Agentic Skill & Report Design – Ep. 530

Episode 530 explores how agentic coding is reshaping report development, from faster prototyping to new expectations for model and visual design. Mike and Tommy connect the May 2026 Power BI updates to a practical question: which skills still matter most when AI builds more of the stack.

News & Announcements

• Power BI May 2026 Feature Summary — The May release is packed with updates across Explore, Copilot, semantic modeling, and authoring quality-of-life improvements. Mike and Tommy focus on the practical impact: features like Explore perspectives and numeric input slicers can directly improve adoption when business users need simpler, guided reporting experiences.

• New Power Query experience in Power BI Desktop (Preview) — Microsoft introduced a redesigned Get Data flow with clearer source discovery, smoother connection/auth setup, and accessibility improvements like keyboard navigation and dark mode support. The hosts call out how this reduces onboarding friction for both new users and teams moving between Desktop and Fabric experiences.

• Five Minutes to Wow — Kurt Buhler’s piece frames agentic development as a new “lightning moment,” similar to early Power BI’s rapid feedback loop. The episode uses this article as a springboard to discuss where AI-assisted app building is already delivering value and where teams still need stronger validation habits.

Main Discussion

The core conversation asks whether report developer skills are changing or simply being repackaged by AI tooling. Mike and Tommy debate where agentic workflows accelerate delivery, where Power BI still needs better UX for advanced builders, and how teams should balance speed with verification.

• Agentic coding vs. “vibe coding”: Mike argues for intentional agentic development over ad hoc prompting, with tighter requirements and clearer outcomes.
• Explore perspectives are promising, but incomplete: they like the direction, but push for multi-perspective switching and stronger role-aware behavior to make the feature usable at scale.
• May update highlights with real user impact: numeric input slicer enhancements, faster web modeling, and GA status for visual calculations/custom totals all reduce friction in day-to-day authoring.
• Visual calculations still raise portability concerns: both hosts note that visual-level logic can become trapped in one visual unless reuse patterns improve.
• Desktop vs. Service remains a live tradeoff: Tommy emphasizes Desktop depth; Mike points to a steady shift toward service + agent workflows for speed.
• The skill shift is toward architecture and validation: even when AI generates code, developers still need to define constraints, verify outputs, and reason about model behavior.
• Report design still matters deeply: no matter how code is produced, model clarity, perspective design, and user-facing experience continue to determine whether reports get trusted and adopted.

Looking Forward

Use one current report this week to test an agentic workflow end-to-end, then document what AI accelerated and what still required manual model or UX judgment.

Episode Transcript

0:23 Hello everyone and welcome back to the Explicit Measures podcast with Tommy and Mike. Good morning Tommy, how are you doing? doing? morning Mike, how are you doing? I’m doing well, thank you very much. Appreciate you asking. We are just clipping along today. It is a Thursday. We have reached episode 530 and we’re going to jump in with some really interesting things today. Let’s talk about an article that our main topic today will be all around going through Kurt Bueller’s He’s got Data Goblin’s, that’s his blog and he has a 5 minutes to wow, the

0:53 and he has a 5 minutes to wow, the surprising optimistic similarities between vibe coding and now Power BI reports. so he’s he’s now going through this experience and much to what Kurt is doing, I’m also going through much of this 5 minutes to wow and agentic coding is changing how I interact with Fabric. It’s changing how I interact with semantic models and I’m getting a lot more value out of Power BI and reporting in general by leveraging more of these

1:23 in general by leveraging more of these Kurt’s saying vibe coding. I think that term in my mind, I want to kill that term. I don’t really like the word vibe coding. I agree. I like agentic coding, I think a lot better or agentic development. I think that feels a lot more sophisticated than just vibing things together cuz a bit more vibe feels a bit loose. But I feel like agentic coding is a bit more intentional with the design and and some focus to like what you’re with the outcome to be. Well, Mike, the problem with that though is you have to understand we’re not millennials and there are just the millennial term and slangs.

