Fabric's Most Underrated Features - Ep.524 - Power BI tips

In Episode 524 of Explicit Measures, Mike Carlo and Tommy Puglia unpack the latest Power BI and Microsoft Fabric topics from the show. You’ll get a quick read on the episode’s biggest ideas, why they matter, and where to dig deeper in the full conversation.

News & Announcements

• No linked announcements were available in the episode description for this post.

Main Discussion

This episode covers the major themes, opinions, and practical lessons Mike and Tommy surfaced during the conversation. The transcript below captures the full verbatim discussion if you want the exact phrasing and context.

• Mike and Tommy react to the episode’s biggest Power BI and Fabric developments and explain what stood out to them.
• They connect product announcements to day-to-day practitioner decisions instead of treating the news as abstract roadmap chatter.
• The conversation highlights where teams can move quickly, where they should slow down, and what tradeoffs deserve attention.
• They share candid perspective from real project work, which gives the discussion more practical value than a headline recap alone.
• The episode mixes tactical advice, opinionated takes, and a few forward-looking predictions about what listeners should watch next.

Looking Forward

If this episode’s topics affect your current Power BI or Fabric plans, use the transcript and linked resources to identify one concrete change you can test with your team this week.

Episode Transcript

0:01 Dance to the [music] day to laugh in the mix. Fabric and A. I get your feels. Explicit measures. Drop the beat now. H feel the crowd. Explicit [music] measures. Hello everyone and welcome back to the Explicit Measures podcast. I still like that song Tommy. It still still rings. The kids aren’t asking for it as much except when someone news in the car like

0:32 except when someone news in the car like if a grandparent’s like, “Oh, you got to show the Pulia song.” Because yeah, explicit measures podcast right now. So that’s amazing. A little far away from their tongues. But my daughter is actually now doing data visualization. They’re doing like line graphs and I’m very it’s very hard for me not to become my father like because he’s an engineer. They go, “Wow,, this is my field.” Oh,, and then immediately they’re going to go, “Nope, not interested.” Yeah. You don’t you don’t want to you don’t want to push them too hard. You want to let them get there on their own. One thing I I pushed a little

1:03 their own. One thing I I pushed a little bit Well, I don’t think I pushed hard too hard, but I when my son was younger, he was like,, really wanted I think he was like around 11 or 12. He was like, “I really want to build video games.” Like, “Oh, great.” Like, he you you said it. Let’s jump in. Let’s go. And started showing him Scratch. And I started saying like, “You need to learn you need to get you need to learn Python. You need to learn programming.” All these things. He’s like, “Yeah, I don’t really like that.” [laughter] Then he turned 16 and things started clicking for him. That’s a 16-year-old. That’s crazy. So, yeah, he’s driving now. So,

1:34 So, yeah, he’s driving now. So, that’s when things clicked on again and he was like, “Oh, this is really exciting.” And and it seemed like there was like another wave of excitement and now he and I have been working on some I’ll call it lightweight AI projects and things. That’s cool. I want to do that with my kids. Well, I I honestly think, Tommy, I think we have it. I think we owe it to ourselves. The AI is changing how I think about just education. I think about knowledge. I think about what’s going on. Like I was on a talking with someone the other day and they’re like,, how how comfortable are you with

2:05 how how comfortable are you with switching out the back end of your apps or doing something with an app? We’re like, well, I have no problem with it actually because I pay for this 100% knowledgeable knows every bit of language machine. I’ve I’ve built an entire program in Rust before. I don’t write Rust directly, but I can talk to my AI about it. And I think there’s this there’s this side of like you you get to a place if you are able to prompt engineer things correctly. I want to coin a new phrase, Tommy, with you on this and I think this will resonate with you. you. I like new phrases.

2:36 I like new phrases. Well, it and it’s it’s not actually not a phrase, it’s more like a term basically., people are are saying like the the agent the agent engineer or the agent optimizer or something something along those lines. I think the agent creator is creator. I think people are going to probably lean on like the agent architect or the agent engineer. They’re going to be other people that are going to like use those terms because those are architects and engineers are common terms like for people who are experts at what they do, right? I think but I I really think the agent creator

3:07 think the agent creator and maybe it’s not a good term yet. It’s it’s where you use agents to create things, right? I don’t want agents to resolve and give get me the data. I want agents to assist me in storing a lot of links, storing a lot of articles, holding a bunch of information for me, right? I wanted to create the app. I know I want a report. Yeah, Yeah, I know I want to go back to this information over and over again. I just don’t want to make all the clicks to make the report. I I want it to just for me. So like that’s where I’m at right

3:37 me. So like that’s where I’m at right now is I want agents to create the things and I want to be really good at wielding the agent to help me create stuff. stuff. I like that because there’s a different side to that. I think when I really started with the agent it was everything was creating something but and I know I keep talking about notion but man alive like it’s not it is a freaking awesome tool right now and what it’s doing for me. It’s not creating you it’s doing for me. It’s not creating applications or the tools but know applications or the tools but it is absolutely my assistant my secretary I so I created a agent in

4:09 secretary I so I created a agent in notion it’s called the consultants consultant and after every meeting which is transcribed I change the status of the meeting to say agent pending it will write a checklist of do you want to update the milestone do we need to update the claw instructions do we need to change the scope here and I set the instructions to say if I do a check mark it’ll do it if not I can write something say leave this in the backlog. So all my notes, all my organization is this really this this partnership with the agent

4:40 this partnership with the agent and that’s where my that’s why I keep talking about notion because it’s not just hey I need this information in now single it is part of an entire workflow this context where really I’m like hey let’s take a look at the the diagram what are we missing here or based on the meeting that we had go ahead and update our instructions that we need to do and what how does this change the scope of our project and Mike I know that’s not creating but there’s another side of that too and this is completely agentic but so you’re seeing the

5:12 agentic but so you’re seeing the different I think Asian creator is one and there’s that Asian assistant too. too. Yeah. So Yeah. So yeah, I’m I’m gonna kind like this to your point. It’s it’s a lot of these I have a workflow that I should be doing or I’m what I think we’re starting to see. I think to your point too, Tommy, is is our new workflows and I think this is going to come out a little bit in in our our talk as well, the workflows that we’re starting to build that we’re creating now are actually evolving into something that is very heavily

5:42 something that is very heavily agent-based in general. And again,, I was talking with Matias one. Yeah. Yeah. Yeah. Yeah. I’ve been seeing the shorts, man. They’re they’re fun. They’re great. The the article I wanted to like touch on here really briefly is like because people are now leveraging the creator agent or the agent creator experience, the amount of GitHub repos have just skyrocketed. GitHub went down like I think it was last week, the end of last week or something like that. something

6:12 week or something like that. something GitHub went down at the end of the week and it was just a sheer volume of more commits, more repos, more I’m doing many more commits to a repo in a single day now with my agent and I’m having my agent work on things when I’m not attached to it. So I’m I’m having it build issues and work on things and it’s it’s really changing my workflow. And I think in the new economy or the new skill I guess or feature skill, whatever you want to call it, that we need to learn now is how to be efficient with

