Kicking Off Fabric the Right Way - Ep. 508

Mike and Tommy use this episode to pressure-test a familiar problem: organizations want to move fast with Microsoft Fabric, but they often skip the setup work that keeps the platform useful, governable, and actually adopted. The conversation frames Fabric kickoff work as a sequence of practical decisions around trust, readiness, scope, and operating habits—not a random pile of features to turn on.

News & Announcements

• Update to required permissions for Semantic Models in Fabric Data Agents — Microsoft is simplifying how permissions work when people interact with semantic models through Fabric data agents. Starting April 6, creators and consumers only need Read access on the semantic model to add it to a data agent or ask questions through that experience, instead of needing broader workspace access and Build permissions. The change lowers friction for adoption while keeping Write permissions reserved for model changes and features like Prep for AI.

Main Discussion

The core theme of this episode is that getting started with Fabric well is less about chasing the newest capability and more about creating the right operating model from day one. Mike and Tommy push on the idea that teams need to be intentional about how they introduce Fabric, who owns the experience, how much trust they place in the platform, and what the first implementation should prove.

• A strong Fabric kickoff starts with narrowing the first win. Rather than trying to stand up every workload at once, the better move is to define a practical starting point that proves the platform can solve a real business problem.

• The conversation reinforces that architecture decisions made in the first few weeks tend to echo for a long time. Naming, workspace structure, semantic model ownership, and deployment habits all become much harder to unwind after teams start publishing content at scale.

• Mike leans into the importance of trusting the platform enough to use it, while still keeping your eyes open. That means avoiding analysis paralysis, but also not treating Fabric like magic that will automatically clean up weak data practices or unclear requirements.

• Tommy brings the organizational angle into the discussion by grounding the rollout in human behavior. A Fabric implementation only works when people understand the boundaries, know where to build, and have a path for support when something breaks.

• The episode treats governance as a practical enabler rather than a bureaucratic tax. Small decisions—like clarifying permissions, deployment flow, support expectations, and who can create what—help teams move faster because they remove ambiguity.

• Another thread in the discussion is sequencing. Fabric can do a lot, but trying to introduce ingestion, transformation, semantic modeling, reporting, real-time patterns, and AI workloads all at once usually creates confusion instead of momentum.

• The hosts also frame kickoff work as expectation management. Success is not “we turned Fabric on,” but “we created a reliable way for people to build, publish, and trust the data products that come out of the platform.”

• The practical takeaway is to start with a scoped use case, create a few clear guardrails, and let early delivery teach the team what needs to be standardized next. That is a much healthier pattern than overdesigning the future or improvising everything in production.

Looking Forward

If you’re about to start a Fabric rollout, the best next step is to define one meaningful business use case, the team that owns it, and the minimum governance decisions you need before delivery begins—then use that first implementation to shape the rest of the platform playbook.

Episode Transcript

0:02 Tommy and Mike lighting up the sky. Dance to the day [music] to laugh in the mix. Fabric and A. I get your fix. Explicit measures. Drop the [music] beat now. Pumpkins feel the crowd. Explicit measures. Good morning and welcome back everyone to the Explicit Measures podcast. I’m going to start out with an apology just really quick here. Still dealing with some cold. I’m going to have [clears throat] to adjust my my voice

0:32 [clears throat] to adjust my my voice here a lot. So, I apologize. Just not feeling the greatest. Got some head stuff going on. So, anyways, apologize for the voice and the sound. It’s going to sound a little bit different than normal, but other than that, we are here. We made it. We’re already into March. Tommy, listen. And everyone’s going to be sick next week because we spring forward on Sunday. So, don’t forget, even though your phones remember forest now, and that always throws everybody off. You’re just tired. Yeah. Why do we need to do this? Stop moving the times around. It doesn’t I don’t

1:03 the times around. It doesn’t I don’t even see the light of day anyways. It doesn’t really matter what time it is. It’s just the same no matter what I do. I’m stuck in my basement. It’s just I will tell you this. It is always more depressing when you spend the full day in the basement and sometimes you’re just in the mode. You’re in the groove and then you come back up and it’s dark. Like, oh, Like, oh, you feel like you skipped the day. I missed the whole day. Yeah. Yeah. I went to work and it was dark and I came back and it was dark. Yes. Yeah. That is that is the winters that are in Wisconsin. So, I get it.

1:33 are in Wisconsin. So, I get it. Chicago, too. But we’re getting there. Awesome. Let’s let’s jump into our main topic today. So, our topic today will be let’s talk about kicking fabric off the right way, right? What do you do when you initially want to start rolling out fabric? What does that look like?, how would what are some practices that Tommy and I have done or we have used to start using fabric from the very beginning, right? How how do you get going with what fabric is going to be doing? How do you open it up? What people do you give access to? Is there

2:04 people do you give access to? Is there any training that we need to like preemptively go through before we get started? Just things like that. So, let’s that’ll be our question today. We’ll unpack a little bit getting started with fabric the right way or at least Tommy and Mike’s opinion. There’s probably like no one right way., but we can hopefully give you some pitfalls to avoid when you roll out fabric for the first time. That being said, let’s jump over to some news. Tommy, you got some news for us? All right, so we got a article. It’s not terribly impactful in terms of like the

2:35 terribly impactful in terms of like the amount of features, but I think this is important for those who are using data agents. We have now changed or Microsoft has now changed me and you, we we went through Microsoft hacked it and changed permissions. [clears throat] So, Microsoft has now changed that it’s required for anyone who’s using data agents and semantic models that the person reading it must have or using a data agent connected to semantic model must have read access to the semantic model. Surprisingly, this was not a

3:06 model. Surprisingly, this was not a feature. So in a nutshell, if you’re using a data agent or talking to a data agent and it’s relying on a semantic model that you do not have read access to, it will not give you any of the underlying data or talk really to you. It’ll just tell you, hey, you don’t have access to this. Previously, as long as I had ability to access the data agent, I could access the data behind it. So the experience is pretty simple. Yep. And this is starting as of today or yesterday March 4th which is yesterday

3:37 yesterday March 4th which is yesterday and simply that’s the big change here I believe is there a [clears throat] is there an actual documentation like a Microsoft docs that goes along with this and there’s a couple things here talking about like prep for AI I see some permissions to interact with semantic models I don’t know there’s some additional pieces here Tommy but all this to say this makes sense to me, this makes sense with like the Microsoft security model, right? The data agent is acting

4:08 right? The data agent is acting as like a service principle on top of the data to some degree. And all this is doing is adding another layer of checks to when the user. So this is something that I think needed to get built and definitely needed to be there before data agents go like GA which is if I show up to a data agent the data agent knows who I am because of credentials and then therefore it should know okay we’ll pass that context of who the user is down to the model or models right

4:38 is down to the model or models right because again think of this too Tommy we’re not just doing a semantic model here this could be multiple semantic models now this is interesting Because I’d also like to understand like okay this is talking about semantic model permissions but what about like you can add anology you can add a lakehouse you can add tables like what are all the other security things that are out there. So to me this is like a good starting point but I also think this is a very small portion of a bigger story which is data agents can touch a lot of data. How do

5:08 agents can touch a lot of data. How do you manage the permissions across all those data assets and give the user permissions down to those underlining security items? Does that make sense what I’m describing there, Tommy? No, 100%. And it’s funny because so I have a few projects right now, Mike, that are data agent related. And the primary recommendation I’m having on the early ones is we’re going to use a semantic model to have your data agent basically recognized. It is so efficient to go through and especially the types

