? How do you get it to run? How do you run the notebook? What does it look like? One thing that I don’t see very clearly on the website which I think would be

16:15 the website which I think would be wise for Microsoft to add you have fabric jump start you have catalog you have getting started it’d be nice to have how to contribute and go through in detail how do you how do you so so so to your point like what you’re I think I think what you’re asking for is look you have a vision for how you would want to use this thing in a way that says how do I leverage this right right to get people through a training session and so and so I don’t see fabric jump start as being a training thing. However, However, the technology of deploying the right

16:46 the technology of deploying the right things at the right time using the scripts they’re doing, that feels like something you would want to reuse for the class, right? So, okay, here we go. We’re starting day two of the class. Run the script initially. That sets up everything we need to do for what we’re going to work on today. Right. Off to the races. Exactly. Yeah. It installs the items. It links things together. It then hydrates some data. and then you can then jump into the course and say, “Okay, now we’re going to work work on real-time data. Let’s do something like that.” So, I think in that regard that would be helpful helpful and then and then so you might have to borrow it as

17:16 so you might have to borrow it as opposed to actually use it. That’s okay. I like I I like borrowing. So, that’s good. No, I I love that. hopefully that happens. So, that’s good. so, that’s everything I got on Jump Start. Mike, do you have anything else or do you want to go jump right in? Yeah, let’s I think this is actually a good transition, Tommy. Honestly thinking through like the jump start items and then this is this is probably one of the areas where people are spending a bit too much time on items for what what they’re working

17:47 on items for what what they’re working on. So fabric jump start is a great timesaver. All right. don’t go out and build architectures and piece together every single thing. Go use off-the-shelf templates., another one, Tommy, that I’ll point out here as we start getting into our main topic, which is it goes along with jump start is Alex Powers Task Flow Assistance. Right. Right. Right. That’s also another really good one. Hey, just tell it talk to an agent. Say, “Hey, I want these things. This is what I want to build.”, both Alex and I

18:17 I want to build.”, both Alex and I last week did a demo of it. I did one where I said, “Hey, build me an architecture where I’m loading SharePoint files.” Alex built one where he’s taking real-time data and trying to shove it into reporting with the solution. So, the two items here are like really useful and I think these are really good time savers when you’re trying to get stuff started from a project from scratch. So, I’ll just with that, let’s get into the main topic. Let’s talk about our main topic. Where in fabric do we think we’re spending too much time and where

18:47 we’re spending too much time and where should we start becoming more efficient? Well, let me ask you, Mike, just to introduce the topic. How did we come up with this? because I think there were a few things you and I were talking a lot offline about basically what people are doing. There’s some things on the community that we’ve been seeing as well and we are at a point right now where it’s incredibly easy to build in fabric in single item. Sure. But I think there’s two ways I’m looking at this where there’s either we’re having trouble building the

19:19 either we’re having trouble building the items if we don’t have the skill or the way we build it begins to make us waste a lot of time after the fact. And that’s where I think our offline conversations recalling. But Mike, where did this come in your head when it came to actually wanting to talk about this? I just think in general there’s when you’re working in any system or any process or things there’s things and there’s things you just don’t know like I just

19:49 things you just don’t know like I just did things a certain way in Excel for a long period of time because I didn’t know there was a better way. Yeah. Yeah. Even today I I will go into organizations and say are you using tables the tables inside Excel? Excel? Yeah. And there’s many people like table. table. No, I I don’t use those. And I’m like, did that the formula? So, this is again for me one of the aha moments. I love tables. I was in bronze. So, there’s this company called Delta Associates. In Delta Associates, this is a class you take around belt. Yeah. Black belt. They’ve

20:19 Yeah. Black belt. They’ve changed the lane of the the class a little bit, but the idea is if you’re doing competitive analysis for products, right, category management, right? I I sell a product and a lot of other companies sell a similar product. How do we know which one of our products is gaining the most amount of traction market share analysis? And so it really was helping me fundamentally think about things in a more comprehensive way. So in that class I we were doing a ton of Excel work, lots of formulas and

20:51 of Excel work, lots of formulas and everything was done a certain way and you had like this you way and you had like this format to get through the analysis know format to get through the analysis and build things. And I found very early on that by turning my data into tables, using tables made all my formulas much easier to correct, validate, create correctly across the the spans of a column. column. And some of the analysis was like thousands of products, right? If you got thousands of products you’re trying to do analysis for, you got to make sure all the formulas are correct. And so tables was a nice easy way to write

21:21 tables was a nice easy way to write formulas where I could see column A times column B do these like the math is right there in front of me. I had a much easier time of doing the category analysis by using tables. I love tables so much every single thing in the workshop was just build tables on everything. The the instructors watched my exam and then the output of my exam they’re like this is really pretty interesting. This kid’s doing tables on everything. and they started they well they used them but they didn’t use them as much as or maybe as extensively as I was doing them.

21:51 extensively as I was doing them. So then they decided hey this is probably smart for us to teach the rest of our students and I think they slightly not change the coursework but they had to added more emphasis on hey when you build these tables you these calculations right build them in a table and here’s why it’s easier and gave some more explanation there. For me, that was a feature I never really knew existed., the same thing for Power Query. It was there. I could go get an extension. I could add it in, but I didn’t really understand it was there until someone shared it with me. Hey,

22:22 until someone shared it with me. Hey, did this is available to you? And even now when I do a training, I’ll ask people, do how to use tables? Are you using Power Query? And many people are not. And so these are hidden features you don’t know. And so when I think about fabric now, it’s you just build things the way you’ve been building them, but you don’t know there’s a better way. And so this is be conversation of like here’s some other alternatives. Yeah. Yeah. To help you save time where you don’t even know you’re thinking about it. That’s where I was thinking this is going to go. Yeah. And with that in mind, I’m going to ask I want to have the conversation