6:43 learn now is how to be efficient with our agents. I think right now we’ve been just throwing a lot of tokens at them. Yeah. And just say hand it because the tokens have been almost relatively free. free. I think people are getting very wise about this and especially now that there’s no free there’s no more free lunch. GitHub Copilot is now changing its licensing model., Anthropic has just kicked out Opus out of their lower-end models and now you can only use Opus at the higher paid models. So, you’re now having to spend more money and everything is moving away from the license like user user

7:14 from the license like user user licensing. Everything’s moving towards a pay for what you use model, right? If you use a lot, you pay for a lot. If you use a little, you pay for a little. I have a question for you and again a reason I love this podcast. We had two topics. We’re diving into something way more interesting here. is the onus on a consultant because we’re talking you and I individually about our token usage or our agentic usage. Sure. Sure. But again, if it’s this is really going to make an impact in organization, you to make an impact in organization,, it can’t it’s not just going to be know, it can’t it’s not just going to be me using it. I want a team to be using this,

7:44 this, right? So the onus on teaching and providing that skill because again I completely agree with you. Yep. This should be something eventually that’s on your resume as a proficient skill. skill. It will be. How does that get taught? Is that on the honestly is that a BI developer helping with that because if you’re utilizing the company’s data I am feeling more and more that in two years I need to be able to provide that as a service. Yeah. In some capacity. I think you’re way where is that going to be? Yeah. I think I think you’re way beyond two

8:15 I think I think you’re way beyond two years. I I think it’s way shorter than that. I think it’s like six months to three months from now. But is it going to be our space or is it going to be a completely separate consultant space? Yeah. I don’t Tommy that’s a great question. I I think there’s I think in the same way when you go interact with companies and you have like when you talk to someone at the business right they have specific domain knowledge around their business their tables where data comes from how do you get it in right they have their own like if you’re in the healthcare industry they’re going to be very narrow around like not narrow but like they’re going to be very focused on these are the tables

8:46 focused on these are the tables that we deal with right when you when you generalize data in like the the data space in general like if you step back and say well if you’re man manufacturing train, if you’re healthcare, if you’re distribution, whatever that is, all of them use data. All of them need some form of like a lakehouse, something like that, right? And and so if you talk generally around data structures and how you load data, that’s very transferable across every I think industry. So we

9:16 across every I think industry. So we have a lot of clients across multiple industries. when you get into specific industries, the experts there really know what to build for their specific area. And I feel like this is the same thing. There’s going to be a lot of generalist AI people, but they’re going to lack the language to speak through like, okay, how do you apply AI in the best way for data engineering? How do you apply AI in the best way with Microsoft Fabric? What does AI look like when you’re using data bricks? How do how does their tool interact? because everything the the while in in general

9:49 everything the the while in in general terms AI should be used in all these different places and you could probably not to your to this effect Tommy right right AI is so smart it already understands all of data engineering let’s be real like it knows all the patterns if you say I want a slowly slowly changing dimension type two like it knows exactly what that means and it can explain it to you can you explain this to me and I’m finding now I’m doing a mix of learning and de and working at the same time, right? right? U was just reviewing some legal

10:20 U was just reviewing some legal contracts and there’s some language that was requested by another lawyer about one of our contracts. I was like, hm, I don’t really agree with this statement. What risk does this provide to me if I accept this statement here? I’m going to I’m going I’m going and I go to right away to the highest most expensive model. I’m going right to Opus 4. 6 six high reasoning and saying, “Okay, opus, here’s what here’s what I’m given.

10:45 here’s what here’s what I’m given. They’re using this word versus that word. What is what is the difference here?” And help me understand the context. And he says, “Oh, by the way, here’s the actual grammatical language of how this would work.” Like, okay, great. I’ll I’ll make sense to me. And then I can go,, if I if I have still have gaps, I can still Google things or I can say, “Go find references to this information. Make sure you you to this information. Make sure you, tell me what it is.” So I think know, tell me what it is.” So I think there’s going to be a skill in general. I think everyone Tommy is going to need some level of agentic space. What I don’t know is how we’re going to

11:15 don’t know is how we’re going to implement in the harnesses that we’re given. given. And this is a tough one. That’s the part that I think will vary widely., notion is just a harsh a harness around a large language model. But there But there and an article I was just reading is harnesses are way more important than the large language models. I would completely agree with that and and again you have to understand what is you have to be able to look at something or a system that has a large language model in it. You have to be identify what part of this system is the

11:45 identify what part of this system is the harness versus what is the large language model. Yeah. And to your point Tommy like you love notion. notion. Notion’s an amazing harness around an a large language model that does a lot of orchestration of things. So I don’t know this this is a really interesting new world that we’re stepping into. There’s a lot of things going on here. yeah, any other news items you want to cover? I think this for going off of that and honestly a segue to a segue here is one of the skills that I I think are so important is with

12:17 that I I think are so important is with notion I have a page called cloud instructions for this certain project I’m working on. Sure. Sure. And anytime the scope changes or we have a meeting I go make sure you provide to the top what to focus on because again it’s a large page for instructions for claude. So, and you understanding how context works for an agent., if you just provide the whole thing, it is going to prioritize things differently. So, the other day I was like, I need to know if I can create if there’s a way to have a relationship between these two tables. That was not part of the original scope. That’s something they want.

12:47 want. Wrote the instructions, went to the claw MCP, read the page, had the model open, played catch with my son. He’s learning how to pitch and came back. It’s like, hey, it’s complete. Yeah. So,, he’s learning how to pitch as in like, okay, your left leg goes up, guy, not your other leg thing. But,, this is like baseball 101 stuff., you must be just eating this up right now. now. I’m a very proud papa right now thing because [laughter] I also taught him too. So, like he’s always he has to wear the Yankee hat when he goes out to play. So, whenever a

13:17 when he goes out to play. So, whenever a car comes down the street, I’ve taught him because there’s a Derek Jeter commercial. No, he tips the hat to them. Oh, it’s adorable. So, yeah. 5-year-old going. But but the fact though too with the MCP like I can just say take a look at this. Focus on the first part of the instructions. They were updated. Claude already knows that we’re working with notion and I can get something done again whether it’s for work or being spending time with my family and it was like yeah we are able to do this. by the way here are some

13:48 this. by the way here are some potential barriers if you’re trying to do this. Yeah. And I said, “Go ahead and write back to the,, to notion what you found.” Then I work back in Notion. But the MCP, Mike, I’m at a point right now, I would feel lost if the MCP of Microsoft said, “Nope, we’re not supporting MCPS has gone away for PowerBI.” PowerBI.” I Yes, Tommy. Even something’s good when you’re trying to do a little fix fix and even in a little fix, you’re like, ”, I need to add a date.” So, so I had