1:54 and slangs. Yeah, bet., like I don’t like stuffy. My kids say that. No cap. It’s a lie. Or that’s cap. Good. Or you got mord or something. I don’t know. This If you’re young, you’re probably not listening anymore. But [laughter] No cap. All right. No cap, yeah. Right. Right. Let’s move on. so, with that being said, we have a ton of really interesting news announcements. A huge update that came out from Microsoft. so, let’s go check into that. Let’s go check out the the update here and see what we got for that one as well. Tommy,

2:25 what we got for that one as well. Tommy, kick us off with the news and or updates. Let’s go through those first. So, the major one is the May Power BI updates are out and it’s a ton. If you have been wanting, I wish Microsoft published 25 to 35 major updates to Power BI. I feel like you got your wish here in May. I really thought they were impressive what they did here. And we’ll just go through I think the ones that we really liked. I want to start off with someone

2:56 I want to start off with someone something that doesn’t sound as exciting. exciting. But it is right off the bat and I think it’s going to be really important from an adoption point of view. And it’s the set the perspectives and explore. So, we know what perspectives are. Perspectives are a way to show and hide certain parts of the models for certain users. Previously up until this point, if you wanted to use perspectives, the let’s say you had just the marketing tables. The only way to do that was the cust custom personalize a visual. Which

3:26 Which not a lot of people use. Now, finally, they were combining that with the explorer feature. If someone wants to look into a model, you can actually set a perspective there. And say, what? I want when people are exploring not to look at the model the way I built it, but just only show certain columns or tables or measures. And for me, I really love this feature because you can say, “Hey, you’re viewing a subset of the data called the download perspective.” Or you can set the hey marketing leadership perspective. So they’re only going to

3:57 perspective. So they’re only going to see what’s probably most relevant. Like we’ve talked about the explore feature. We I know we’ve done episodes on it. Yeah. Yeah. Something I do use quite a bit and when I go through the adoption process is something I really encourage. It’s It’s part of the training I do with the consumers. What do you think about this? What’s your take about that feature now available? Perspectives has been a long under served area of Power BI the man the semantic model portion of this. So I

4:27 semantic model portion of this. So I think this is good that it’s getting some love. It’s getting some highlighting here. One word of note here. We build a product called Intellexos. Mhm. Mhm. We have there is no way to embed a exploration that Power BI gives you. So we had to rebuild or reimagine what an exploration would look like for an embedded application. We built a custom experience. Holy Holy custom. Here’s the dimensions. Here’s the measures. Yep. And we in our tool use perspectives to highlight what should be shown. It’s makes a lot of sense.

4:57 It’s makes a lot of sense. Mhm. Mhm. The weakness of this feature and so I like this feature. I think this is good. But when I read the feature I’m going to I’m going to need to play with this directly Tommy because it says your if you’re using a large semantic model cuz again a lot of tables could be just junk. There might be tables that don’t aren’t linked to anything. You don’t want the user to see these things. Right? You can provide consumers and in the explore feature a focused list of tables or fields measures or things like that that the perspective can show, right? Great.

5:27 right? Great. When I’m in the exploration, what happens if I have three or four perspectives? Can the user select which perspective they’re looking at? That’s what I don’t see. Yeah. Yeah. And to me And to me It’s You choose a perspective. That’s what Steve Yeah. The assumption here is there’s one perspective that you’re going to use in only the one exploration. I’m like, “Well, I that’s not how I use it.” point. That’s not the point of the point. point. point is have many perspectives with many options here. So, this is a really solid feature. We’re heading the right

5:57 solid feature. We’re heading the right direction. I’m hoping there’s more feature coming with this one, and I’m going to directly provide feedback on the community blog about this feature specifically that’s like, “Look, one perspective is not enough.” Right. Right. You need to have multiple perspectives, and you need to provide me a drop-down menu at the top of the data panel, the data pane data pane with a little perspectives drop. See See And And that needs to be there, and it needs to be in desktop when I build models and reports, and it also needs to be in the service when I edit and build