5:38 to go through and especially the types of queries and questions that people have or what they want to get out from the data agent. Sure, Sure, I have that structure with a semantic model. Not that I don’t care about the lakehouse or database, but it is so much more data that’s necessary usually what a semantic model has. But no, I I completely I I need to see a table. I need to see a table permissions based on the data source, what a data agent can do. I’m sure it’s there in the Microsoft docs., but again, if you’re using a data agent, you need to you need

6:08 using a data agent, you need to you need to know that. I agree. Anyways, just be aware things are changing. [clears throat], if you’re just starting to figure out data agents anyways,, this is just something to add to your knowledge base., the the link for the article will be here. I’ll put it in the chat window right now, just in case you want to read up on it a bit more. This is the full article link here as well, just in case you want to check on it. Okay, good topic there., Tommy, back over to you., I think you’ve got a beat from the street here. So, I’m going to give an argument here that I know that you’re going to disagree with, so these are always fun,

6:39 disagree with, so these are always fun, but I am going to make the case, the title of my argument is I’m going to make the case for Power Apps. The underlying context here is I have the same ones with the data agents a few projects that are actually dealing with rightback actually multiple ones who been requesting we want to do right back in some way in PowerBI and obviously again let’s think about the three ways in a sense I can talk to my data I can build my own application

7:09 build my own application trans analytical or power apps to me are the three primary platforms or foundations for accomplishing that said goal. goal. Yes, I would agree with that statement. Custom application, power app or trans analytical and I I would want to just very briefly hang on your thought, Tommy, just so I understand where you’re thinking with this when you’re serving data back to people. Are you expecting to go through the semantic model or something else? That’s one very Yeah. So

7:41 That’s one very Yeah. So it used to be only we could serve data back easily through PowerBI using only the semantic model but I think this is changing changing with newer tech that we have in the space. Yeah. So what’s your when you describe those three areas what is your expected way to serve the data back? Yeah. So for me especially if it’s a fabric integration which we very recommend if they want to do any of this now is I’m going to push it back to the lakehouse. I’m going to update, modify, and push a job in a lakehouse.

8:11 and push a job in a lakehouse. Preferably, preferably, preferably it’s a SQL database. If I know an organization is wanting to do a lot of write a lot of write back, it’s easier with SQL database than it is a lakehouse. It’s much easier to write really. And I wouldn’t say so much that the only argument here is Power Apps, but I I’ve broken into three buckets here in terms of which use case. Small one-time changes where I only need in PowerBI to have a single update. Trans analytical serves a purpose. If I

8:42 Trans analytical serves a purpose. If I do want to do a little more bulk or this is more going to be integrated with the PowerBI report, then I’m looking at Power Apps here. If it’s something even more then it’s not even a powerbi report it’s really a custom application and I think I know that you have felt that with AI and just with trans analytical do that where is let me ask actually ask you where you where does power apps have that better than others what does it do better than the other two to you

9:18 well I’m [clears throat] gonna I’m gonna poke And I think the part that I don’t love about Power Apps is it’s I haven’t spent enough time learning it. Mhm. Mhm. Sadly, right? but I think that’s primarily due to it comes with an additional licensing cost. So if you have like Dynamics 365 or you have something like that, you get some power apps by default, right? Yeah. Yeah. So from that regard, you could just use it. But, I would argue most organizations don’t have just Dynamics

9:49 organizations don’t have just Dynamics laying around. They don’t have extra licenses laying around. So,, I’m paying $14 per user for PowerBI or,, some fabric capacity. capacity. When I get into Power Apps, I now need to have another license and then I have to think about the design of the system, right? So, Power Apps doesn’t freely connect to lakehouse data. Power Apps can connect to semantic models. Power Apps can and also connect SQL databases, but most of everything I see in Power Apps now is all premium

10:19 in Power Apps now is all premium connectors. There’s very little I can build on top of Power Apps that isn’t a premium connector. It feels like you only get like SharePoint and a handful of files you can use as sources for the Power App app. And now I’m going to go back to like does it does it really make sense for me to do that? And so [snorts] Transit, Transit, yes, there’s cost associated with it, but there’s no additional licensing, right? I can just build it out of the box, no problem., and then Tommy, you said custom applications, right?

10:49 said custom applications, right? Previously, it would have been really hard for people to build custom applications, but I think as of like no def December this last year, 2025, building applications is like nothing. Like it’s actually easier to build an entire application to build a Power App now because the AI is so good at building what you want. It’s you you do

11:09 building what you want. It’s you you do need to learn how to leverage the AI to build what you need, but you can build a full-on application. Again, no additional licensing. I can now use semantic models or SQL databases as part of the custom application. So, I me personally, I think the Power Apps is falling third in a race here. It’s falling behind compared to the other solutions. If you’re building custom apps, you can build them as a static website. So you can literally pay $15 a month to run it in Azure.

11:42 $15 a month to run it in Azure. That’s just that alone. That’s like half the price of what a Power Apps license is. is. Sure. Sure. So So when and that’s for all users, right? You build the app and you’re done. [clears throat] So I’m going to go back on like I like the idea of Power Apps. The fact that I can’t vibe code or vibe build it and power pages is also very interesting, but it’s very too it’s way too pricey. Oh, that is way tooome. It’s it so I just feel like it’s a it was a solution that was too early for

12:12 was a solution that was too early for its time. It was a good solution. It made a lot of sense, but it was too early. And now once we had this whole idea of being able to build apps with agents, it basically made all of my opinion, it made all of Power Apps obsolete. And I have cheaper, faster, more customizable solutions that I can just build off the out of the box with whatever I want. And I I it now the limitation is my imagination. It’s not how how do I have to understand the framework of Power Apps to get it to do what I want to do.

12:42 what I want to do. I’m going to give you a counterpoint. I’m going to refute the other point. First off, yes, you can vibe co Power Apps because you can actually create in a Power App and part make it part of a repo, which the whole thing is a bunch of YAML really on the back end. It’s not like readily you see it in the UI, but I can say, “Hey, connect this Power App to a repo.” Next thing, I have the full code just like they did with Timle. They have the same thing with Power Apps. So theoretically you can do vibe

13:12 Apps. So theoretically you can do vibe coding from scratch say hey create an app in this repo that’s fair and yeah so that’s more now again I will agree with you to a point where it’s not universal like create javascript like you have to really give it a skill so it’s got to follow some you it’s got to follow some nonuniversal rules these are know nonuniversal rules these are specific to power apps yes yes the other side of the coin too you mentioned this $15 for a website sure but if a company’s dealing with their own data They want different views and more dynamic. There’s a difference,

13:43 more dynamic. There’s a difference, Mike, that I will argue between a app and a business app. And this is really from my experience building business applications. It’s not so much the customization that people are looking for. It is the dynamic views based on the user who goes in. Not to say you cannot do this in an application, but Microsoft has made this dead easy to do in a Power App. Hey, if a user from sales goes into the app, show them these four tabs and make that dynamic. And

14:15 four tabs and make that dynamic. And also make sure that they know their notification. Here are the things that are grayed out based on the user who signed in. Not again, I know that can happen in an app. I can run open claw and do that overnight in to an extent, but Microsoft was made for this. Office 365 was made for that types of rules. And that’s really where Power Apps truly shines. shines. Can can I power app edit data in the lakehouse? Oh yeah. Oh yeah. actually I have not tried the lakehouse.