22:53 to ask I want to have the conversation like where do we feel people are wasting time? Like let’s identify the problem. And I’m gonna ask both of us to have a mental exercise for the first half of this of our conversation that we’re not allowed to use AI as oh just use AI you’re not wasting time because I think there’s a process no matter what even if you are using AI you can still waste time and so I don’t want to do a

23:16 time and so I don’t want to do a catchall and I am all again I’m thinking of the peoples the peoples who listen to us who don’t have that ability for us. We’ll get into it because I think there’s a lot of AI solutions here. Yeah, there’s some there’s going to be some some I believe it’s going to change how we traditionally work. Oh, yeah. Yeah. And that’s, you and I are as nerdy as we get on the AI stuff, but I think when I look at how people are wasting time and fabric right now, I want to do one right off the bat where I’m seeing it because it’s not so

23:46 where I’m seeing it because it’s not so much the individual who wastes time. And I’m finding this more and more. I think people who are pretty familiar with fabric are it’s pretty straightforward., the medallion approach,, I’m not going to say the errors people do, but if you’re doing a medallion approach, copy jobs, pipelines, you’re pretty set up. But if you’re working in a team environment, it’s really really easy to waste time. And Mike, I don’t see people using Git nearly as much as they should or source control. Mhm.

24:16 control. Mhm. And I I think for me when I look at you cannot assume you’re doing fabric in an on an island. No matter if you felt like you did that in PowerBI, if you’re building fabric in an island, you are going to begin to waste time at some point because at some point there is going to be collaboration more than we’ve ever had to deal with. We are now building software, right? In a sense, rather than in PowerBI, I can do everything on my own. There’s always the and that goes into

24:47 There’s always the and that goes into technical debt. I will shelf that because I think we’re going to talk about that too. Sure. Sure. But there’s a lot I focus on what I see is people are building as if they’re the only one who’s doing it and that leads to so many complications. I’d agree with that. I think that’s a good I think Git integration is a huge upsell to what fabric is doing. And there’s also been some recent features around the Git integration. It’s been you’re getting more UI based items.

25:17 you’re getting more UI based items. you want to do more branching strategies and how well that works now. Yeah, it’s getting much much more the investment is being richer there from the Microsoft side which I think is just making it easier for us to then build as a team. There’s still some there’s still some gaps I think that are still in there but I I believe believe right right from what I see the direction of the platform. So one of the things I’ll just note here is there’s there’s different teams inside fabric. So good. Yeah. Yeah. There’s there’s different teams and there’s one called the platform team

25:47 and there’s one called the platform team and I believe the platform team owns things like holistic experiences across the entire platform. So like one lake catalog owned by the platform team I believe git integration owned by the platform team. And so as we look at like okay if I’m looking at data flows that’s owned by data engineering team or data integration team I think is the technical name of the team. So all these different teams are doing different things in fabric and they all have to coordinate together to get things to work and so the the platform

26:18 things to work and so the the platform team items I feel like have been be are really good at listening to like what the audience needs. Yeah. Yeah. And they’re adjusting the platform pieces in a way that makes it easier for us to use them. So overall I’ll just say I like what the platform team is doing., and I and I think the get integration is one of these value ads that yeah, yeah, as a BI developer, we’re not familiar with them, that’s not doesn’t come second nature, but the amount of times that saved me from reworking something or someone

26:49 from reworking something or someone overwriting some changes or having a legit process around saving all the work that is done and how do you revert when something goes wrong? That’s that’s what the the get supposed to do. And I I think there’s and I want to ask you because there’s this concept of time where are what does actually working faster mean and more importantly like where are teams losing the most time like data prep modeling pipelines reports governance. So let me ask you your take when you work with companies and just your own experience where are

27:20 and just your own experience where are you immediately going wow you could really do better here and work a lot faster or you’re just seeing that bottlenecks where are those occurring that that there’s better ways better solutions to work faster. faster. I think sometimes people tend to lean a little too hard on data flows gen two than going over just straight to notebooks in general., this is an interesting one that you say that that it’s not a waste time perspective, but it’s more of a

27:51 it’s more of a [snorts] [snorts] data flows gen two seems to work really well on like small to mediumsiz data. When you start getting to like larger, more comprehensive things or you’re trying to build out dimension and factual pieces, sometimes it turns into a bit of a time waste just because it takes longer for it to process and and transform these things for you as opposed to going directly through the notebook. So I think that might be an area that would be a a part of this that I’d say is potentially we’re spending too much time on that area. we should just migrate over to

28:21 area. we should just migrate over to notebooks and and be done with it like just move forward right and you have some SQL knowledge let’s just use spark SQL inside these notebooks let’s bring in a project and see if we can just execute against that I think there’s a hesitation I believe that the spark notebook is like oh this I don’t really want to learn that it’s a whole new technology stack I it’s a bit far out out but I I think the the jump between what knowledge you already have to what the notebooks can do is actually a very minimal jump., it’s not too bad.