6:27 be in the service when I edit and build reports there. I saw it a bit differently. I see what you’re saying. To me, I feel this almost should be part of like role level security in in regards to the process to set up users to a role level security. Because to me, it’s hey, team A should see perspective A. Like when they go into it, rather than having to do the drop-down, team A sees perspective A. Team B sees perspective B, etc. Because I like the drop-down feature, but there needs to be more user personalization, so to speak, or user

6:58 personalization, so to speak, or user specific, team specific. That’s how in my head perspectives work best. Can you bind a perspective to an individual user and have all the UIs adhere to that? So, Tommy, what you’re explaining is like I would maybe argue is like a a level two of what I described, right? I just want the feature to work. Right. Right. even I’m not even there. You’re already adding like available, yeah. You’re You’re already adding like we need to have security, and we want to bind it to like, “Okay, when Mike shows

7:28 bind it to like, “Okay, when Mike shows up, he gets this perspective on this,, report build and this, you , report build and this,, all the things.” Makes I agree know, all the things.” Makes I agree with you, Tommy. Like the feature you’re describing is right, right? But I we’re missing we’re missing like we’re trying to drive a car with no wheels at this point. Like Yeah, I’m like, “Where where’s the stereo?” stereo?” You missed it. Like I There’s no engine in this thing. Like we can’t use it. So, I I Me personally, I will not use this because I having a single perspective driving only the one exploration not

7:59 driving only the one exploration not really helpful. I need to have many perspectives. again, here’s the reasoning, Tommy. Like if I think about this, every fact table has a list of dimensions that it attaches to. Mhm. Mhm. And so, you almost need, especially in large large models, you almost need a single perspective per fact table. Because when I build the report, I’m when I’m So, when I’m building a report, I’m going to have fact table one

8:29 report, I’m going to have fact table one supporting one page. Now, people are going to argue with me, people in the chat are going to be like, “No, that doesn’t work.” There are exceptions to the rules. This is not a hard fast rule. This is a general line of thinking. You build a fact table, and then most of the data from that fact table comes out on one or two pages in the report. When you build a second or third page, there might be a different fact table that you’re drilling into. There’s maybe a shared dimension, but usually you’re you’re using one or two fact tables. The only exception to this rule, I think, and I’m going to call it out,

9:00 I think, and I’m going to call it out, is the overview page. The overview page typically has a lot of dimensional time-bound things Right. Right. with many pieces or many measures from many fact tables, right? I’m showing a a summary of a lot of data in one single place. place. Well, to your point though, too, a semantic model usually a good semantic model can support multiple reports, right? And Correct. Correct. rather than even a single page. Like I To your point, I can literally think of right now a semantic model I’m working on that what you’re saying would be

9:30 on that what you’re saying would be perfect for it. Yep, exactly. I know. This is what I’m saying. saying. Yeah. Yeah. This is my point. My point is the feature is so real just by speaking this our audience is going to automatically be like, “Yeah, I could have used it here. I could use it there. I could like that That’s a good feature then. That’s a good feature because you can into existence. Yes. Yes. [laughter] The whole There’s a couple parts Tommy. This is going to I’m going to rant here for a minute as I If I hasn’t been ranting the whole time. Rant away.

10:00 Rant away. Rant [laughter] away. my rant here is my threshold for inadequate software per what I think it can do is getting so low. And what by this is any piece of software that doesn’t do exactly what I want the way I want to or be able to let me modify it in a way that works well for my process or my system I’m getting impatient with the amount of speed that development’s happening. Let me give you an example Tommy. I was in I don’t know how this is working, but I was in

10:33 working, but I was in a Grok Academy. And in Grok Academy they had this new they had this new agents Grok Academy agents area which is basically a big whiteboard. You can just have a bunch of images. You can select an image and then prompt to get new images out of it which is really cool by the way. But oh speaking of which I just made a new side note. If you want to go out and grab our new swag we have a brand new

10:54 grab our new swag we have a brand new short shirt that has just been created. I created one last night. I was just sending this to Tommy recently. I made my nice question right now. Yeah, exactly. So Tommy has the filters maketh the model. That was his phrase that we use and I thought this other one would would be appropriate. It’s called I build agents. So now that I all that I do pretty much is I build agents all the time. So I’ll put the the link for the shirt here in the chat window if you’d like to. I’ll maybe able to throw it in the description as well. But I made a really cool logo called I build agents