14:45 actually I have not tried the lakehouse. We’ve I’ve always focused on databases. Okay. So Okay. So you can trigger a job. So yeah this is this is one of my friction points here around power apps. Right. Perplexity. Yeah. Continue. But yeah I believe you can connect to lakehouse storage tables. I think that works., works., but I’m gonna go back here to argue it’s the licensing thing for me, Tommy. That’s what steers me away from the solution. I don’t I don’t like the fact that it’s so clunky., I I

15:18 fact that it’s so clunky., I I do agree with the repo stance because everything we do in PowerBI is repo based. I get it, understand, but I don’t want to have to go spend more money on another license to get something else to run. And that’s to me that’s an immediate turnoff for getting into these rightback solutions. I’d rather rather I’d rather build a solution that’s a bit more more custom owned by me, either web app or trans trans analytical to actually do it, right? So I I would actually argue those are be better value

15:49 actually argue those are be better value ads for me than I’m not paying for the extra licensing per person, per user. It’s it’s like a per user thing. So the difference here is like I can build what I want and then I have no restrictions. Like Like I want to connect a SQL, no problem. I can do that. I can have data being entered in there. I can edit it back, write it back., there’s no for me the the rightback solution is best when you have a custom app. You can build whatever you need there. And the barrier to do it with agents and helping you out with aic things is so low now. Anything

16:19 with aic things is so low now. Anything that causes a little bit of friction for me is like I don’t really want to do it. Like I’d rather go to like don’t want to do it, but your client may. may. But who’s gonna like who’s gonna have to maintain it? Are they willing to pay for the extra licensing? Right. When you lay out I don’t have to tell you that when you lay out the options it’s not as I don’t think it’s that hard of a jump for them. for them. I don’t want to use monday. com but some of my clients do. So guess who has a subscription too right? Like in ter or guess who signed up for monday. com

16:50 guess who signed up for monday. com because you want me to use monday. com you pay for my subscription client. That’s what that’s how this game I’m I didn’t say my project app. I’m using their project app. I don’t pay for it. But it. But so, so sometimes you just play, you’re playing in their playgrounds, so to speak. You can recommend things, but all this to be said, like I do agree with you. We’re getting to the point. I am very frustrated that Power Apps does not have a native way to write back to a Delta table. To me, that’s you’re missing to you’re losing customers that

17:20 missing to you’re losing customers that way right now. I think actively losing customers. Maybe, maybe not. Like, I don’t know if that’s really the best way to think about it. I again I’d also argue your top position customers this the way you service data is the most important thing here right so in in any regard no matter what you’re going to do you’re going to have to figure out how to do direct query on top of something something so it’s direct query to some data data and so if you’re going to need to have editable or writeback data it’s if you want the feedback right away inside the application inside the system

17:50 inside the application inside the system you have to wait for the semantic model to refresh or something along those lines or you need a mix of tables in the semantic model that are direct query and some that are import that are slower to change. change. Yeah, Yeah, this to me is just friction and there’s not an easy way to push data into the model anymore because they got rid of it. That was there there was a push data set thing. You could literally push data into a data set and it would just show up immediately on dashboards and things and it would just show up. We don’t have that anymore. That’s that’s been deprecated. No one’s using it. So,

18:21 been deprecated. No one’s using it. So,, where I look at this system, I say anytime you’re talking about rightback, I’m probably immediately thinking SQL Server somewhere, and then I’m looking at what’s my licensing options, what’s the easiest way to get things done. And even if you did have Power Apps, Tommy, the ability for you to agentically build things anymore is so fast nowadays. I know. I’m just saying. I’m just going to push back new. Yeah. I’m just saying the the amount of extra time I would need to

18:51 the amount of extra time I would need to spend to get the Power Apps to work to me. I’ve actually like I’ve moved on. It’s it’s for me it’s not for me anymore. I won’t I won’t take a project anymore. If someone comes to me and says, “Hey, we have Power Apps. We need to build something with Rightback.” I’m not going to take that project. I’m going to going to forward that email to me then. So, [laughter] you won’t get any cuz they’re not going to need them. [clears throat], if no one’s if anyone’s asking me about it, like we’re not going to use it. It’s just not a good product anymore. That Microsoft missed the boat here. I I cannot disagree with you more and I think you again for me and you. I don’t

19:23 think you again for me and you. I don’t need to build a power app for myself, but I’m also a company of two, right? And the second is my wife. So, it’s and it’s not really u one. Let’s be real. It’s one. It’s one. I know. I She doesn’t She doesn’t build power. She did dashboard in a day. Oh, add moment here. But I built her a model for her her new her new work and it’s a nice consulting place that does also like designs and stuff perfect for

19:54 also like designs and stuff perfect for her and they were billable hours all that yada yada and she’s like this is amazing and I had to do some more training with PowerBI but she forgot to save her file and it was one of those moments because I had to make changes to it on her computer but she didn’t save it and she didn’t know where she put it and it was just one of those moments of like I’ve dealt with clients like this and I cannot now say what I would say to a client or a colleague to my wife [laughter] and you just have to reward you have to find a different way to say the words. You need to take a quick walk and then

20:24 You need to take a quick walk and then come back and say and then say something calm. calm. Yeah. [clears throat] All right. I think this was good though. I I I enjoy the conversation. I think we’re going to agree to disagree on Power Apps, but we we probably will. And, if I don’t get any requirements around these, I it’s fine by me. I will not build Power Apps anymore. I will tell people to build other stuff that are way more capable. One thing I will also point out too is GitHub Copilot actually has its own AI generation app building solution. It’s called GitHub Spark. GitHub Copilot Spark.

20:55 GitHub Spark. GitHub Copilot Spark. Yeah. Yeah. Which actually lets you, I would argue, do the same thing you would do with Power Apps and YAML. You show up, you tell the app what you want, you give it some product requirements, and boom, it writes a whole bunch of code, builds a whole bunch of pages, and then you have an app that does things. I’ve been actually building. So this is this is the reason why I feel this way strongly now is because I’ve actually been building real apps on top of fabric SQL. SQL. Yeah, Yeah, this is amazing to me. You [snorts] know, previously you’d have to go build

21:25 know, previously you’d have to go build something in Azure. The data would be locked away somewhere. You’d use some database that you’d have to go talk to or pay for. It’s included in my subscription now. I love this. I love the fact that I can go out and just talk to something and get something built that is functional that then has real hooks into real data that I can go back and see and now I can make the decision on does this data only live in the SQL server or do I take some of the data out and put it back into a semantic model and then serve reports back into my

21:55 and then serve reports back into my application. So now I have this really interesting,, pattern of I’m now building APIs that are doing custom SQL calls that are doing filtered conditions, all this other stuff that you normally would have to build and I don’t have to write any code. I just talk to the agent to do this stuff like that. Regardless of how you think about Power Apps or anything like that,

22:15 about Power Apps or anything like that, whatever solution you pick, I want to be able to communicate my requirements at a high level and get whatever built done. Whether you pick Power Apps, whether you pick a standalone app, does not matter. You’ve got to have like to me the the things I’m going to die on are the hills I will die on are you need to have the data stored in a place that is easy and almost like part of your ecosystem for data. data. Yeah. Yeah. To me it’s like a no-brainer to say turn on fabric land your operational and you

22:47 on fabric land your operational and you on fabric land your operational and reporting data all in one platform. know reporting data all in one platform. You’ve got enough tools at your disposal it will work. You’ve got GraphQL inside the SQL database. You’ve got Cosmos DB if you want record type stuff. You’ve got SQL databases for traditional transactional stuff, table based information. Yeah, Yeah, it just all lives there. Like everything you need is there to help you build stuff. So it’s all secured. It’s easy to access and I just turn it on and it runs. That’s really nice. So I like the ease of that. Combine that with okay, let’s talk about what we need to build