14:20 ”, I need to add a date.” So, so I had an experience yesterday just exactly around the MCP server within within Fabric. I think this is going to lead into our actual topic here of the main the main show here. But I had a an aha moment. I was looking at this model. Oh, and Tommy, this this this will this maybe maybe will make you laugh. I’ll be a little bit revealing here as well. I was I was working on a semantic model. It’s coming directly from a fabric SQL database. So one there’s the fabric SQL. I’m doing some data changes there like operationally data and then that’s being mirrored down and you can from that SQL

14:52 mirrored down and you can from that SQL database you can create a semantic model that is direct lake so the tables that are in the SQL database automatically update the lakehouse tables those are then able to be built into a semantic model using direct lake Tommy I I got into the MCP I was like even little things anymore ah I need to add a date table I should just use the MCP server I’ll just spin that up right it’s It’s easier to now just talk directly to I love talking to computers and just saying here’s what I need and writing

15:22 saying here’s what I need and writing out exactly what I want instead of having to like type out the the actual DAX code. So, hey, I want a date calendar. It’s going to be have a date field. It’s going to attach to this column in this other table. I need other columns that are going to be doing this, this, and this. Here’s the data type of them. Here’s the formatting I want for each of them. And it did a pretty good job of getting through and building the date table. It made the date table, put it in the model. The only downside, Tommy, was how this is this issue? This is a very nuance issue. Remember when you you have

15:53 nuance issue. Remember when you you have a date table and you have data that’s actually has your your created at date, it’s not just a date, it’s a created at date time. So I gave a feature that was just renounced that which was lakehouse with calculated columns. I was like no one should ever do that. Never make that happen. happen. Well dog on it Tommy. I could not get my date table to connect to my other table because one column was

16:24 to my other table because one column was date time and my date table was only date. Yeah. It was so annoying to me and I was like dog gone it. I need one more column in my source table that is just not the created date or created at time date time. I needed the created at day, right? I created at day is what I needed and I wanted to add a calculated. It only had 6, 000 rows in it. I was like dog on it. Yep. Yep. Yesterday I was just giving calculated columns in direct lake a really hard time. And here today I’m like man it

16:55 time. And here today I’m like man it would have been really nice to have a calculated column right [laughter] now. that would have solved my model and I would have been off to the races. So, there is a time sometimes it was I understand fully what needed to be done and built. I didn’t want to leave the direct link experience to to have to calculate that column and plus I didn’t want to make there was a whole bunch of things Tommy it just see SQL views are not materialized in the lakehouse you can’t use them in a direct lake model there’s all these things that are like I don’t really love

17:26 things that are like I don’t really love this and I don’t want to take my operational database and turn it into more of a reporting database so my now it’s just was there was some totally gotcha there was the operational database and I needed my reporting database reporting tables, right? The the model was not quite semantic model ready. And this happened to every project, Tommy. I know. I know. And you it does happen too with the lakehouse because sometimes when you’re using PI Spark, there are different date types or date data types

17:58 different date types or date data types for PI Spark versus Pandas, which can become incredibly frustrating. And also the UI sometimes. So, we had a another sorry other random really really random bug. Yeah. Yeah, you’re good. We had a notebook at one point that was writing data down into date timestamps and we were trying to do some like dduplication on the date timestamps. The UI the the table that was resolving the date timestamps was like truncating the dates. So it had like the date time stamp with like a number and then there

18:28 stamp with like a number and then there had like seconds, hours, minutes, seconds and like milliseconds. the milliseconds decimals was actually longer on some of the data and some of the functions were truncating that some of them were not. So what we found was we were having no duplicate records and we were trying to overwrite records and it was just a it was just wild but even the UI of the table that was resolving didn’t have the right precision for me to see the difference in datetimes and so we actually had to really go into the

18:58 so we actually had to really go into the problem print it out using like a print screen to actually see the full datetime stamp. Oh like oh they actually are are missing. We have to actually truncate these date time stamps to make sure that they’re the same piece of data or study record. Anyways, jeez. Yeah, it’s just it’s just nuances of data engineering. Like these are things that you just experience and you’re like it’s the little it’s the little things that you go you log it away. You’ll need it later in the future or something like that. Anyways, any other topics here, Tommy? I think I’m ready for a pretty a pretty

19:28 I think I’m ready for a pretty a pretty fun topic today. ramp us up here, Tommy. So, we’re going to we’re going to talk about our main topic now, which is fabric’s kind main topic now, which is fabric’s more underrated features. We these of more underrated features. We these are things that we find value from that we don’t really hear a lot on social media about. Maybe people aren’t blogging about it very much or they came out and they just hang around. So, let me just I’ll just like leave it there. Tommy, give us some other runup here. Tommy, what else do you want to talk about this topic? Yeah, I I think this is important here because I wanted to scope it out with you on okay, what constitutes something

19:59 you on okay, what constitutes something that’s underrated? And I think there’s two ways to look at this is whether there’s a lot of buzz and I think who is it for or like if we’re actually seeing clients wanting to use it if we’re actually actively working on it. So there’s kind actively working on it. So there’s two ways to take that again whether of two ways to take that again whether we’re seeing Microsoft talk really put a lot of releases on it. what are we seeing on X Twitter and blog articles and again whether or not that’s a actual interest from organizations and trying to implement that be it for their team or the organization as a whole

20:31 or the organization as a whole now I put out here Mike I have a few things here that it can be preview it can be things that are not necessarily in GA yet it could have just come out as well but that’s what I want to do and I think it’s important to say there are I wish I had the actual number of how many actual features or products or parts of fabric out there. It’s got to be over I I don’t want to just put a number out there but I I if you were to actually break down each thing you could do in fabric from VS Code to user defined functions to

21:03 defined functions to lakehouse to mirroring easy over 50 potentially hundred now and not all of them get the respect that they probably deserve or the recognition that they deserve. Yeah. And I think this is a sometimes you need to spend time. It’s also hard Tommy especially when I work with new clients around fabric. You step into fabric you’re like okay what’s

21:26 into fabric you’re like okay what’s what’s the best parts of like let’s build something. This is the story of well we already are powerbi users we want to step in we want to use data flows gen two. Okay that’s one way of loading data. Yeah. And the the common feedback I hear from people is there’s so many different ways to do things. We want to build it right or close to right the first time. We don’t want to spend a lot of time doing extra work. And I think there’s this there’s some right architectures on things. There’s also some liberty to change the architecture based on what

21:57 change the architecture based on what your team already knows. So fabric has the ability to do notebooks and SQL servers and data warehouses. And so if your if your team is more SQL friendly, you may spend more time on the data warehouse side of things that that there there’s a little bit of I think a price point trade-off. I think data warehousing or SQL side of things is tends to be a little bit more than what you would do if you ran Spark and notebooks. And I’m a I’m a fan of having whatever tool I’m using having a

22:28 having whatever tool I’m using having a large amount of ability to control how much compute I’m using or not using. So like Spark notebooks come with Python or you have a proper Spark SQL notebook or just Spark notebooks or Python PIS spark notebooks as well. Sure. Sure. So you have a lot of options here to like pick which sol works best for you, but you you’re trying to pick a system that the team will understand and be able to learn and maintain, right? I I don’t want to walk into anyone and be like, “Hey, we’re going to do this thing.” And you’re like, ”,

22:58 thing.” And you’re like, ”, I don’t know how to maintain this when I leave.” Dave, like the whole goal here is like educate your team into like you now need you now know what I know. That’s that’s the whole purpose of consulting, I think. All right. So, do you want to go first? You want me to kick it off? Yeah, I’ll go first, Tommy., I’m going to I’m going to be really wild here. I’m going to probably say an underrated feature, and this is something that I find myself leaning more on over and over again inside Fabric. It’s going to be the It’s going to be the fabric modeling MCP server.