11:25 really cool logo called I build agents and now we have another piece of swag that will likely be showing up on the podcast here soon. I’m going to have a new shirt with a logo. I’m refreshing some of my wardrobe. So in lieu of that, there’ll be a link out here shortly for everyone to go check that out. nice sweatshirt yourself right now, conditional formatting. Right, exactly. So I think it’s a fun shirt. All right, that being said, what what the heck were we talking about Tommy? I got totally distracted. [laughter] We were talking about the inability to scroll. was ranting about something.

11:55 was ranting about something. Low threshold for for for features on reports and things. Anymore Tommy, I just want to rebuild all my software. I want to just it the the barrier to get pieces of software made anymore is just so low. And actually this is going to be a good lead into our main conversation today, which is if I have an idea in my head of how something should work, I can literally go to Figma or go to some program, stub out even a rough diagram or image of it, throw it over to an AI agent and say, “Here’s the requirements, go on the internet, research these

12:25 go on the internet, research these things, build what I want.” And it’ll just just and it just builds what I want. And and the the gap between what I expect it to build and a good item that it creates is getting better it’s getting the the gap is is getting shorter and shorter. So I can I’m getting better getting better prompts, I’m getting better at understanding how to communicate to an agent and get it what it needs so I can give it to what you want. And I know you do the same thing Tommy. Yeah, 100%. All right, I want to hear what what you got by the way on update

12:55 what what you got by the way on update because there’s a there’s a ton I want to make sure we get to. I don’t even I haven’t had had time to really discuss Good Give me another one Tommy and I’ll I’ll give you give you I’ll give you another one. This one I really like. This is the input slicer that we have now supports numeric columns. And what’s really neat about this rather than just saying a number, I actually think this is going to be really useful. You can see not only like the number one or the number two, you can actually use your go back to your arithmetic or your geomet your arithmetic and you can say two

13:26 your arithmetic and you can say two through seven or greater than 11. And that supports that feature with I think it’s awesome., if you’re doing like order IDs or you’re like, “Hey, I just want to see all the things that were over 600.” It supports all of that. I think that is awesome. That is needed. That is something that I know people have asked for in the past, so I’m very happy to see that. I think it’s a good feature, too. I’m trying to find one here that I really liked, Tommy. Okay.

13:56 Okay. of things here that stand out to me. One of them is You’re going to like this one, Tommy. Oh boy. Oh boy. faster web modeling in the semantic model updates, right? So, I I’m big on How long are you going to push this, man? man? Dude, it’s great. I just like it. I’m I’m using desktop less and less. I’m switching over to agents to do more and more. more. And I’m using the service substantially more over time. So, so again, the

14:26 more over time. So, so again, the it it’s been a It’s not like it’s been like this hard stop when I’ve been like I just stopped using desktop altogether, but it’s been like this progression. Like it’s just I’m spending more time in other places and not so much directly in desktop. desktop. Yeah. Yeah. So, anyways, that was one that that stood out to me. one that I’m actually quite happy that is actually landing here and they’re figuring things out. visual calculations and custom totals are now fully GA. Like that should just be a thing. Yes, agree with that. all those items should just be there. on

14:56 those items should just be there. on the visual themselves. I think that’s been a long-running joke for Microsoft. Like the totals of things don’t quite make sense. Can we just make it add up to whatever’s in the visual? Like it just made sense to have something very simple that was just on the visual that made sense. Mike, I got a hard question for you about this. Mhm. Mhm. Do you use visual calculations? I try and stay away from them as much as possible. only because I don’t like the idea of the visual having its logic locked into a single visual. visual. Yeah.