23:17 let’s talk about what we need to build on the front end side of things. You now have a lot of flexibility there and I can just talk requirements to it. build a product requirements document and then get what I want relatively fast. I’m building three or four apps a day now with just building out requirements like full apps like that do stuff like this is insane what it can do. So we’re just at the beginning of this. It’s only going to get better and I think everything else is like those are the two things that I’m going to continue to use if Power App steps up to the plate

23:49 use if Power App steps up to the plate and adjusts itself to be more like you and adjusts itself to be more like GitHub Copilot Spark or it’s it’s know GitHub Copilot Spark or it’s it’s more easier to use if the licensing process gets less. Like there’s a lot of things that could be happening here that would make it more attractive for me to use it. But right now now I’m going to die on the hill of I want to be able to talk in English to my model to get what I want. Fair enough. Whatever language you talk. Yeah, fair enough, my friend. All right, I think we’ve hit this one pretty well. We got a great topic today. We do. We have our main topic. So, our

24:19 We do. We have our main topic. So, our main topic today is kicking off fabric the right way. Tommy, you want to give us the the read through on this one? Yeah. So, this again, mailbag March Mailbag Mania is what we’re calling this month because we’ve just got so many good mailbags here. Agreed. Agreed. Let’s dive right in. And again, no one’s putting their names. I know we already like picked the topics and I already know if people have names or not, but put your names. I’m going to keep saying that. Starting out with PowerBI, our company’s reporting team has been using import mode exclusively and we have not

24:49 import mode exclusively and we have not been reusing models very commonly. There’s a lot of redundant data storage and I am sure there are many conflicting DAX definitions and model relationships. I’ve started learning about fabric and I’m realizing we will need to rebuild some semantic models from the ground up. thing is exactly right. I’m not sure our team will have the patience to do things right. I suspect they will dive in and start creating more redundancies and our work will continue to be fragmented. How can I

25:19 continue to be fragmented. How can I kick things off on the right foot and influence the team culture to work together with a single source of truth rather than individual sources? A lot to unpack here. There is a lot to unpack here. I do want to hang on the governance piece. So, I think really we’re going to probably like towards the end here we’re talking about about team culture. I really want to I’m actually going to go here and bold that in my notes. That’s really what we’re talking about here. There’s a lot of team culture pieces here that will probably

25:49 that will probably this could be highly debated on how you want to build or how you want to grow team culture in your company. I think this is a you have flexibility here, but at the end of the day, your culture of your team will really dictate how you get stuff done in the in the business. All right. So, that being said, where do you want to take this, Tommy? What do you want to start with?, honestly, just hearing the first part of this paragraph got me sick. So, [laughter] there’s there’s a lot here that I think that’s really important for us to

26:19 that’s really important for us to understand before we even talk about fabric. fabric. The first part Yeah. Go ahead. Yeah, I just want to before I I agree with you, Tommy, on your first statement there. I also would agree. I think a lot of companies coming [snorts] from PowerBI PowerBI that start with a bottomup approach, right? This is a business-led initiative. Hey, we need PowerBI. Great. We all have Microsoft 365 M365 licenses or whatever the the whatever

26:49 or whatever the the whatever the license is that gives you free PowerBI Pro. Great. Everyone comes to start building their own stuff. People find value in different ways, but everything to your point, Tommy, disjointed, repeated, not consistent, all this stuff. I think there’s a lot of organizations that are operating that way. It’s not wrong. It’s just how you started. So, I just want to just make that quick note there. there. And I I think one of the things you and I have been doing more on this mailbag is we try to make assumptions about who’s writing in the situation they’re in. It’s a fun little exercise here. And

27:19 in. It’s a fun little exercise here. And to your point, this is this is a reporting team. This is not just random people using PowerBI at the organization and they’re like, “Oh, I didn’t know you were working on a sales report.” This is a centralized team or a version of that that that’s how I’m hearing this. But to the point, they’re just probably working what’s the loudest. There is no collaboration or a ticket system that I think is actually working or more importantly, there’s no communication around the team. This to me screams of the who he he who yells the

27:51 the who he he who yells the loudest gets what they want type of culture because it’s like hey we need this and I’m going to tell you over teams okay we’ll build this so talk about the governance side of this yeah if you can’t do this right the odds are fabric’s going to be fragmented too because there’s also that very big part on what are we working towards what is the team working towards are we providing value and I’m just going to stop there because to This is dead door number one on what we

28:21 This is dead door number one on what we a company needs to focus on or a team needs to focus on. So I want to respond to your are we adding value? Obviously they are because the team exists given. given. Yeah. [laughter] Is it? No, it exists. Someone’s made the call. The fact that the team exists means it’s adding value. There’s there’s a there’s a BI reporting team and they build stuff. Great. The value is being added. Now, is it the is it the maximum value? That’s probably debatable, but someone has decided at the leadership level that there’s a bucket of money

28:52 level that there’s a bucket of money that we’ve got to spend to get reporting out the door. This is how we’re going to do it. So, at the end of the day, like I’m not going to argue that point. It’s adding value 100%., where I go after this one is you’re used to building everything in an importonly world. So when I look at this,, we have we have probably lots of repeated DAX definitions. We have a lot of model relationships that are probably being repeated. We have redundant data storage. Now, let me give you a little p perspective in a pro license. In a pro

29:23 perspective in a pro license. In a pro license, you’re not penalized for having repeated data. There’s no penalty there. The penalty becomes how many times am I hitting the source system and can the source system handle all the repeated queries for the same amount of data. That’s that is the one that is like where I would argue is the penalty of pro licenses. You can have as many models as you need. There is no threshold or limit. They all have to be under a certain size under one gig and it actually incentivizes you to build smaller models and more of them to keep

29:55 smaller models and more of them to keep it under the one gig limit. So in a pro license, if that’s what you’re doing in this organization, you’re ex you’re incentivized to build the repeating things because it keeps cost down. You’re under a threshold. It it allows you to keep working the way you’re working. So if if pro is where you’re going to stay, this is going to have minimal value back to you. If you’re going into fabric now, that’s where things I think start

30:25 that’s where things I think start changing and shifting and now you’re trying to optimize differently., just pause there. What do you think about that comment? Well, a lot of things you said are true for me. The only thing I’m disagreeing with is I don’t think it really pertains exactly to the situation we’re talking about. You said in the beginning your first reaction to me was spot-on in terms of honestly where the pain is is they started as a BI team. They just started working on things but there was no process involved in the beginning on

30:57 no process involved in the beginning on structuring semantic models and at some point you either have to pause all work to build from the ground up as the mailbag said or you’re just going to keep going to these weird weeds and roots because it’s very hard. No way that you’re not going to pause and go you’re not going to pause and start over. You’re going to strategically allow some things to be built. You can’t there’s no way like there’s no way you would ever be able to pause an entire team and say we’re not building anything until we have it all fixed. Like that’s what I’m saying. I’m not saying yeah like we’re hey the BI

31:28 not saying yeah like we’re hey the BI team will be back in two months. Please leave a message. I’m not saying that. Obviously Obviously you’re gonna have to you’re going to have to figure out how to migrate the plane as it’s you have to build the plane as it’s flying here a little bit. you’ve already you’ve already launched the system. You got to build in flight here for a bit. So, I I would really argue like that’s what you’re going to have to figure out how to do. Okay. A better way to say it is the production that you were producing if you want to start having the right structure is going to be smaller than it was because at some point like Mike, I I