28:52 minimal jump., it’s not too bad. So, I’m I’m picking up what you’re putting down, but I want to lean into what you’re saying here because you you’re raising an interesting point because part of wasting time here, I think, is also that skill, right? Like what is one of the first things that we do when we are in a new fabric project is the discovery and the evaluation of the current skill set of the team. the team. Sure. Yes. Right. So, and that matters so much when it comes to again how if you’re wasting time because if you only know data flows to yourself, you’re not really wasting

29:23 to yourself, you’re not really wasting time, right? Because you built a very efficient power query. You only again back to you only know what. You what. You only know what. Yeah. That was Yeah. So, you only know what and there’s a lot of those known or known unknowns. I know notebooks exist. I know pipelines exist, but I don’t know it. So that’s a known unknown to you whoever the guy was who said that he was he was in the government at one point. Rumsel Rumsel. Yeah. So the Anyways, but this idea here of you

29:53 Anyways, but this idea here of you are currently doing something that feels very optimized. Maybe you did the medallion you’re using data flows gen 2. But then the other side of the coin I think the other argument there is you were then asking people for long-term not wasting time not short term. Because if I only do data flows because that’s where I’m coming from and this has been our eternal conversation. So I feel like I would be wasting time learning Python because I’m not just going to spend a two days learning notebooks and then everything’s going to

30:24 notebooks and then everything’s going to move over, right? It would take time. This is a this is how long would you reasonably expect someone to learn Python and then apply it in a notebook, right? right? Yeah. I’d like to I’d like to hold that question until we get to the AI statements because I think that I know I was thinking that too. There there’s there’s some of this stuff where I’m like so I’m going to you said don’t answer against that. So I want to I want to hold that question till later if we can. Let’s bank that

30:54 till later if we can. Let’s bank that question question for what for a later question. All right. So I guess it would be how how how to learn Python, right? I got you. Or notebooks, I guess, would be the the experiment here. U Okay, we’ll put that for later. Okay, awesome. This is maybe one a bit more around architecture and maybe just general structure that I find in newer organizations. We do a good job, I think, sometimes of bronze, silver, gold, but sometimes you

31:25 bronze, silver, gold, but sometimes you don’t really need bronze, silver, gold for everything. And I think this gets a little bit hard because people like to show up and be like, “We’re doing medallion.” That’s just a principle or a concept to help you like get through the data processing, right? There’s methods to like bronze and the silver and the different layers. Well, what should be happening in each of the layers? layers? And some people get really hung up on like, well, I’m going to have three tables. There’s going to be one in bronze, one in silver, and then one in gold.

31:56 gold. It doesn’t have to be that rigid. I I think sometimes we spend too much time slowing down and pretty straightforward tables are straightforward. I’ve seen success with just let’s call it bronze gold, right? Get the data in, do some transformations to it and get some gold tables out the door. Like we don’t need all these different layers. Layers are I think a mental construct as to different categories of types of tables. I’ve also seen a lot of times where we bring in bronze and we’ll have,, that bronze table will

32:28 have,, that bronze table will support three or four silver tables. Yeah. Yeah. And so one table coming in can turn into four tables either in sequence or in parallel, parallel, right? right? And then that can turn that that potentially turns down to one or two tables down in the gold layer gold layer as well. So there’s a lot of patterns there and I think sometimes we spend maybe not the right amount of time or effort or thinking through that data transformation stage. So that’s that’s one a bit more on like the table side.

32:59 one a bit more on like the table side. And then last my last maybe point here is this is much more PowerBI onetoone semantic models and reports. Oh. Oh. So So yeah, yeah, I’ve I’ve seen teams just go in and build Okay, well we’ve built you the model. Okay, I’m just gonna here’s the report. You need here’s the report and let’s build you the next report. And they basically copy the same model down and then do it again and here’s a new model and a new report. And that’s just a very immature way of thinking about modeling and reporting because now you have five

33:31 and reporting because now you have five reports, five models. Okay, fine. That that works. But how do you update which measure across all five reports and models? and you start having drift in each of the models and then a year down the road these models look somewhat different and now you’re like how do we consolidate them down to the same thing., this is such a good point and Mike, I’m gonna this fires me up because in the year of 2026, if you call yourself, if you get paid for PowerBI and you do this like most of the time,

34:03 and you do this like most of the time, you don’t think when you build a semantic model of scalability, flexibility, and, the accuracy, I don’t know why you’re getting paid. And I know that’s a super hot take and mean, but I where we’re at today is you need to have that in mind in the envisionment phase of a semantic model. Yeah. Yeah. Always like we’re we are there are no excuses excuses to not build a semantic model for more than one use case because there’s always going to be more than use case or planning that way. I’m not saying you

34:34 planning that way. I’m not saying you have to build two reports together in a semantic model, but you have to envision two months from now. you have to envision the future of this model. How can I make this product this semantic model have a long life cycle and more importantly to your point there is going to be either some self-service I know I have some quota here and I know

34:55 I know I have some quota here and I know I have some actuals here there’s a lot we can do here what else are we doing I don’t want wasting time is not just about the work I’m putting in right now it’s about eliminating work that I’ve done in the past or eliminating work that we’ve already done too because to your point I may build a semantic model with the right mindset. But if I didn’t do my research on what it’s available, what we already have to say, do we need to merge things together? Can I use something existing here? This is so much of the bloat that we’re going to unfortunately

35:26 bloat that we’re going to unfortunately deal with. And this fires me up so much because I see this all the time. I go into an organization, it’s like 25 models, 25 reports. Yes. And then you look at the data and 80% of them are the same tables. and you’re like, “This could be probably six models. models. We could simp we could simplify this down. down. I could build you 50 reports and I could keep the same models, but pay me the same.” [laughter] Well, the the opposite is also true. I Well, the the opposite is also true., sometimes the opposite is also mean, sometimes the opposite is also true. Sometimes they come in, there’s like a monolithic model. Oh, yeah. And and then the users come into the

35:56 And and then the users come into the reports are like and they’re like, “Okay, we we have a mega model, but like we don’t understand how to use it.” Like it the model’s too big. There’s too many fact tables. There’s too many measures. there’s too many dimensions. And so there there’s too much information to to answer like particular reporting. So again, it’s this domainbased modeling that feels like a good blend between,, it’s not the entire companywide semantic model, but it’s like specific to what kinds of reports you’re going to be producing., and this is where I think an area of Microsoft that’s generally weak right