15:26 Yeah. My one downside of visual calculations is the the knowledge or information or DAX that you’re writing is locked to a single visual. And there’s not a really good way of porting that logic. Let’s just say I have a a report and this is the time with many tables in it. And one of those tables has the logic I need in it. Where is the I get that I I understand that that one visual has the visual calculation in it. Mhm. Mhm. Why why isn’t that a parameter

15:58 Why why isn’t that a parameter that I can reuse anywhere in the model on any report on any table that’s inside the in any of the visuals. So to me, there’s a missing feature of [snorts] [snorts] in this report here’s the summary of all the visual calculations. And would you like to add them to any other visual? Like that to me that’s a missing piece of this. Again, it’s all these like it’s all these like little cuts of like that’s a that’s a friction point. That’s a friction point. Don’t have perspectives. Don’t have like all

16:29 have perspectives. Don’t have like all these little things that I’m saying to myself myself when’s the threshold when I start rebuilding Power BI Desktop and building my own PBIX and PBIP files without without even using Desktop anymore? Like when does that happen? We’re closer to that. that. So So What? What? Hold on. Let’s not get crazy here. But let me ask you so let me rephrase. I I Listen, Guy, I know where we’re going with that and you can get away with it, but Desktop is still Desktop. And right now it still to me reigns in terms of the features available, the

17:00 in terms of the features available, the experience, it is still a better experience. Not saying you cannot do or achieve in the service what you can do in Desktop because the the similarities of parallel features are close. It doesn’t have everything. Now, what I want to focus on is what you said about the visual calculations because I agree with you. I don’t use them. it is a different language than DAX, right? So, it has its own functions, its own, and it’s not context, but, its own input the parameters.

17:30 , its own input the parameters. Let me ask you this, would you be more open or do you think you would use them more frequently if it wasn’t a different language? If it wasn’t its own visual calculation language, if it was more plug and play rather than actually having to put in a formula? Because to me I think I would be more willing. Visual calculations require another set of skills because think about it, Mike. In Power BI right now, there are six languages you need to know. Power Query,

18:00 Power Query, DAX, DAX, visual calculations, user-defined functions or user data functions, and then if you want to add in translated cool task flows. So, that’s actually five. But still, that’s requiring you to be competent in multiple languages if you want to have a full Power BI experience. Just dealing with DAX is enough alone. I would I would say but the fact that I need to know, not be an expert in visual calculations, but you got to know what you’re doing,

18:31 you’re doing, right? And that’s not a simple oh, I need to do this. I’ve done it once, so I can get it done. No, it requires the same type of experience and I don’t not a fan of that. that. I’m I’m I understand your comment, Tommy. I’m going to push back on the ability here we go. I’m going to push back on we need to know these languages anymore. We used to know need to know the languages. I think we’re very rapidly moving away from needing to know any languages in anything anymore.

19:01 languages in anything anymore. No, no, no, no, no, no, no. No, no, no. Who Who validates it, dude? Like so, I hear what you’re saying. To build it output of it. You don’t validate the code anymore. You’re not going to You’re You’re You’re You are You’re going through what you’re describing here. I see your I see your reaction. reaction. This is the exact same reaction I’ve been getting from all of my developers, all of my interns, anyone who’s I’ve been talking to, and I’m telling you this is this is a You’re in You’re in the program, Tommy. You’re in the

19:31 the program, Tommy. You’re in the program, and you’re you’re It’s fun to know the DACs. It’s fun to know the things that are in there. there. It’s It’s not just has nothing to do with that. It’s not going to be It’s At some point, you’re going to let go let go of the wheel, and you’re just going to let the AI drive. At some point, but we’re not there. Let me give you some hard cold hard facts. Issues in GitHub have gone up 70% since the introduction of GitHub agents.