31:58 was because at some point like Mike, I I dealt with this at my previous company. We had nine models for nine sales reports and eventually we realized we needed a gold model. Well, someone has to build that gold model with all the feature requirements and people have to get provide their input and then there is that migration. So, this is not there is that part of like, hey, we are building a better way here. So, we’re not going to be able to do all the tickets we were doing before. Again, not saying that the BI team is going on vacation to figure this out, but there is that point of where our

32:29 but there is that point of where our resources are going to go and they’re not going to go to keep building reports and doing the same thing. And the problem is the more you wait, the harder this gets. So the person who’s submitting the mailbag is definitely at a point where it’s like, oh boy. Like they’re looking at everything going, oh no, there’s all these DAX definitions. Then we have to figure out, okay, are you saying sales the same way I am? Are they two different definitions? There is Sherlock Holmes types investigations here before you even start building. And

33:00 here before you even start building. And that requires resources, which means that you can’t use resources the way you were before. And whether or not you’re willing to do that for the long term, it keeps becoming this weed of data. I agree with you. How can you argue that? [laughter] I’m going to agree with you. I’m going to agree with you to some point, but I’m

33:22 to agree with you to some point, but I’m going to also substantially disagree with a lot of what you’re saying as well. I I believe if you stay the development course of not using MCP servers and not using agentic spaces Oh, dude. Okay. You’re you’re you’re going like to your point, you’re right. If you take that off the table, what you speak of is very accurate. We need to sit down. We need to merge things. We need to pull things together. together. But if I’ve got an MCP server and I’m connecting to two different models or I’m using two PBIPs, like let’s go back to your example. I’ve got nine different

33:52 to your example. I’ve got nine different reports and we need a golden model of all of them. Okay, I’m not scared anymore of this stuff. Like honestly, the nine definitions of what is required has already been defined. We just don’t know how to merge them. And AI is actually really good at merging multiple models and telling me what is really duplicating and how many times are we saying sales or customer or what those things are defined continually. So, I’d also argue too, Tommy, I understand your

34:22 also argue too, Tommy, I understand your point around this one, and there does need to be a human in the loop for this. There has to be conversations, but I think your first pass at this without changing anything of your existing structure is take all nine models, drop them down as PBIP, and say, “Hey, give the agent instructions. I’m looking to merge these things down and make data domains. Tell me of this what should be common tables that are across all models. Tell me where there’s redundancy and repeatedness. Tell me where I could centralize and make common objects. Are there any junk dimensions I should be moving out into dimension tables or

34:53 moving out into dimension tables or flattening table? You can ask a lot of these things. Now, it won’t be perfect. I 100% agree it won’t be, but it’s going to give you like a short list of things you can go after immediately and start tackling them right away. I think that will be highly effective for that team. And then to your point, Tommy, you still need some sleuththing. There is some discovery work that has to come back and say look let’s really define what customer is look we have a number of measures that mean the same thing or are oversimplified the name is too simple based on the

35:24 the name is too simple based on the calculation being provided in the DAX it really should be named this that and the other thing so now you at least have you other thing so now you at least have the noise where it was just a know the noise where it was just a loudspeaker with a lot of noise in it you now have cut through a lot of the noise and you’re now down to a couple of the key items and I would argue when You do these deep dives and investigations into model consolidation, figuring out things. Most of the model, let’s call it 80% of the model everyone agrees about. Not a problem. I

35:55 agrees about. Not a problem. I think there’s a large portion, maybe that’s a bit high, but there’s a lot of portion of the model that people are like, “Yeah, I get it. That makes sense. That’s how we think about customers. That’s how we think about sales.” So, it’s not the 80% that I care about. The the agentic side can handle that part. It’s the 20%. It’s the these are the areas we’re having friction and you say you want data by week but that doesn’t mean the same thing to this person and how do we build that what will work for both teams we’re looking at the data two different ways like those are the harder conversations

36:25 like those are the harder conversations to have and that’s where you should be applying your people so I would agree to your point some level Tommy but some level level the the the tooling that we have been recently given in the last this is again you asked me this in November I would have agreed with You ask me this now, I think I’ve changed my stance on this based on the new tooling that I’ve been given., it’s taken me 500 plus episodes in 5 years to understand the difference between Mike Carlo and Mike D. Carlo. And what I’ve realized is Mike

36:55 D. Carlo. And what I’ve realized is Mike Darlo is binary. Why isn’t everyone using MCP? Everyone, every company should be using an MCP. You’re making a lot of assumptions here, guy. You’re making the assumptions that they have all the access to an MCP. Now, for me, I would push every company to do so at this point or to get there. But that’s not a lot of luxuries. Again, you and I live in this I sometimes feel like you and I live in this like but the case is so compelling though. Like, so I understand LA world where we can get any AI we want and pay for it

37:26 can get any AI we want and pay for it because it’s our own companies, which is great, but we who’s your IT team? It’s you. you. Well, yes. Yes. And I I get it, but like so my point is what is the value proposition that you’re pushing back to your team? Like I understand that Tommy but like let’s be real here right now. I’m not going to tell anyone to do this. This is a big no no. Time out. This is a big no no. Don’t anyone do this. Don’t say it because I know what you’re going to say. Hold on. Hold on. [music] [music] Let’s Let’s Let’s imagine. Let’s imagine

37:58 Let’s Let’s Let’s imagine. Let’s imagine for a minute. I’m entering the realm of things you should never do. Right? This is not things I want you to do. A new segment. A new bit. We’re going to do do a new bit. We’re doing things Mike says you should never do. Right. This is the Carlo. This is the Carlo. Let’s do this one. this one. Let’s imagine you did have these nine models. They’re on a computer somewhere. How hard is it? And again, this is a this is probably a failing of AI computers and aentic stuff. Anyways, if you really wanted to use these AI

38:29 if you really wanted to use these AI tools, let’s again, let’s just play I’m playing devil’s advocate for some reason, Tommy. Like, let’s just say I’m going to go buy my own. I There’s nothing stopping me from buying my own claw code anywhere. I could go buy that on my personal computer, whatever I want. I just need the PBX files, which some companies are doing a good job controlling them. I would argue a lot are not. So, what is stopping someone from going, I’m going to download these nine PowerBI files. I’m going to go get a free copy

38:59 files. I’m going to go get a free copy of PowerBI desktop and export them to PBIP and I’m going to go use a version of Claude for $20 a month and go talk to it about these things. What’s what’s what is the barrier? The barrier is so small to my opinion. So, one, this is a c this is a cautionary tale for two reasons. One, this is a cautionary tale for the employee saying you think you can do this, but that’s really probably violating your company’s data policies and probably not a good idea. The second thing I want to say is leadership of these companies, you are

39:30 leadership of these companies, you are handcuffing your people by not embracing an AI forward fronting strategy around you either give them AI or they’re going to do shadow IT to AI and they’re going to go take their files, move it to a computer that gives them value and they’re going to go do this. So, if you’re a leader, you better listen up because if you don’t give this stuff to people, it’s going to create problems and you’re going to have AI data leakage in your company. I’ve seen it happen.