23:28 to be the fabric modeling MCP server. I’m gonna take that one off the table right away. All right. what? That was one of mine, but I figured you were gonna do that one, too. Tommy, it’s more of this. I’ve used it a couple times now. I like what it does. It helps me. It assists me with a lot of things I know that need to be done., and and Tommy, it’s also it’s that and I can start building skills around common patterns I do in models, right? So it’s it’s like I can work with it and then I actually am able to leverage what I’ve learned or the or the

24:00 leverage what I’ve learned or the or the system or the process that we’re going through and actually make skills to do better evaluations on things. So I think it’s an underrated feature in that in that regard, right? It’s this is the first tooling that we can enhance ourselves and get more out of it just by spending time with it. There’s nothing else like this in fabric. fabric. There’s nothing else like this anywhere else. It’s it’s this idea of being able to use that. And so maybe,, secondarily, right, the first thing I’m using a lot more is the modeling MCP server. My main usage right now, I just

24:31 server. My main usage right now, I just did an experiment yesterday with modeling in the web. So I had the the model open in the service and I connected PowerBI desktop to the service model and started making changes that way. way. That was that was okay to me. I felt I I think I feel like I’m more apt to have PowerBI desktop be the source of the model and then using the modeling server locally through VS Code directly to that. I feel like that feels a little bit more fluid. It’s a bit safer it feels like to me as opposed to going right to the service. So all in all, I’m

25:03 right to the service. So all in all, I’m using PowerBI modeling MPC server on my VS Code and I’m using PowerBI desktop to do that work. What do you think, Tommy? I love that because the what you’re really saying here is not just the fact that AI is going to perform a task for you but really again compared to nowhere else in fabric with co-pilot is that you can in a sense customize it or really enhance it based on your own company your own skill

25:35 on your own company your own skill because of using skills because it can provide this additional context that’s not just in a fabric user interface I not just in a fabric user interface what that opens opens up right mean what that opens opens up right because I I now have I think four or five skills that I’ve created for power query for translating things from notebook to semantic models and also again again working with my own notion projects that it’s using those skills along with the MCP and so the again what you’re limited by is to use a cliche is

26:05 you’re limited by is to use a cliche is your own imagination of what’s possible because it there are no limitations on using a skill with an MCP which is absolutely insane. in terms of what you can do can do and there’s already and how do I describe this right as your company builds things there’s already in your mind regardless whether you’re doing it or not whether you’re writing them down or not you’re already forming you’re forming habits and patterns on how you structure the semantic model for your users let me let me give you one example here Tommy do you use the measures table

26:37 Tommy do you use the measures table Tommy like when you build a semantic model do you make a separate table and put put originally what all the measures What I try to do is in my fact table I try to only include columns that are either for a key a relationship or for aggregation and then hide those which then it becomes a measure. measure. Yes. Correct. Okay. So yeah but the idea I’m becoming more and more con so Tommy I’m I’m very much in that camp right now. I’m I’m really in the camp of make mult the measures really should

27:08 make mult the measures really should exist inside the fact table and where they’re calculated. And the reason I say this is because just organizationally in my mind, let’s just say you’ve got multiple fact tables in the semantic model, there’s a couple situations where a measure may span two fact tables for whatever reason. Then I think that’s a a bit it happens, but I think it’s rarer than just having a bunch of measures in a single fact table. Oh yeah. yeah. That calculate what you want. And so I feel like it’s really useful to have the

27:39 feel like it’s really useful to have the different factual pieces, the tables that you’re trying to aggregate data from. The measures that live there should function on that fact table. And if you think about star schema design, you really don’t want measures up in the dimension level of your models because it messes everything up. It makes it you put them on pages, it doesn’t those filter contexts don’t get pushed down to the fact pieces. Like it just doesn’t make sense to me as much. So, I really like having the measures where they’re being calculated in the table that they need to live in. And then,, that helps my mental model a lot. I get

28:11 helps my mental model a lot. I get really confused and I get really not it’s not it’s not sad, but it’s just more like it’s just difficult to go into a single table that has just a ton of measures in it. Like, what does this apply to? What tables does this work? You can’t you can’t physically see where it comes from. There’s no what I think should happen, Tommy. Maybe this is a tool I should build at some point with my agent. my agent. Yeah. Yeah. You should be able to to look at a measure and you should be able to like hover a measure and it should show you

28:42 hover a measure and it should show you either in diagram form or at least in text form this measure is calculating this from this fact from this table and has this impact on these other dimensions. Like it’s almost like Tommy remember how we had this metrics view thing? thing? Oh yeah. We were we were all happy about metrics views and I have a measure and then this measure has what it can talk to like what this measure can be impacted by right when I use this measure these are the columns that I will use to against that measure

29:13 will use to against that measure that’s something that we don’t have anymore and I feel like that needs to exist for the report creator experience when I grab this measure here’s all the relevant things that you could use with this measure and you don’t need another tool to do this. It’s already built into the semantic model anyways, but that would be something I would really like to see from an ease of use or ease of building standpoint. Anyways, just pause that right there. We’re getting there. We’re getting there. So, I love the MCPS. I’m curious why you say it is an underrated tool

29:44 why you say it is an underrated tool because, you and I talk about it all the time, but I guess it’s not really a huge part of,, other people buzz. I know Kurt is always talking about the MCP. Well, I think I think we’re on bleeding edge a lot of this stuff, right? When you walk into organizations and say, “Are you using the MPC server?” One, they’re like, “What are you talking about?” And then two, it’s like, “My organization won’t allow me to use MCP servers or my organization won’t turn on GitHub co-pilot or let me have an agent.” agent.” So, I think the barrier to entry here is

30:15 So, I think the barrier to entry here is more around I got I got the other companies that I think it’s a culture company issue at this point. Right. I think it is really scared of agents and what they can do and the power that they can they wield a lot of power, but there also needs a lot of control. Right. And I think Microsoft’s trying to really sift out this story of like how do you control your agents? How do what they do? I don’t know if you saw this announcement, Tommy, but do you use Azure Foundry at all?