36:26 Microsoft that’s generally weak right now, now, there’s not the enterprise version of a model. We don’t get that capability of like like that’s something that’s not there right now. Well, I I Michael, I’ve opened models and it growled at me. Literally, I open up PowerBI desktop. I heard you’re like, “Oh, okay.” Yeah. But to your point, the like we should have perspect like I don’t want to talk about what doesn’t exist, but we could have enterprise models right now if perspectives were more integrated in everything. I think that the architecture is there to do something

36:57 architecture is there to do something like that. The the there are pieces that exist. Yeah. Yeah. To support it. I’m not. But we can’t deploy it. I really I really want my my biggest clamoring feature I think right now is I want a semantic model without any data and I need a tool that supports the autogeneration of the data and the model and the like so that to me that’s something that needs

37:27 that to me that’s something that needs to exist. That’s what data bricks and snowflake are doing with their semantic layers as well. I just saw actually speaking of which Tommy as a side note here oh man what was I I just bookmarked something from Snowflake that came out. there is a new standard semantic model layering item that just was announced announced recently that from us. We talked about that like a model MD file. We had that on the podcast. Yeah. It’s not quite it’s like a open

37:58 Yeah. It’s not quite it’s like a open standard though. It’s always said there should be something that could apply to anything. anything. We were maybe just ahead of it. I just saw that announcement. Always ahead of it. [laughter] Keep keep talking, Tommy. I’m going to go see if I can go find this this word of document and see if I can go find this this actual documentation. I’m going to be really upset or maybe Okay, I found it first. First googling was was I got it there. Okay, so the open semantic interchange OSI is what this is called. [ __ ] And so the OSI specification has been

38:30 And so the OSI specification has been officially finalized. It’s a GitHub project you can go use. And the idea of this is it’s a general I’ll get the the article from Snowflake here. I’ll put that here in the chat window as well. So here’s a here’s the article in the chat window and then I’ll also put it here the the OSI specification on GitHub as well. But to your point, Tommy, this describes a lot more around what we’re talking about here. It’s a generalized way of

39:01 about here. It’s a generalized way of why should it matter if it’s data bricks or snowflake or PowerBI semantic models. There should be a way of describing the the the business data that the logic in a general way. And so I think this OSI open semantic interchange is what we’re looking for. Okay. Okay. Yeah, I’m looking at this., let me see. Let’s see what the specification is. So, YAML. So, they did it in YAML. Okay. Okay. Which which again, YAML is like on everything anymore. That makes sense.

39:32 everything anymore. That makes sense. The whole the whole idea is that standardation is creating a uniform language and definitions., the idea is it facilitates seamless data exchange between different platforms. And it’s extensible, meaning the model will adapt as the data needs evolve. And so, if you laugh, there’s a here’s another another article, Tommy, here. This is the one that actually has the diagram., oh, this is hilarious. Look at the diagram they show you on this other article here. So, it’s it’s basically the monolithic semantic model written out as YAML. Honestly, the image

40:04 written out as YAML. Honestly, the image here is like, hey, I have semantic model A, B, and C. And they’re even going saying they’re even going farther and saying, hey, look, it’s partner A semantic model, partner B semantic model, partner C semantic model. These semantic models are expected to the the spec of this is saying all of these models are expected to live across different data platforms. And so the idea is this YAML OSI format allows you to have a generalized lens across all these platforms and then you can build out from there. This is exactly what I

40:34 out from there. This is exactly what I was talking about. I need the shopping cart of I want this measure, I want this table, I want that measure and you put it in your basket and you say build me the model and it just in every platform it understands how this stuff all relates. relates. Yeah., it’s interesting. Microsoft did not help out or isn’t part of this at all. So that’s one of my picky points here on this. this. I’m not happy about this. Whoever is listening the you guys missed an opportunity on this one. Well, I I can’t tell if this is

41:06 one. Well, I I can’t tell if this is Microsoft protecting their own like protecting their semantic model because they got they got same stuff. That’s but how so Tommy maybe we take this upon ourselves. How do you protect the one to many? I [snorts] don’t know. But like I look at this going like there’s already a lot of things in here that we could just reuse and borrow from what the semenic model is already doing. So like why why not just not just it is a symmetric model. It is a semantic model. So again this is what I was describing. It’s relationships. It’s column names. It’s descriptions. It’s measures. It’s

41:37 descriptions. It’s measures. It’s calculations. All these aggregate all that stuff. Everything here is into AI that we’re not talking about yet. yet. We’re not talking about that yet. Hang on. We’ll get there. We’re about to It’s coming close. [laughter] We’re seeing it. So, this is something here that I see becoming like I I I like where this is going. I feel this is incredibly needed for the broader system. And even let’s just let’s even ignore this. Let’s ignore different platforms. Let’s ignore data bricks and snowflake and fabric.

42:08 bricks and snowflake and fabric. Le let’s just even at a high level Tommy let’s just zoom back out and say what if I had a definition of a calculation right customer total customers customers right right oh my gosh think think of what how you like think think about what happens in even just the PowerBI ecosystem or the fabric ecosystem right in fabric we have lakehouse tables that would have information about them that we’re going to want to calculate something against it Mhm. it Mhm. What happens when I bring in a SQL user?