40:00 in your company. I’ve seen it happen. It’s going to happen. It’s the barrier is so little. And the reason this occurs, let me give you why this happens. happens. It happens because the little amount of effort I put in, I get tremendous amounts of results back. Like going back to our example earlier, the 80%, right? I’ve got nine different models. I need to consolidate them. There may be same similar tables. What about this customer’s table versus that customer’s table which has the number of columns, right? right? The AI could just figure it all out in like seconds, in minutes. And so the

40:31 like seconds, in minutes. And so the value returned from an AI is so great, people are going to clamor to figure out how to go do it regardless whether you jump in or not. So I think it’s better for a company to say, “Look, we sanction these tools to use AI to help you do your job. You have to use these things.” Push back Push back you and then you can like,, you mitigate some of this. And then you’ve got actual traceability. Who’s using the tool? Is the data going somewhere you trust? Is it going to,, open

41:03 trust? Is it going to,, open AI on a personal account, which means they can see every prompt, every piece of data? Like there’s already company heavy policies that says you can’t you do this stuff. I am not I am not going to push back. I can only pull I’m going to give you a quick scenario here to say just how effective everything you said. I am doubling down because I’ll give you a scenario. I was doing one of those experimentations on things and it got complicated. I had a semantic model that was pushing to a lakehouse that was creating two more tables for me in the lakehouse and then I was pushing another

41:33 lakehouse and then I was pushing another semantic model to pull that new lakehouse data. On top of that, I had measures in both new semantic models. Very complicated stuff. Semantic model that was adding to a lakehouse using a notebook to create new tables creating a new semantic model off of that with report level measures. And it was just a pain. And I’m like, I could just do this all in one now. I figured out a way what I didn’t need. And all I did, Mike, I didn’t even open MCP. This was a tindle and GitHub copilot. I said, “Hey,

42:03 and GitHub copilot. I said, “Hey, take a look at these two folders, these two directories.” I complic and I said, I told it, I said, “I complicated things. things. I want this to only be a single semantic model. I want any report level measures or those measures to be added into the new one. So everything’s going to be import mode from this source.” And it was done. And I’m like, this would have taken me I don’t know how long. So, I’m going to completely double down to what you’re saying. And honestly, I am getting to that point too where if your

42:33 getting to that point too where if your leadership is going, “No, we can’t do AI or it is pushing back.” You need to really consider how much you want like no what you can do for your colleagues, your employees, the value you can give them because they have the skill, but they can work on other things. We’re talking about a project this nine semantic model idea that’s going to take months probably two months I would say if you’re not putting all your resources to it or it can take your company a week what are you doing you

43:05 company a week what are you doing you like in terms of so what are you doing so so but this is the point though I think there’s this nervousness around like what is AI doing are we going to lose skill are we going to have brain drain like I don’t think that’s the case I really don’t and I am I’m heavily investing in learning how to change how I do work. It’s making me more productive. I can do more things. I have more projects running. I can have t fully custo I said this a couple times before, Tommy, and it came up again in conversation. Hyperpersonalization

43:36 Hyperpersonalization of apps and workloads is going to be a thing. The AI makes hyperpersonalization almost a non-negotiable. You’re going to build like if it works good for Mike and his pattern and how he likes to do work, great. Build a custom app. I’m literally seeing advertisements now. I don’t remember who the who the company is. It’s maybe it’s maybe it’s cursor. I think it’s the company, but there’s a commercial happening that someone’s at a desk and they said Pam just built an app. Pam just built an app. Pam built an app.

44:06 just built an app. Pam built an app. Oh, it was in the Super Bowl. It was during Super Bowl. And so it’s been now running multiple times now and now it’s showing up on YouTube. But this this is this is [clears throat] this is the new way honestly like it’s going to be these little mini apps that are going to help build business process and to again going back to your comment earlier Tommy about power apps right back it’s exactly that it’s the exact same thing it’s

44:29 that it’s the exact same thing it’s power apps it’s right back it’s all these things however the difference now is we now have this ability to be able to build apps with just talking and prompting things as opposed to having to go through a framework that is power apps Again, I think Power Apps was like 5 years too far ahead of its time. I’m not going to argue. I’m I I’ve done with that conversation. Had Power Apps come out now, it probably would have like really resonated very well with people. So, so that’s where I’m I’m sitting on these things. Yeah. So, let’s get to the the real question here on how do you actually do fabric, right? If let’s say you are

45:00 fabric, right? If let’s say you are migrating to fabric and how can you make sure you kick it off the right way? Because the nice thing for this user and for the co company that’s in the mailbag is they are getting a fresh start because we’re going to migrate to fabric. So we get an opportunity rather than trying to fix our broken features here is really creating a new process create a new pipeline of how our data is going to flow through the company. And I’m gonna completely agree with the mailbag question here where

45:30 with the mailbag question here where or the you what they said is yeah it’s going to be more than it’s going to be easier to create fragmented data in fabric than even in PowerBI to create the semantic models without without the right governance in without the right planning. It’s going to make things more complicated if you just give everyone access and say just start building. Mike, I’m gonna start with a very basic concept here. A very basic answer that yet on its surface again seems like

46:01 yet on its surface again seems like obvious but to me is so important if you’re building a fabric diagram. How does you how do you want your data to flow? If I have sales data and operations data and employee data like my semantic models, write out the lakeouses. write out that this is where we will be pulling information from. If you need to see what operations is doing, we will have an operations data lakehouse who wants to take that on. Okay, great.

46:31 wants to take that on. Okay, great. That’s where you pull from and you make sure that the reporting team’s accountable to that. But you have to set that structure up that this is the way we are going to be pulling data. We have this great idea to there’s there is a cognitive model of what a lakehouse is that is separate than a semantic model. And let me expand on that. When I have a lakehouse, I am really providing for people more than just a data flow. These real true sources of

47:02 data flow. These real true sources of truth where I can build my semantic models, my sales lakehouse, my operations lakehouse, however you decide to go ahead with it. Well, as long as that is how it’s built and as long as you build accountability into this to say if you’re not using the operations lakehouse for operations, we’re going to have a talk. This will be penalized for you in your yearly review or whatever the case may be. But you have to start that way. that way. Lakeouses not built the right way is

47:33 Lakeouses not built the right way is going to be the hardest thing to try to migrate out of and try to do what we do semantic model. So few things there but just from the onset would you start like if this where are we starting where are you starting yeah this I’ve been as you were talking I was trying to make a couple little notes of things like a little bulleted list here of how how I would approach this how I’m going to go start this what I think is the right way so first and foremost assess what you have you’ve got a bunch of models this is where I would again I’m going to I I’m if I’m

48:03 would again I’m going to I I’m if I’m doing this I’m throwing an agent at this I’m downloading things I’m going to go get a bunch of what do we have currently. Where’s the data domains? And only not only assess what you have, get real numbers on it. Mhm. Mhm. Meaning, go look at the usage reports of the things you already have. There’s probably a bunch of stuff you need to kill because you planned really well to do create, read, and update, but you did not plan really well to do deletes or deprecation or get rid of stuff. [snorts] So,, go look at what

48:34 stuff. [snorts] So,, go look at what you have. go get some usage metrics running. Figure out which reports have the most views, the most interactions by people. That’s important. Once you have that, then you start looking at, okay, let’s evaluate. Again, to your point, Tommy, here’s the whiteboard. Look at your data domains. What domains of data do we have? Is it by department? Is it HR? Is it by sales? Is it by what do these domains of information look like? And the reason you do this is because likely and again

49:04 you do this is because likely and again it feels like this there’s a a rough pattern here. It doesn’t always hold true but rough pattern is a team of people need one or two domains for their team. Inside the domain when you look at the report structure itself you have potentially a semantic model that has many fact tables. In those many fact tables one fact table typically supports a page of data. Not always true but sometimes it happens. that when you look at like okay what fact tables do I need what report pages