30:46 Azure Foundry at all? I not as much as I would like to. Okay. there there is an Azure Foundry blog and then I believe let me just see if I can pull the article here real quick to make sure I I I so two things that Microsoft Foundry is building they’re introducing these things called toolboxes this was just recently announced so there’s a toolbox in fabric so all the tools that you would normally use it’s not quite MCP it’s more about like it could be like a skill it could be MCP it’s it’s it seems like it’s more of this combined

31:18 like it’s more of this combined experience but also So on April 22nd, Microsoft announced a new hosting agents in Foundry agent service secure scalable built for compute agents. So Tommy, we have all these things like these little open claws laying around or whatever you want to do. They’re shipping a microVM with an agent you can plop into it. Like so the thing about it’s almost like they’re building the little infrastructure to make a harness and the

31:48 infrastructure to make a harness and the computer all wrapped into one item where you can just deploy and use this whole item directly inside Foundry. And sure sure sure sure this is I think extremely important for people because the same value Tommy you and I get from using VS Code and an agent and that harness like there’s a file system it can write files down it can make markdown.

32:09 can make markdown. These other agents now hosted in Foundry are going to give us the same effect. It’s going to allow us to be able to use like a file system with an agent which I think really increases the the performance usefulness of that agent. Sure. Yeah. So that’s why I think this is un I think it’s underrated in the fact that like Yeah. I Yeah. So I think we’re still a little farther away from that. We’re on the bleeding edge. It’s going to be useful. Like it’s underrated in the fact that it’s people have even the [clears throat] ability to access it

32:39 [clears throat] ability to access it yet. yet. Mike, you’re going to love mine here. It is userdefined functions here or actually rather why don’t we actually say the my udfs which are user data functions I’m going to say not the DAX one. So user data functions and these are again the serverless Python functions that live in fabric that have their own rest endpoints. the inherent fabrics identity model can be invoked anywhere pipelines notebooks activator semantic model and this goes into the

33:09 semantic model and this goes into the transanalytical task flows so I’m kind transanalytical task flows so I’m going to combine those together we of going to combine those together we still I have not seen a lot of examples or buzz in the community I I haven’t seen a lot of get repos that have kind seen a lot of get repos that have taken off I’m not seeing a lot of of taken off I’m not seeing a lot of people really write about them either so but the possibil and I think it’s honestly because it’s such a open you honestly because it’s such a open slate it’s such a open box know slate it’s such a open box of what’s possible. That is one of those like, well, what to choose, what to do. What are those actually use cases? But let’s just think about what this actually replaces. And we’ve talked

33:40 actually replaces. And we’ve talked about this before, but things that are now in the graveyard to me. Power apps that are embedded,, the power automate button, thirdparty custom visuals,, any custom web apps or deep linking or even trying to do or your SMS or SSRS right back that was pageionated. those are in a sense kaput if you use transanalytical task flows. and again now that we also have AI to code if you are not a Python developer

34:10 code if you are not a Python developer it’s pretty easy to begin to scope out what you’re trying to do. I think the biggest thing where the the barrier for a lot of people or where the in a sense that buzz is is how best to use it like what are you actually trying to do u some of the use cases that are easy you want to refresh your semantic model on demand you can do that with a transanal task flow it’s a lot easier than power automate you can do that but this really sets a lot more capabilities I don’t like I don’t like having to pay

34:40 I don’t like I don’t like having to pay for power automate license to be clear right you can that too and that because that PowerBI is a premium,., you can do it with fabric. You can do it with Power Automate, but I don’t really love paying for something else when I can just go back to fabric and just do it there and just pay for this to use. And the the UDF’s fabric user data functions are really cheap to run. They’re you can run them a lot without a lot of extra overhead that too. And I think the other thing too is you can do on your customer like drill through pages if you have an

35:11 drill through pages if you have an account or a customer or a product that you’re looking at and you can add in since your drop down to say okay the customer is now a premium member or you customer is now a premium member or we need to flag this know we need to flag this all those fields can go into a button to say update the record. So you’re really dealing with that application. That’s really hard to do with Power Apps and it’s and again you need a Power App custom visual in there. It’s not just a button to do that. So this sets up a ton of cap again possibilities and then you can use VS code with the UDF extension to easily create that.

35:43 UDF extension to easily create that. So you are not you have all the tools available to create them. It’s just testing it out and allowing AI to begin that process. I think you’re you’re right about this Tommy. I think I think the power of the user data functions is extremely well developed. It seems like it can it can do a lot. It gives a lot of flexibility. What I what I think is lacking there is to your point, Tommy, it’s not really talked about in the community very well. It’s not talked about there’s not a lot of blogs on it. There’s not a lot of

36:14 lot of blogs on it. There’s not a lot of information coming direct. I don’t think I think it’s going to stay and live. It’s going to stay around. one thing I I wish it would do, I don’t know if this is actually possible or not, Tommy, and maybe someone in the chat could tell us if they’ve done something like this as well. I I would like to see Tommy the ability for a user data function to be called from a DAX function. So like I’m in a semantic model. I have DAX. I want to create a custom

36:45 I have DAX. I want to create a custom DAX function and in that DAX function I’m able to pass into that function some information, right? A table, a measure, a single value, whatever that is. I want to be able to pass something to it and have that go fire against the user data function in fabric and return some results. I don’t know if that exists. Yeah, that’s a little tough, but I think there’s I I’ve had some people just recently asking about how can we enhance the semantic

37:17 about how can we enhance the semantic model,, running on demand custom queries and like well we have to use direct for those things. So it it’s interesting. it feels like the user data functions are more developed around in concert with like the report, right? Yeah. Yeah. Here’s data from a report. I can pass that to the user data function using the report layer and then the user data function can go do something that the semantic model can see or or use, right?

37:48 semantic model can see or or use, right? Push the data back to a SQL server. Semantic model can then directly pull that real-time data back in. So there there’s a you could have a really quick turnaround time from data out of report to do something in use user data functions and then back to report again. Do you see like I think where the UDFs are now, do you see this that should be part of more than say 20% of reports? Is this something that you think? I don’t think so. No. Why? No. Why? I think there’s just a very large

38:18 I think there’s just a very large barrier around the to me user data functions are very developer centric. It’s a developer world. world. You write the whole thing using DAX. There’s no there’s no graphical interface. Oh, sorry. You’re you’re my I misspoke. you’re writing the whole thing with Python, right? It’s it’s functions. It’s defining Python functions, doing things offline. I think there are certain use cases and I also I think too, Tommy, there’s a little bit of setup with it. It’s easy to set up if you get the pattern right, but if it’s not working

38:50 pattern right, but if it’s not working correctly, if you need some debugging, if you’re it extends the experience a lot, but I think you really need to know the use case on how you want to use the user data function. Okay? Okay? Right? You’re not just going to be like, “Oh, I’m going to try it and just throw it in here and see if it adds value.” You you really want like the need is got to be one of these specific things, right? I need to send data from a report to somewhere else. I need to I need to capture comments and bring it back to the report.