42:39 What happens when I bring in a SQL user? The SQL user needs to be able to understand the relationships of those tables and be able to write the right SQL to sum the correct column and filter the right data to get to the answer. Semantics. Those are all semantic definitions. Those are all business semantic definitions. H add to that now the same user a different user showing up and saying I want to build the same thing in a semantic model. The semantic model does the same thing. There are relationships. There are column definitions, there’s DAX measures, and so that defines how

43:09 DAX measures, and so that defines how they want to interact with the data. So at the end of the day, you should be able to resolve the same numbers between here’s a user coming in to the system with SQL. Here’s a user coming into the system using DAX and PowerBI reports and semantic models caching the data. Both users should be able to drive to the same answer on top of the same data. The semantics layer, the OSI that we’re talking about here,, the open semantic interchange is just a definition. I’m not even talking about other tools. How do I get to the same answer using a

43:41 How do I get to the same answer using a SQL and tables system and a semantic model for PowerBI? That should be achievable. Mike, Mike, the language is either DAX or SQL. It should be interchangeable. And they literally have a directory here for converters. Which one should be Tindle or Microsoft or RPI? Don’t care. Here’s the thing. Now all I know what I’m going to do with this because this is pretty good. I’m listening to my notion agent and they’re going to build a skill and I’m going to use this with notion to feed him my MCP. Cool. Why don’t we have anyway? Okay, I’m not going to

44:13 going to I know I know we have I know we have so much sway here, [snorts] but guys, I’m fine if they’re not part of this as long as you do your own thing. That’s not just Tim. Again, this is human readable and a AI readable and computer readable. what you’re trying to achieve as something I can send to you, Tommy, Tommy, or to someone else. This this is verbatim like this is exactly what I want. I want a converter for tim.

44:43 for tim. Yeah. Yeah. Right. So, and this this thing should be able to take the code that’s in this this general language, this OSI language. It should just say convert this general spec, this YAML file of definitions of things and turn it into a PowerBI file. Could you imagine just going to organizations like look, we’ll get started. Just send us your OSI and we’ll get started or or or give us give us your actual stuff. Like I think I think this is so okay let me let me bring this full circle here because we went on a little bit of a tangent of this because

45:13 little bit of a tangent of this because so so the the time wasted here conversation is this. This is part of the time wasted conversation. Time wasted is we’re spending so much time trying to like rip out tables and pull them together and like what model does it look like? Is it a small model? Is it a big model? Is it domain based? There’s a lot of time being executed here, right? I think if you were able to spend more time making the monolithic enterprise semantic model that lives maybe in the OSI interchange, right? right? Let’s bring it back home. Right. If you did more of that at the high level,

45:44 did more of that at the high level, building individual domain based models should be like a drop of the hat. It should be a converter type thing. 100%. 100%. Again, I’m I’m really going back at this idea of idea of I think there’s a huge advantage in organizations thinking about what how does all of our data relate together and a large non-data semantic model. What does that look like? Spend time there. And then when you give your business users the need to go do something, they can go to the catalog, the area, the space, and say, “I need to build a

46:16 the space, and say, “I need to build a semantic model quickly off of this.” So what should I do? I should be able to go to my menu, grab the things I care about, right? The calculations I want, right? right? Sum of customers, sum of revenue, sum of products, sum of sales, sum of outstanding sales, some of return products. products. These are regular calculations that you

46:34 These are regular calculations that you can just define and put somewhere. But the model should be able to say, look, I know what you’re asking for. I can get you all the tables that you need, hide the tables you don’t need, and then give you the output in a new simplified pre-built semantic model. Cuz this this is a force multiplier in my mind. I do the design once and then everyone can reuse the design. And I shouldn’t care how many smaller semantic models exist. Yeah, there are certain jabs here though at Microsoft on the actual site

47:05 though at Microsoft on the actual site for open semantic interchange. They’re like Easter egg things. For example, it says we’re 100% neutral, which basically means 98% neutral. it’s and again it’s not called model. They call it an interchange. People like just call it open semantic model, but they’re not going to do that. and you can’t bloat the word a little bit, but it but it is more of an interchange though. It’s it’s more of what model do. So there’s two things. There’s an input side and there’s an output side. The input side is any

47:35 output side. The input side is any semantics, right? It could be data bricks. It could be the data bricks unity catalog. It could be snowflakes catalog. Like the I think they’re trying to make it a bit it’s the idea is there’s enough generalized terms of semantics semantics modeling that you can generalize most of it. And occasionally there’ll be like tweaks for different sorts of systems like data bricks might have a couple extra things in their OSI that make sense to them and Snowflake might have a couple things that they do that’s spec specific to iceberg tables and how

48:07 specific to iceberg tables and how they handle it. At the end of the day, this principle exists. And so now it and the part that really lights up my brain in this one is, oh wow, they’ve got the the converters that makes sense. There’s a Salesforce converter, right? Hey, how do I take their semantics model and generalize it and pull that back out into what’s going on over there? Like Polaris, good data. Like there’s these companies that are like this is a good thing. We just need data bricks

48:37 a good thing. We just need data bricks to show up and we need PowerBI to show up up is on it. Data Bricks is one of the collaborators. Oh, it’s one of the collaborators, but they don’t have a converter yet for that one. Okay, they will. I’m sure they will. Well, I Well, I we can talk about this. I do want to give an honorable mention for something from a wasting time point of view. So, to shift or move move forward and then and then let’s move forward. Talk about what we all want to talk about., but no. Or just what Tommy and Mike want to talk about. Yeah. Exactly. thing. [snorts] But I people really always go back to this but they don’t start with this and it’s your

49:08 don’t start with this and it’s your organization. It’s your workspace organization and how you name and locate your artifacts here man because we saw this in PowerBI. We’re seeing we’ve talked about this before but it hasn’t gotten better. How do you organize a medallion? How do you organize a pipeline? How do you organize a project? There are task flows. That’s one way. Great. I know you and I are not really huge proponents of it. Like it’s nice to diagram