49:34 fact tables do I need what report pages are we using you can roughly map out what information you need on what pages that’s something you can all get now one of the downsides and one of the reasons why I love using agents is if you have a report it’s very difficult to map what columns on what page of what report there are tools out there now you can pay for but AI makes this really easy I just did this yesterday preparing for my demo in Fabcon on which was here’s a report PBIR format. Here’s a semantic model with TIMDLE. Go look up

50:05 semantic model with TIMDLE. Go look up all the columns in TIMDLE. Go find all the areas in the report and give me an output a summary page by page what was used on what page and then I can clearly see what part of the model is being used in different areas. Awesome. Really good. Then my next phase is plan, plan, right? Write it down. And this is about planning your workspace strategy. How are you going to build workspaces based on control? Do you need a central workspace that’s only data related and fabric items? Plan on the onion effect,

50:36 fabric items? Plan on the onion effect, right? How you distribute data. Do you distribute data only on the report level only? Do you distribute data at the model level, build your own reports or do you distribute tables directly? This will vary between teams and there may not be everyone on every team that has every access to the layer of the onion that you can supply. But when you think about this, you just walk your way back through what is the stuff we what is the data contract I’m giving to the business team. team. Then I’m looking at all my data sources.

51:06 Then I’m looking at all my data sources. How do you refresh them? Is there full load incremental? Just high level doing that. You should already know that because of your existing models should be already set up. And then the next part is leverage one late catalog. That’s your discovery area when you think about what you’re building. This means semantic models with descriptions, semantic models with detailed tables, lakehouses with information in them. All as much as you can to describe what is the data that you’re bringing to the one lake and then in concert with the one lake, you start using certifications. This stuff is

51:37 using certifications. This stuff is certified, this stuff is promoted. Those are things that really help I think lock down what you’re actually responsible for. What domains do you really care about? So all these are very tactical things. I really think you have to plan this stuff out and then go into leadership and say we are BI, we’re central BI. We need a a federated approach mixing some heavy IT people with the business side and this is our plan. And when you come with this plan to the the leadership, they almost

52:07 to the the leadership, they almost always buy in. And I’ve had really good success with clients using this technique. Yeah. doc documenting, planning, communicating back up to leadership, getting the buy in where people have moved positions and roles and done a lot more now with fabric and now they’re building everything in fabric [clears throat] because they’re taking the next step to show I know what I’m thinking about. I’m I’m planning for proper usage and and handling things. So, that’s I sorry I’m sorry I said a lot of things there. That’s where I’m landing on that one. So I’m going to go on the

52:39 on that one. So I’m going to go on the other side of the coin. Not the disagreement, but everything you talked about to your point was it was tactical and no notes there. I that’s the same idea that I think is a very very best practice approach is very much aligned in the way that I I plan as well. But the there’s the other side that we need to consider and it’s really what you bolded also in the in our notes here and this is the team culture that you cannot disregard or undervalue how important this side is in the

53:09 important this side is in the integration side right because here’s the thing Mike we can talk about lakeous and we’re going to have the what domains but you have roles and people who at some capacity their skills are going to be changing when we get to fabric This is a great opportunity not only to get a refresh of your model but allow people to shine and allow people to take on responsibilities and grow in different ways because I have to consider as a director of BI or a manager the people below me and say who

53:41 manager the people below me and say who do I trust right and who can I give a great opportunity for to advance their own career because if I’m using fabric Mike odds are in our assumptions you’re going to be using notebooks Odds are you’re going to be you, you Odds are you’re going to be you,, there are new conceptual things know, there are new conceptual things that we have to learn. So, am I going to make everyone learn that and just build everyone off or am I going to give some more responsibility to some other people for handling those gold lakehouses, handling that pipeline, the medallion,

54:11 handling that pipeline, the medallion, right? And to me, this is a great opportunity to really align my team and to structure and organize my team. It’s like, look, you’ve been a rock star, so I’m going to let you lead this project here. I’m going to let you lead how the lakeouses are built around these domains. And this can’t be undervalued. And it’s funny, there will be studies done in the future about, hey, how AI affects team culture. I don’t know what that is yet because this is also I think that’s

54:41 because this is also I think that’s another little cap that’s a little variable that we have to add to this, too. So, I’m going to stop and I I would like the rest of the conversation to talk about this team culture and make because here’s the thing. At the end of the day, you can write a diagram, you can write a tactical approach, you can have everything laid out, but if people don’t follow it, I don’t want to say that word. It goes for not. So, with all that, Mike, what’s your opinion on that? Do you agree? What do you want to add or disagree with? No, I I agree with this one. I I also

55:12 No, I I agree with this one. I I also would argue, Tommy, I think I like this one., I also think that having metered access to this new world, you start small, especially when you’re going. You don’t give the whole team access to everything, right? You need to absorb what is a best practice in building this stuff. So, that has to happen regardless. You’re going to have to do that., so don’t give free access to everyone all the time., to

55:37 access to everyone all the time., to your point, Tommy, I think you’re right. There’s are there are a couple people or individuals that are going to be at the upper echelon or the high end of this that is like hey you are able to handle this I need to give you a special project right let them work ahead there’s I’m not going to assume everyone on the team as equal playing field right I would really recommend strongly not everyone gets a trophy right yeah so yeah but I would also recommend strongly that when we’re talking about things that are happening in these

56:08 things that are happening in these spaces, spaces, I would really strongly recommend like you have to be like willing to learn these net new things. Like if we’re going to talk about consolidating, what does that look like? Where do we put it? Does it make sense to go into lakehouse or do I need something else that’s going to be a bit more fluid like a data warehouse or a SQL database? You may be pulling data in ways that you weren’t usually doing them. This is all net new for you. So, you got to learn up. So go to learn. microsoft docs, go there, go

56:38 to learn. microsoft docs, go there, go get your PL300 exam, go get your PL600 exam. These aren’t like these aren’t the end all solutions of what you need to know for everything. But it helps you at least get to a place where you at least know the technology. what it’s supposed to be there for. You have the vision from Microsoft as to where I would apply this. And then you can rule out real time for certain things because you just don’t need it. Or you can rule out KQL databases because you just don’t need it right now, right? you can really hone in on where do you think we should work and I think for me that’s where a lot of this

57:09 think for me that’s where a lot of this team culture comes from leadership needs to be able to allow the knowledge learning education to grow right right and then you have to build process it’s got to have you’re going to you have to be able to build here’s how we’re going to market this is what we’re going to do if we don’t follow this we’re going to ask you to rebuild it like you’re going to have to have some gatekeepers here you can’t just keep doing what you were doing in the in the prior world which was let everyone have a free-for-all and just go build things. So again, going back to our my plan plan yes

57:41 back to our my plan plan yes note here earlier. A lot of this is just workspace strategy, right? If you have a workspace strategy, that really defines a lot of like where you put the data, how do you get it in, are we using pipelines, what when we transform data, is it notebooks or is it dataf flows gen 2? Those are going to impact your bottom line dollar costs on what you choose. And so if you’re choosing something that’s just easy to use, for example, data flows gen two, be prepared to pay more than using a notebook, plain and simple. So without someone telling you

58:12 simple. So without someone telling you that, you don’t know what to do. And so I’d also recommend listen to the explicit measures podcast. We’re going to tell you a lot of things you should do that are going to help you build better stuff. And we’re doing this at scale with lots of customers. So, you scale with lots of customers. So,, like you need to be plugged into know, like you need to be plugged into people that are building these things and learning from them because Microsoft builds the framework, but the community decides what’s best practice and that kind kind Mike as we dance on the edge of the volcano. Exactly. So,, but I I want