39:22 But Mike, we have AI. You have a VS Code extension and co-pilot. So, I know you say it’s developer, but I’m going to go on your angle here where we’re talking about and I think ma Matias said this too. Matias said this on the AI podcast where like there is no coding anymore or there’s no I forgot what his phrase was but there’s no need for hard coding or the skill for code. So, wouldn’t that necessarily eliminate that barrier? I see what you’re saying here.

39:52 I see what you’re saying here. I’m going to use your own podcast against [laughter] you. Well, Well, to your point around writing the code, I think I would agree, but you need to also understand what’s the use case, where are we applying it, right? And I think I think the need or the effort to be able to do that is less or or not as well known or configured, right? So, let me say it this way. Even though it is

40:23 this way. Even though it is straightforward to use Python to help you write the Python code inside the user data functions not not arguing there right there there is a little bit of skilling that I think is lacking in that area right because there are pattern so one thing if you just say if you go to an agent and say write me a Python function that does XYZ thing it will do it of course of course but what it doesn’t know is it doesn’t understand the fabric way of connecting to data connecting to lakehouses connect. So there’s a little bit of

40:54 connect. So there’s a little bit of examples. So when you actually run a user data function, you have to go to the examples and pull down like here’s the example of how to connect to this. Here’s a connection. Here’s how to talk to an API. So there’s there are patterns they’re giving you like a starting pattern. Those patterns you need to bundle them up and give them to your agent so your agent understands, okay, when I connect to things, this is how I do it. So for me, I feel like that’s just a little bit of a barrier. the agent isn’t immediately aware of what’s going on for

41:24 immediately aware of what’s going on for that that that function. All right, let me ask it a different way. is if you have a client in a project that they are really looking to that solution would be a great one with UDFs is that something that you’re going to implement and then train them or are you still at a point right now where this is still a hard thing that I’m going to push to a a single client that’s a great question Tommy again I’m going to go back to the identification of the problem right I know I I have some very in my mind I

41:55 I know I I have some very in my mind I have like three or four really clear areas where using those things makes sense, right? I want to take data out of the report and I need to go do something else with it. If the client is saying something specific around that, I would recommend it purely from the fact that it already exists, you can use it. It’s all supported by Microsoft, right? So, I will recommend and build the solution for them. Also, once it’s built and you see how it works, it’s actually not too intimidating. It’s just the do you have all the bells and whistles set up all along the path to

42:26 whistles set up all along the path to get it to to right to to to be built, right? It’s not a it’s not as simple as drop a single button on a canvas in a dropown menu, select something and call it done, right? You have to do multiple steps to get it all together. Absolutely. That’s why I would like really hone in on the use case to make sure we really understand what it’s for. Plus, I don’t want to build this thing and walk away from it and be like, “Okay, we’re done.”

42:54 from it and be like, “Okay, we’re done.” And the client only uses it a handful of times, right? It’s a heavier effort thing. So, I would say I want to make sure it’s going to be it’s going to really solve the problem the customer has. Okay. All right. So, that’s that is my and I think we’re we’re getting there. So, what do you have for your next underrated feature, my friend? I’m gonna So, Tommy, I I know this has been around here for a while., I think with agents this gets even easier. I think my underrated feature is DANB.

43:24 I think my underrated feature is DANB. Now, maybe you could wrap this up into like two parts. Custom visuals and particularly the custom visual called DAB. DAB. It’s incredible, dude. Like the amount of stuff you can do with the DAB visual inside a PowerBI experience, I think is so powerful., I like the ability of being able to write visuals using Vega or Vegaite language. AI models are actually really good at building Vegaite visuals these days. Yeah.

43:54 Yeah. So, I I think that’s an underrated feature in general. I think this is something where if you’re looking for high customization, high stylization on certain parts of visuals or want to do something really unique and creative. I see a lot of people doing a lot of trying to use a lot of visuals out of the box and conform them to what you’re doing. I just disagree with the whole,, there’s so much click tax on the visuals. I really wish I really wish existing visuals had some JSON structure that was exposed to the user

44:24 structure that was exposed to the user that I could go modify and manipulate all at the same time. So, for me, I I think this is an underrated feature in general is some of these custom visuals and the one that I think I like the most or the one I find myself using the most even with clients building things for them. So, them. So, especially when I’m taking especially when I’m building really polished dashboards like if it’s a report that’s really polished or and especially maybe this comes up for me more Tommy than other companies because I do a lot of embedding for external clients and in the embedding space when you embed things you sometimes the customer just

44:55 things you sometimes the customer just wants a really spec the designer and the customer like want a very specific thing and they’re like we really want it to look like this and sometimes you need a lot more control around what you’re building for So for me, I really think the the the db and custom visuals is like a secret hidden thing here., one other note around this one. So maybe it’s broader more broadly custom visuals. I think I’ve seen a couple people on the on the the community. I think Kurt

45:26 on the the community. I think Kurt is one of these individuals. If you give PowerBI or or your agent the definition on how to build a custom visual, I’ve seen people vibe coding or or agentically coding their own custom visuals, just just building it. And this was typically a process that would take a long time to build without agents. You had to like write all the code, understand how it all worked, had to compile it down, make the PBI viz file, and then upload it. I think with agents, you’re able to shortcut this cycle

45:56 you’re able to shortcut this cycle substantially. So custom visuals I think become much more accessible to whatever you want to build. So I think there at some level there’s there’s an opportunity there as well. If you want to build custom visuals for your company or custom visuals that you want to embed into experiences that have a highly stylized specific way of doing something. I think now more than ever it’s easier for you to build those things. Again way technical. Yeah. Yeah. This is where I was going to push back on you because what you just

46:26 push back on you because what you just said about UDFs and now what you’re bringing about DANB I feel like if anything the DAB is a harder one too because that is an integrated part of a PowerBI report and if someone wants to make a change there there’s a lot of customization usually a lot of updates to a visual a user business user is going to do that I think that to me that falls into the same thing you were talking about with the barrier that you have with UDF right now and let me and let me maybe I can round up maybe more of my idea of why I’m saying that versus

46:57 of my idea of why I’m saying that versus UDFs. if UDFs I think are good but if you think about the the surface area of what I have to know like what I interact with if I’m talking about fabric udfs or doing task analytical task flows right if I’m doing that I need to have knowledge of the report buttons configuration the text input visual as well so you’re using you’re using a button you’re using like potentially a text input field on the page then you have to go make the UDF

47:27 page then you have to go make the UDF thing and then you have to figure out where like what is the action and maybe there’s like a SQL database or something behind the scenes. So if you think about the surface area of the things you must know or touch to get it to work there’s a lot more items that you need to be aware of. I think that’s more challenging than hey I’m going to put in this single visual and you’re going to focus all your attention on this one thing learn this one thing. Yes, to your point, both of them pretty technical, but the fact that you’re staying in one single tool to build out the DAB spec or

47:57 single tool to build out the DAB spec or visual spec, plus also, I would argue there’s a lot more examples around Vega and Vega light all across the internet. So, there’s a lot of really good examples to start from that you can go use those specs in your visuals and things. So maybe that’s where I’m gonna lean a little bit more, Tommy, is on on that idea is is the concept of it it’s how many different things you need to touch. UDF’s like three or four things, whereas in DAB or custom visuals, we’re talking about like kind visuals, we’re talking about like one central thing.