49:39 it. Like it’s nice to diagram things, but we don’t live and die by it. But with our display folders or or not display with normal workspace folders, with how you create workspaces and how you name things, Mike, this could be a hu this is such a time waster. Not because you’re going to spend hours on this, but it adds up because everything you do is where was that thing? I need to find this. Hey, I have this, you to find this. Hey, I have this,, initial connector. What did I do know, initial connector. What did I do after? What’s the one that actually transforms the data? Which notebook is it? And it’s one of those annoyances

50:11 it? And it’s one of those annoyances almost like when you don’t know keyboard shortcuts. It’s not going to make or break your life. But I’ll tell you what, you’re the frequency of this one is so essential because you’re doing this all day. day. Yeah, I’m gonna I definitely see your point. Oh, Oh, no. No, no. I agree with this one a lot. I was actually writing down some notes here. here. The whole idea of like data discovery, data lineage, like there’s a lot of as your team grows in size, as you build more things, this just gets to be a

50:41 more things, this just gets to be a larger pile of information and data. And yeah, I I agree with this one too, Tommy. People spend a lot of time looking for things or how to organize it and how to,, underutilized in this area is certified or promoted tagging on things. underutilized in this area is domains using a domain specific and delegating that. I also think there’s there’s something too that and maybe this is a bit more of my opinion. I’m a very big federated approach to building things. And so

51:11 approach to building things. And so sometimes when the central team takes on all the work, you get bottlenecks. And so whatever whatever data discovery or data lineage that you’re producing for your company, you need to think about how do you integrate it with experts in other teams across your organization so you can get them up to speed and running on you’re now self-sustaining. You can now build things that you need in your area. area. Well, even if you go to the basics is the nomenclature, the naming of your artifacts. And I think you actually taught me this approach or I might have

51:43 taught me this approach or I might have modified it. But people even in the naming if you have 20 items and you try to name things that can throw you off to try to find what you’re looking for. So quick tip, free tip here, the first part is everything’s underscores. The first thing is what step is it in the process. So let’s say I had something coming into a pipeline and load to load it in a lakehouse. I had two notebooks that I had to do had to do and then I’m putting it into maybe power query. I’m just throwing things

52:13 power query. I’m just throwing things out there. That’s four steps. Yep. Yep. Everything starts with 001 underscore project name or category underscore the description. So the first part of that is structured. Why the number in the beginning? It’s alphabetical in PowerBI. So 001 002 oh this is the second thing that happens in this data process. Oh, and this is part of,, GitHub integration or whatever you want to call it. One, easily searched and then two, it’s organized, self-organized, too.

52:46 organized, self-organized, too. I did this. You had something like this. Yeah. Yeah. This is this is a pattern I’ve used a lot of times for notebooks. So, a lot of times I’ll build many many notebooks that,, step one, step two, you bronze, silver, gold notebooks or I’m transforming data by customer and I need a notebook per customer. Just all kinds of different patterns. Yeah, it gets really complicated when you have many many notebooks in a single place and you’re like, well, I don’t know. There’s not a really good place to observe which notebook is calling the next notebook.

53:16 next notebook. Task flow. Yeah. Yeah. So, you can do like you’d but you’d have to build that in the task flow. flow. I want an automated one like a lineage. Yeah. And so, why don’t we have that in in in my world? I’ve been looking at this going like there’s this is another I think maybe not time waster but this is maybe where tooling lacks and we don’t have a good lineage of this right right right you have the data source of where you’re pulling things in you have the pipeline or copy job or something you’re doing to leverage the data you’re potentially transforming the data

53:46 you’re potentially transforming the data multiple times with notebooks you’re going to a semantic model and you go down to the report there’s no really good way of having column lineage from the column in the report in the visual all the way back up through the system where you can cross reference if I change this measure, if I change this column or this table, what is the downstream impact to that change? So, a lot of times I’ve talked about this for years. This is something that doesn’t exist. Stay tuned.

54:16 Stay tuned. Okay, Okay, Michael may be solving this with a workload sooner than later. I’m coding the heck out of this. So yeah, yeah, this is so but this is something data lineage, right? And throwing I need to be able to throw a lot of code at an AI and say here’s all the code for different stages of what I’m building. You figure out the lineage. You figure out where the column came from. If two columns came together to make a third column, great. Describe it for me. Document that. Put it down somewhere. and so I think a lot of that would be

54:46 and so I think a lot of that would be useful. And having like a proper ondemand when you need to generate some lineage. Hey, I’m in this report. Where did this column come from? You need to see the lineage of that column and be able to walk through where in code did that exist. And I think that’s important important a thousand%. All right. So, I just want to give us I want to give us kudos because we spent 55 minutes not talking about AI. about AI. Now, let’s talk about AI. We’ll talk. [laughter]

55:16 This is a two-hour episode. The [laughter] Gosh, if only. All right. So Mike, we’ve talked about I think a lot of things that you could do preai, but it it would be remiss of us to I think it would not be a quality episode because let’s be real here. If you don’t want to waste time in 2026, then you have to use some agentic solutions. You don’t have to be a vibe coder. You don’t have to be like, “Oh, I’m an AI bro or dude or girl.” Well, it doesn’t matter. But let’s be hon like

55:47 doesn’t matter. But let’s be hon like like let’s be honest here though, too. Yes, we love it and we and I talk about and we like being on the forefront. Yeah. But Yeah. But the it we’re at the point now or we’re getting very close to the point where it’s like if you don’t use it, it’s like, oh, I don’t like Yeah. It’s like, well, okay, you don’t like the internet. Like using Google is a tool that we all rely on rely on nowadays. It’s not cheating. It’s not the advanced thing. And I think where we’re at with AI, there are a ton of things to put it in the same

56:17 things to put it in the same vein. If you use Google to find resources, why can’t you not apply that in the what we’re in today? So, with that framework, Mike, and I know we’re already,, the time, let’s quick hit here. How do we not waste time or work faster when it comes to fabric and now having the whole spectrum available to us, the whole garden of options with AI and what’s so not?