58:44 volcano. Exactly. So,, but I I want to add two fundamental pieces here and they’re more they’re not as tactical as much as they are or they’re not as technical as they are tactical. two things that I don’t think you mentioned and I apologize if you did but from the onset when you are starting any of these types of projects and especially with fabric especially with fabric you need to work with your team on expectation and goals what are we trying to get out of fabric are we just building everything is everything in a lakehouse why just because we can no like

59:16 why just because we can no like understand what is the end goal here or what is our first milestone here that we’re trying to do what’s the expectation for the team and for you because that’s also going to roll in the accountabilities but also it’s one of those like everyone has a shared goal right because I for me Mike if someone says you just can’t do this now this is not part of your thing you cannot access the data at lake house like I go the other way with that but if we start working on things like hey we want

59:46 start working on things like hey we want people to step up here who wants to step up our goal here is to start building our gold semantic mo our gold lakees is here. does anyone have that skill? Does anyone want to do that? This is going to be the way we do that. Who wants to help implement this? You get this opportunity to I think really enhance your team. But there’s that there’s that fine line between the accountability and then also rewarding people here that goes into team culture and is a to your something you said it’s a sword I will die on regardless of the

60:18 a sword I will die on regardless of the technology but especially when you get the you have such a great opportunity here with fabric because you you are starting fresh it’s not like we have to take everything and reconfigure it’s like I’m giving you a new land kind it’s like I’m giving you a new land thing to to build so take take of thing to to build so take take advantage of that and take advantage of building your team and that maturity. I think I like that idea, Tommy. I like the define. I would agree with you from a leadership standpoint. Yes, there needs to be a new vision statement. This this is a vision.

60:48 statement. This this is a vision. Yeah, Yeah, we’ve got a lot of PowerBI. Our vision is to consolidate and build more value with this net new thing, right? There’s needs to be a simple vision statement from leadership that’s telling everyone this is how we’re going to market. And then you can then come back behind that and have some real tactical, okay, we’re not going to bring everything. We’re going to go attack the most used data sets and start working on those things first. And then I think you have to also get some buyin from the teams of well why are you taking away my reports? Why are you changing things? What are we doing? And I think the the value you add

61:18 doing? And I think the the value you add there is you really speak to the onion effect again, right? We’ve been traditionally only giving you just the report. report. What we’re doing is going to enable us to build more things faster. It’s going to help us to be less redundant with what we build. But we can also give you create your own reports on top of a model we’ll share with you. For those of you who are really good in finance team and you want to do table level details, we’ll just give you table level access. So you really have to speak to fundamentals of like what’s really

61:48 fundamentals of like what’s really happening here. And I think the strategy is the story here is we are making a better data contract with us the data collectors. I love that. I love that. And you the business users. And so the reason you go down this pain paindriven path is because it’s going to be better on the other side. You’re going to give more access, more things, having faster data. Like all these things drive towards making it better for the team to work with the data that

62:18 for the team to work with the data that you’re providing to them. I think that’s really the MO here. You need to really emphasize that message and overcommunicate that to the team so they know this is the right way to go moving forward. 100%. Dude, I love today’s conversation., honestly, I don’t have a lot of closing thoughts. I think just to highlight some of the things we said in terms of it’s a fresh opportunity. Plan, plan, plan as you said and set goals and expectations is really I think what I want to highlight here. And also to your point, just use AI and don’t do any work. So, I’m going to throw down a little bit of

62:48 I’m going to throw down a little bit of like you don’t know what you don’t know, right? So, if this is something you’re doing brand new and you need just need, I would recommend go finding some consultant in not a big firm because they seem to they seem to do a good job on the sales side and not a good job on the execution side. Giant blueprint, generalized blueprint. Yeah, Yeah, I would recommend go to the MVP mvp. microsoft. com. Go find an MVP in your area and go ask them questions or hire them for a short contract, four, eight hours, something

63:19 contract, four, eight hours, something small, a contract. just get another set of eyes that that that will be worth its price. Even though MVPs are expensive and they do a lot of expensive things because they they are like the best of the best, but they have seen some things. They’ve seen some war stories. They know what’s going on. I I would I They know what’s going on. I I would, mean, you don’t know what you don’t know. You can learn this stuff all on your own or you can have someone come in and like at least talk to you about it. So, I really would recommend find an MVP. Tommy and I are always doing this stuff all the time for all kinds of organizations. We’re just showing up and saying here’s what we think. here’s what we’ve built,

63:50 we think. here’s what we’ve built, here’s how we found things that work well. Just doing that already sets your team up for big wins. Yeah. Just because you’re you’re you’re shortcutting a lot of pitfalls and things you would have to learn over weeks of time of experimenting and building and creating stuff. You can just cut to the chase and get something that’s going to add value much faster. And that’s that sells the project. Yeah. Yeah. Right. If you can cut through the noise and get right to we’re adding value now, let’s go. Right. Mike, I thought of a great idea for us.

64:22 Mike, I thought of a great idea for us. We should do live road shows of the podcast at different organizations. Just do the podcast. We can’t be out. We can’t be outing these guys when they’re when they have just such bad data culture. That would be that would be such a bad idea. Conversation. This team doesn’t know what they’re doing yet. They’re so lost here. Why would you ever buy their products? They can’t even engineer their way out of a paper bag. That would be the worst idea ever, Tommy. That No one would hire us back, Tommy. Tommy. That first one would be great. It would be entertaining and it would never happen again. Yes, 100%. All right, dude. This is

64:52 Yes, 100%. All right, dude. This is awesome. awesome. Well, thank you so much for listening to today’s podcast. This was a super great topic today. I really like this. Kicking off fabric the right way.,, moving from PowerBI solely into fabric, I think, is going to be something a lot of organizations need to go through and figure out. It’s definitely the right approach. I think you’re going to get a lot of value from it and a lot more efficiency and you add more capability for your business users. I I’m really excited about this. That being said, thank you so much for listening., we are recording a number of episodes because Tommy and I are traveling a lot in March. So, stay tuned. There will be

65:22 in March. So, stay tuned. There will be a lot of episodes coming out for members. So, if you are a member of the channel, we’d love you to become a member. You get all of our episodes early. As soon as we record them, the day we record them for the rest of March, you’re going to see them coming out. So, we’re actually doing we’re recording four episodes a week right now for the next two weeks. I think we’re week two of week three week of these., but there’s going to be three more episodes coming out that will be pre-recorded and you’ll be able to get them immediately on our members channel. So, if you like this stuff and you want to get access to it early, please be consider becoming a member. Tommy, where else can you go?

65:52 Tommy, where else can you go? Yeah, you’ll know when we’re sick of each other when we start really snapping back at each other. It’s like, oh yeah, that’s probably their 16th episode in two weeks two weeks in a row. [laughter] Yeah. All right. You can find us on Apple, Spotify, wherever you at your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. Do you have a question, idea, or topic that you want us to talk about in a future episode? Head over to power. tiff/mpodcast, 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 of

66:22 join the conversation on all of PowerB. tip social media channels. Thank you all so much, and we’ll see you next time. Explicit measures. [music] Pump it up. Be it high. Tommy and Mike lighting up the sky. Dance to the day to laugh in the mix. Fabric and A. I get your fix. [music] Explicit measures. Drop the beat now. Pumpkins feel the crowd. Explicit [music] measures.

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.