48:28 of one central thing. Okay. And all right. Yeah, they’re both pretty technical. I still would say yeah, they’re both pretty technical. So let’s do something a little less technical here. And I’m actually going to lean a little here on co-pilot and the copilot narrative visual, but also the copilot feature in PowerBI once you set up a like prepped for AI feature because when it’s set up properly and and unfortunately Mike it there I think you do need the necessary setup mod like with the

48:58 necessary setup mod like with the feature in desktop to prep your it’s in web now too but once you do that being able to have that chatbot with your data, but also the new feature, the narrative visual that now supports co-pilot. The narrative visual by itself was already something I I did a training last week. Yeah, last week. And I had that in one of my reports and someone I brushed past it because I’m like, I’m not sure if this is one of those things that’s going away with those AI,, visuals that

49:28 those AI,, visuals that Microsoft said that they’re getting rid of. I don’t know if it’s going to be just co-pilot or they’re going to keep that narrative side too. And they’re someone’s like, “Can you go back?” They’re like, “What was that?” I like, “Yeah, that’s just that autogener.” It’s called a narrative visual. And they were like, “This is that’s awesome.” So now with Copilot to have that same context on a page, which I think is much more powerful than the chatbot because again, the hardest part for Copilot right now with data is what to ask, right?, like in a sense having those set questions. So to have that

49:59 those set questions. So to have that with co-pilot, this is at this point a thing that you have to at least dive into and explore. It is worth exploring and really vetting out. If I’m not saying it’s a game changer or essential part, but it is more than worth the squeeze in terms to say, can I really incorporate this in my reports there? As much as I don’t love it, my me personally, I don’t love I don’t love it.

50:29 it. However, the amount of times we get asked to write descriptions about what’s happening on the page. Mhm. Mhm. There’s there’s something very fundamental about that pe people the the number of asks I get around this one when I demo this when I show this to people like that really feels like a feature that while you can have all the pretty visuals in there and this is and maybe Tommy this goes a little bit more into

51:00 Tommy this goes a little bit more into when we’re building visuals and the information about like how do I read this page right you can you could throw a bunch of visuals on the page and maybe this speaks a bit more towards data literacy, literacy, right? right? Interesting. Right? It it’s it talks to like what is your team what is the data literacy of how people are going to access your report. Right? If I think about this conceptually, right? I’m going to build a report. I’m going to have visuals on it. There’s a way that me, the author, I’m creating an experience for someone to go look at this report.

51:31 to go look at this report. odds are there’s some story I’m either trying to create or some general insight I’m trying to drive towards., towards., a lot of times I find it’s very useful to put helpful text on the visuals or on the page. When this chart shows this, here’s what it means. When this chart shows this, here’s what it means.,, here’s how to do drill through. Like little details like that in the instructions or or features about the report. I think these co-pilot narratives are

52:03 I think these co-pilot narratives are aiding with that experience. Does that make sense? Okay, I I agree. Yeah, I I I think I completely agree, but again, it’s nothing I see people like the only person I’ve seen write about this really is Chris Webb and I haven’t seen a lot of,, Microsoft has been showing the updates to it, but I would love more for deep dives into the different use cases how best like when we had the it’s still available, but the advanced analytics

52:34 available, but the advanced analytics like the prediction for a line chart, right? I forgot what it’s called and I know it was a seasonal one like SR S O A R S R S you’re talking about forecasting I believe believe forecasting yeah but they but it had prediction models is one of them I thought it was yeah that’s one of them and the other one I know start like S O R like something seasonal you can do for basically monthly anyways but they had on the documentation and on the blog like deep dives into

53:04 blog like deep dives into like how to make it best work even with the dashboard Q&A feature. They said how best to work for your model. Here’s the use cases. Here’s how to prep your model and I want to see that with Copilot right now because again I think it’s one of those it’s just a very empty box and like UDFs in a sense where okay it’s cool but how do I best set this up? What what are what is the minimal viable thing that I need to do so this works correctly? Because it’s so easy to do one thing go okay it doesn’t work. You

53:34 one thing go okay it doesn’t work. You one thing go okay it doesn’t work. it doesn’t provide what I want. I’m know it doesn’t provide what I want. I’m not gonna keep diving in. What do I need to configure? So, I would love to see more in that area. And and maybe Tommy, I don’t know if this is where you’re thinking about this one too, but if I look at this, I’m also looking at it does a good job of summarizing the page like it like that’s one of the main features that are highlighted like summarize this page so I can just look at the visuals and see what’s going on. Yeah. Yeah. 100%. So, and just little

54:04 Yeah. Yeah. 100%. So, and just little things too. You don’t need to just tell you who to call, but you probably could. But just some basic things where again, help a help a person out when you build like this is one thing I’d like to see more interactions with and and maybe I I don’t I have this in the embedded space, but I don’t have a lot of this other places. But it’d be nice to see the usage on that just just in general like when you put it. So, it. So, I’m just hesitant to have like a if I’m doing all this work to put like visuals and things on the page that are trying

54:34 and things on the page that are trying to explain what’s going on. Does it re are people really reading it? Or is it more of a wow, I normally would have to write all this text? That’s complicated. I don’t have to write it anymore. I’ll let co-pilot write it because someone requested it. And now it’s just easier for me to like just throw this at an agent or co-pilot summaries and then I don’t have to worry about it, right? Does that what I’m saying? Tell is anyone actually reading like there’s no interaction with it. you you can maybe track the clicks like there’s

55:04 can maybe track the clicks like there’s some clicks here on like references, right? So it’ll give you like footnotes about what’s inside. Hey, this insight came from this footnote. You click on the footnote. The footnote will highlight something on the page. Cool. Like that, right? Interesting. Sure. Sure. Yeah. But I don’t really have any way of counting the number of clicks on that visual to tell me what people are interacting with. So to some degree, I’m like a little bit not lost there, but I I really would be curious. Are people finding real usage if we throw that visual down? And maybe that’s

55:34 throw that visual down? And maybe that’s why it’s underused. Well, and that’s exactly what I’m saying. I think there’s a lot of setup that people are not realizing to get it to work the way they want. There’s that prep in the model and then it’s also those form of the questions where I think it has just so much potential. I like it. No, I I’ll go with it, Tommy. I’m with you on that one. Do you got anything else? I got one more maybe. You want to move? Okay. Do you want to do one more? We need to wrap. wrap. Let’s do one more. Let’s do one We’ll do one quick one. Okay. This will be a quick one. I I think maybe it’s unrated, maybe it’s