56:47 what’s so not? How do we work faster with AI? AI is it’s not solid yet. The foundation the foundation is getting laid. I think at this moment things are starting patterns are starting to form in the AI space. space. Have we landed the proper,, do this every time and get the same results out every single time with AI? I don’t think so. And this is why we’re seeing tools like Alex P’s Task Flow Assistant show up. We’re seeing things like the the thing that

57:20 like the the thing that came from Microsoft right their version of this one the what’s it called the jump start right you’re seeing jump start show so these are things that you can throw at agents and AI and agented things and say hey I want these to to manipulate these things one thing that I was going to make a note here Tommy about our conversation which was there’s a lot of business logic that lays across the entire ecosystem of what we’re doing I Okay. Yep. Yep. Yep. Yep. Okay. So, there’s business logic. I’m going to call it business logic in the

57:50 going to call it business logic in the layers of the onion, right? That’s kind layers of the onion, right? That’s how I’m going to phrase this because of how I’m going to phrase this because when I think about layers of the onion, there’s like a core layer that’s like, okay, there’s tables in a lakehouse. That’s one layer, right? If you need access to data tables directly, that’s that’s some business logic to shape those tables there, right? right? Then there’s a semantic model. The semantic model is like an abstraction, a layer with additional business logic. How do those tables relate? what are

58:12 How do those tables relate? what are common calculations we’re going to use and then there is further business logic that can be built in thin measures that are in the reports additional and filters and context. So everywhere you look across this pattern from like getting tables in to getting reports out there’s potential for having business logic applied at any level of those steps. And so how do you do this at scale quickly and communicate what the requirements are? And I think this is where AI can really

58:43 And I think this is where AI can really assist with getting things done faster. you’re probably over building your semantic models with DAX that’s too complicated because your tables aren’t simplified enough or you need more tables that are aggregated in different ways to give you what you want on the visuals. So there’s this jockeying back and forth between if you want better performance you pre-calculate all the data by making the tables that happens in notebooks and upstream if you want like I I know I’ve

59:14 if you want like I I know I’ve seen a a model that is doing currency calculations real time in the model using measures and so the amount of people using this model They’re over buying on the size of the fabric skew because every time that someone’s clicking things, it’s recalculating currency all the time for every user. And my opinion is look, if that’s if that’s a thing,

59:47 is look, if that’s if that’s a thing, just calculate the currency conversion at the like just do it. just make the currency conversion calculation part of the a data engineering and then instead of your table having to like do all these lookups and and and recursive calculations all the time, it simplifies everything and your models get smaller. You can drop it down to a smaller skew. People are probably wasting wasting let’s call it hundreds of thousands of dollars in compute across companies of like of enterp of badly designed models, right? That’s just probably just there.

60:18 right? That’s just probably just there. So if when we go into fabric when you’re in PowerBI Pro and you’re in PowerBI premium per user, you are not the pain of having a very poor model other than speed is the only thing you’re impacted by. Yeah. Yeah. When you get to fabric, every size of model increase becomes a pain point for you. You can’t get that large. you have got you’ve got to go to bigger SKs to get more things out of it. the premium per user skew. You get a 100 gigabyte model. That’s insane.

60:49 gigabyte model. That’s insane. Yeah, Yeah, it’s it’s it’s it’s there’s a lot of there’s a lot of poor designed models you can throw at that and it still performs pretty well. Like you can get away with a lot at that level. The size of the model can get really big. You can have a lot of large tables. So from that perspective, like you get a lot of leeway at the pro and premium per user level. When you move over to fabric, you have to be much more careful around what those are going to look like. Mike, where you’re going with the organization here and you’re touching something for me when it comes to this

61:20 something for me when it comes to this is not what I’m about to say is not technical and it doesn’t have to be because people always equate what I’m about to say with doing it wrong, not with wasting time. And that’s your documentation, your discovery, what you’re trying to actually achieve. And because I think with the abilities, the easy agents that we have to be able to transcribe, help you record, outline everything that was talked about when you’re trying to figure out what you’re trying to do in the first dang place.

61:50 the first dang place. Yep. Yep. Right. Like whatever your harness is or you’re writing your notes yourself, what you write down and what you remember from a call is the biggest time waster of them all. Because if you get it wrong or what usually happens is you don’t get it wrong, it’s just not super clear. You’re like, “Oh, I think he said something about a metric, but I don’t remember what it is. I’ll get there later.” You’re like, “Oh, I think they said that table works, but I don’t remember.” Okay, we’ll just work on it. And then there’s revisions, and then

62:20 And then there’s revisions, and then there’s something that’s updated. And then there’s something that wrong. when you’re not clear in the beginning. Mike, what I found today using notion, using the harnesses that I have for every every meeting I have with a client and talking about going through the what the meeting is and having custom agents in notion to go, hey, hey, they know fabric, they know me basically,, they know everything about PowerBI and they know everything that I do. And I go, “Hey, based on this new project,,,

62:50 based on this new project,,, we just had a call and they said that they now this want this other data source. Does that fall into our our project? Does that fall into this scope?” No. Okay. What can we do? D. It’s like, “All right, let’s go to the model definition that I have a notion which I will be implementing OSI in.”, here I think it’ll be useful. You better believe it. You better believe it. I’m gonna I’m going to say today, “Hey, congratulations agent. You have a new skill.”, [laughter] love it. But like but applying all these things, this is the biggest time waster of them all. So Mike, what’s your take?

63:22 of them all. So Mike, what’s your take? Do you agree? Am I being too far off here? here? I I don’t think so. In this way, I think I would also agree with you a lot is while documentation is incredibly useful. So let me give you a story here. I had a client who was doing a lot of like meetings with business users and and capturing information and details. And I said, “Look, I see a world where you’re going to need all that video and transcribing of those conversations. You’re just going to feed it to an AI and just do things with it.” And I think we’re starting to get to your point, Tommy, you just described exactly that.

63:53 Tommy, you just described exactly that. I’m having conversations with clients. I’m talking to them about what they need. You, Tommy, are prompting them in a way that says, “Okay, tell me more about this. Where is the pain? Where is the problem? What do you need to see? What keeps you up at night? Right. We’re having those real person conversations so I can focus on the person. Correct. And you’re distilling that knowledge into text that the agent can go figure out. And so I don’t want to point out something here that I just ran across my radar here. My recommendation to the customer was like, look, start now. AI is going to

64:24 like, look, start now. AI is going to get to a place where it’s going to be able to absorb a lot of your conversational data into a place where you could use it. I said, just throw everything in SharePoint. just keep it there. And so, boom, here we are., this was announced. Let’s see if I have the date. It’s dumb. Microsoft puts dates on their articles now, and it’s like at the very bottom. I can’t even find them anymore on the date of the article, like when it was published. So silly. Okay. Anyways,, April 8th. So, in April 8th, Microsoft published this new article called Getting Started with AI in SharePoint. It’s in preview.

64:55 with AI in SharePoint. It’s in preview. And so the whole idea of this one is you can now throw AI directly inside SharePoint. And the one the reason why I think this is interesting, it’s not just like an AI agent that’s inside SharePoint. This AI will write PowerShell scripts, do search things, it will it’ll build are two years off, right? Guys, you should have been This is what I wanted to see a while ago. So this is a this is an AI that builds tooling inside SharePoint that it can use to go scan and search and go through

65:26 use to go scan and search and go through all these things like yeah this is what it should be doing like so it’s it’s nice to see that Microsoft is coming around with this one where you can now leverage this AI directly inside SharePoint and it now creates views and you can now automate workflows and you have now structured documentation generation That’s what this should be doing. Like that’s where you should be applying the AI. The AI shouldn’t be the thing that is giving you the answer. It’s the item that’s and I’m going to go back to this. We’ve talked about this before, Tommy.

65:57 We’ve talked about this before, Tommy. It’s the creator agent. The AI needs to be used to help you create tools and systems and processes that are then running and usable over and over again because we know those are deterministic and they build the same output every single time. I love what you said though too about from the meeting too because I didn’t I wasn’t even putting it that way. I’m not wasting time on meetings which nobody wants to do and is the worst time waster like hey can you explain every seven columns and what they do so I can write it down during the meeting okay and

66:27 it down during the meeting okay and you’re right what we can do now is I can focus on the person their pain points y which will have context somewhere else so everything is more productive and I I agree something like confluence or shareepoint or notion whichever choice I think that’s such an essential part of what we’re doing here but Hey, I think we need a part two. I I don’t know about you, but this was great. I think there’s a ton more things we could dive into. especially with all the tooling out there from the creator agent side, especially in fabric here, but I hope a

66:57 especially in fabric here, but I hope a lot of people see today. It’s not just your technical skills that are wasting time in fabric. We talked about some but I think this is very telling for you and I for a lot of people that 60% 70% of what we talked about was not your skill in fabric. It was not your hard code skill. And I think this is very telling on yeah go learn Python, go learn these things or whatever you do but there are a lot of things you can do to waste time and there’s a lot of ways to optimize

67:28 and there’s a lot of ways to optimize your workflow and the value that you provide. So that’s my closing thought. Mike, I like it. I think this is good. I think my closing thought here, Tommy, is there’s a lot of things that you don’t know you’re wasting your time on., on., the the question I have in my mind is let’s let’s say I’m me listening to this podcast or someone else in the audience here, here, I only know what I know. I only know what I’ve built. I only know I only know data flows. I only know PowerBI. I know what I know. And I would just say be

67:58 what I know. And I would just say be have an open mind to continue to learn and and pull into the community and and research things and continually challenge yourself with better ways. There’s a very interesting thought conversation. I don’t know how you want to say it, but like there’s this idea of sometimes you need to slow down to go faster, right? And often when I’m learning something, I have to slow down the pace of what I’m building to just take a moment to digest and figure it out. And then once I’ve

68:28 and figure it out. And then once I’ve digested and understood, I can then ramp back up and go quicker again. So I think I think sometimes we have to be careful with that and be able to be mindful of sometimes a slow down to go faster is is right. we can’t always go in be moving at 1, 000 miles an hour. All right, that being said, thank you very much for listening to the podcast. If you like these links in the description, they’ll be down there in the description for what we’ve been talking about., check us out there. And then Tommy, where else can what else should we find the podcast? Where else would you find us?

68:58 else would you find us? Oh, you can do other things. You can find us on Apple, Spotify, or wherever your podcast. Make sure to scribe rating. It helps us out a ton. Do you have a question, idea, or topic that you want us to talk about? Do you agree with what we said today? Head over to PowerBI tipsodcast. Leave your name and a great question. And finally, join us live every Tuesday and Thursday, a. m. Central, and join the conversation on all Power. TIP social media channels. Thank you all so much, and we’ll see you next time.

69:31 Explicit measures, pump it up, be [music] [music] lighting up the sky. Dance to the day in the mix and I get your fix. [music] Explicit measures, drop the beat now. Has kings feel the crowd. Explicit [music] measures. Explicit measures. Drop it loud.

Thank You

Want to catch us live? Join every Tuesday and Thursday at 7:30 AM Central on YouTube and LinkedIn.

Got a question? Head to powerbi.tips/empodcast and submit your topic ideas.

Listen on Spotify, Apple Podcasts, or wherever you get your podcasts.