56:04 think maybe it’s unrated, maybe it’s not. I really like I use a lot of one leg shortcuts in general. I had that one too. I had this one, too. So this is like our our final one would be leverage one leg shortcuts and I also think Tommy I think about the bronze silver gold stance or even just bronze gold raw process whatever you want to call it like whatever pattern you’re trying to absorb as an organization right how much rigor do you want around that in my in my

56:35 want around that in my in my yesterday example of trying to get a semantic model to report on things I needed an extra column right the data format that was given to me was not in the right format. I needed to ship shape some things. So, constantly I’m doing land data down, pick it up, slightly change it or enhance it, add new columns, format some things. That way it it starts conforming more to I want in the reporting layer. So, because of that, shortcuts are interesting to me because you can use shortcuts as a control surface area for what you give access to users. So, imagine for me,

57:06 access to users. So, imagine for me, you’ve landed a bunch of tables in your gold layer. Your gold layer could be a schema inside a lakehouse. It could be a separate lakehouse that is just gold, a gold lakehouse, whatever that may be. But then imagine you want to start giving access to those raw tables, right? There’s different I I I think of of PowerBI as like different layers of an onion, right? The surface layer is pageionate reports and PowerBI reports. You can view them, you can interact with them, but you can’t edit them, right? The next layer up is now you can edit

57:36 The next layer up is now you can edit reports. The next layer up is now you can build your own semantic models. The next layer up is like creating the semantic model. Right? There are certain members in departments when you start making this federated approach of your data ecosystem. There are opportunities where you do want to give access to tables in a lakehouse but not the original tables. Not the original lakehouse. So what we’re doing now is we’re building like call them satellite lakehouses with

58:07 call them satellite lakehouses with shortcuts back to the original lakehouse. So there’s a central managed certified lakehouse. It has tables in it. And then certain teams are asking for well I just need to see the raw data. I want to build my own semantic models. this is something that we’re comfortable with that department owning this or giving them a handoff of here’s what you’re responsible for now. Here you go. And so now we’re using other lakeouses and we’re just shortcutting back to the original lakehouse which I think is

58:37 original lakehouse which I think is really reasonable actually and and a good good surface control area to let other people build on things without actually be having to actually have the data copied. So that that part alone I really like that. And I think too you you add the shortcuts with like materialized lake views and open mirroring too which is it is something that I think we should I want to see more workflows more best practices approach like in a sense the architecture around this because mirroring I think right now a lot of

59:08 mirroring I think right now a lot of people look at as a as a doing it to a single table or a single a source but I think there is going to be the need like show me examples of the best way to architect this. Do you have the mirroring in its own workspace?,, or is it just whatever the connections needed?, so again, it’s just one of those things. I’m not seeing a lot of, hey, I’ve done,,, this this approach. There’s,, this methodology that works really well., if you’re trying to scale up, we would do this. So, I

59:39 to scale up, we would do this. So, I really want to see that more because it is so powerful. I I definitely agree with this one, Tommy. It definitely helps when you’re it’s part of I think it’s really part of that data governance story. Who is going to control? And what I’m finding a lot more is data governance is probably an understated data topic that we should be talking about because once I have tables, once I have data either at the table level, semantic model level or the report layer, there’s always a transition of responsibility, right? and then as

60:11 responsibility, right? and then as you open up more responsibility to satellite teams, the federated approach, you need to have a better handshake as to we will we will make this semantic model. This will be the the model that we’re going to hand to you business you can build what you want. And I’m going to go back to a little bit Tommy here around the enterprise semantic model, right? We still don’t have that yet. So this is what I think data bricks is doing inside the metrics view which is trying to generate here’s a

60:43 trying to generate here’s a analytical view a metrics view of all the tables and things you care about in your business tables relationships the metrics you care about to to calculate together and it simplifies it for the business user. What they’re doing is they’re building a semantics layer for a business user to write SQL and build visuals and reports. PowerBI has the same thing except all the data is loaded and it’s basically inside the semantic model. But we have all these domain-based semantic models

61:13 all these domain-based semantic models and I don’t really have a single semantic model that describes my entire company. I I think that should exist without any data at this point. And maybe this is where Microsoft’s pushing the ontology or the fabric IQ pace towards, right?, in order for you to talk to an agent and say, “Hey, agent, go find this information from all the semantic models I may have across the world, right? First, it has to go find, okay, where’s the central place to go find all this, right? It doesn’t want to skim or scan all the semantic models independently. It actually needs a central location to go find all the

61:43 central location to go find all the data, right? Everything should be registered to one single location. And then the agent will be smart to figure out, okay, I’m looking for these items, Tommy’s total sales, Tommy’s year-over-year change. It can then go through and comb, oh, here’s all the measures I have found. Would you like me to pull together or show you where the existing models or lineage comes from? So, I think there’s this discoverability piece that doesn’t exist yet. And I think agents are going to need to be incorporated as we as we think about what we do in the future. Tommy, the

62:13 what we do in the future. Tommy, the agentic space is really good around finding getting patterns back to us. And so if we rethink one lake catalog, if we rethink data governance in lie of an agent, I think we need to start having like more of these metadatriven central places what purview is doing. But I don’t really like working with purview and it doesn’t seem to fit our needs very well. I really agree with that. Right. It’s like a purview like thing, but I needed to be more designed for agents, but I also need to be living

62:43 for agents, but I also need to be living in fabric only. There’s there’s a couple weird thoughts here that don’t quite fit for me yet together. Regardless, at the end of the day, shortcuts is fundamental to all this, right? Data governance, distribution, getting information between different systems without having to physically copy the data. I think that’s a win. Anyways, I think it’s an underrated feature and people should explore more around shortcuts. All right. right. Completely agree. That being said, thank you all for listening to this podcast and hopefully you found a couple features here that you resonate with or

63:13 features here that you resonate with or maybe ones that you’re not using that you should maybe spend a little extra time looking into. we thank you very much for listening today. We hope you had a a good,, learning experience, run, bike, whatever, whatever you’re doing during the exercise, commute, mow the lawn. I’ve heard a lot of people saying they mow lawn to listening to us on the Explicit Measures podcast. So hopefully that was fun. Hopefully you enjoyed this. Tommy, where else can you find the podcast? You can find us on Apple, Spotify, wherever get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. Do you have a topic idea or question

63:43 Do you have a topic idea or question that you want us to talk about in a future episode? Head over to PowerBI. tips/mpodcast. And finally, join us live every Tuesday and Thursday, a. m. Central, and join the conversation on all PowerBI. Tips social media channels. Thank you all so much, and we’ll see you next time. Just pump [music] it up. Be it high. Tommy and Mike lighting up the sky. Dance to the day. The laughs in the mix. Fabric and A. I get your feels. Explicit measures. Drop the beat now. His kings

64:15 measures. Drop the beat now. His kings feel the crowd. [music] Explicit 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.