20:02 the introduction of GitHub agents. Since we’ve gone to this agent number coming from? Cuz I’m going to question you on that, too. Okay, I’ll send you the news I I will send you the documentary. Yeah, I’ll send you I’ll send it to you. Happy to. Happy to. Happy to. Seven if seven like 70% fine, but if we’ve had 300% more code showing up Sure. Sure. If the ratio doesn’t fit what it was to what it is now No problem. to you. I’m happy to send it to you, my friend. But, we are seeing more and more and you’ve experienced this yourself. We are trying to do simple math, and it

20:32 are trying to do simple math, and it also hallucinates. I am not saying don’t use it. that that. But, to assume that AI is infallible, right? To use like to use to say that it makes no mistakes, even with the proper instructions I I I think you’re repeating. Not the first time. So, it’s So, who Okay. So, So, I’m going to argue a lot of these systems now are trust, but verify. It’s the same thing you It’s the same

21:02 It’s the same thing you It’s the same thing I would do with a developer. How do you verify it, then? If you don’t know the code If you don’t have an idea of the test. So, this is this is test-driven development 101. So, whether you’re doing this with people or not people, the same thing happens when you have a junior developers on projects as well. They make just as many mistakes as an AI would, and maybe even more mistakes than an AI would, and they do it much slower. So, the argument here is if you’re going to pair up a junior developer on a model and just say go build DAX, you’re going to have to have some verification or validation

21:33 some verification or validation from a senior engineer to go look at it. Okay. Okay. right. That’s not efficient. You have to know the language, though. No, you do not. There’s another model that you’re you need a senior developer on it? Cuz you’re defining the rules that you’re you’re just missing the point here, Tommy. You’re going to You’re

21:48 here, Tommy. You’re going to You’re going to You’re going to die I can see it already. You’re going to die on this this point, and you’re going to lose this one for sure because in the future, there’s not going to be any writing of any language anymore. The programming language will be English. That will be the language we’re going to do. Again, you’re talking about the Matrix. You’re talking about the future. Right now. It’s right now. It’s right now. I don’t answer me why you need a senior developer to write the test? You need someone to be able to write what’s going on and what the answer is. Okay. Okay. At some point, you have to trust something. You can’t you This is not a

22:18 something. You can’t you This is not a zero-trust system all the way through. Right. Right. a developer writes the right DAX, developers write the wrong DAX in certain scenarios. 100% Oh, yeah. How do you catch it? You You give it to other people to test it. You push it out into the real world. Something comes back wrong. Great. How do we capture what is wrong about it, and we come back and we look at it and fix it? Same thing with an agent. [laughter] How can you fix it if you don’t know if if I was writing a math test for AI, and

22:51 if I was writing a math test for AI, and I stunk at math? How would I validate that? You just ask the AI to write it for you, and you’re good. Oh, oh, okay. That’s Okay, that’s Oh, I didn’t think of that. Okay, [laughter] that’s good. So, It’s It’s Oh, how simple it was. I I look at it this way, Tommy. The AI is going to become a black box to many people. And the next generation of people coming up are not going to know how to write languages. They’re not going to know They’re not Their concepts will still exist. Like,

23:21 Their concepts will still exist. Like,, designing systems will still exist. But right now, where I’m already asking junior developers, junior people on my team, team, they’re needing to act like an architect day one out of the box and not actually go in and write the code directly. Right. Right. What What’s changing though is how we verify verify What we What was changing now is that black box is now Okay, inside there is some code. And now what you have to do is you’re learning a skill of how to prompt the AI to go look at the code it wrote,

23:52 to go look at the code it wrote, Mhm. Mhm. figure out what it did, and get to you an answer so you can understand what it’s what it’s doing. So, Right. Right. you have to use the AI to go peer into its own code and describe and draw an architect back with you so that you’re actually able to get to the right answers so you can you can get into is this in fact correct. Now, Now, specific things like DAX, I think you’re going to be a bit more tricky to do some of this stuff, Tommy, cuz you you have to have other ways of validating what this looks like. So, if AI For example,

24:24 this looks like. So, if AI For example, if I’m going from bronze to to gold, right? right? You’re going to like work your way through and make sure you understand all the logic and make Make sure things are proved and right. Especially when you get to the model and the model’s writing new DAX measures. And it may not understand simple star schema modeling and all the techniques that we’ve got that we’ve built for it. Like that we understand as as developers now today. But those skills are going to get programmed into skill those knowledge is going to get programmed into skills and the agents will get better at doing this. And so, all we’re doing is every

24:54 this. And so, all we’re doing is every time we work with an agent, we’re giving it better and better guard rails to get more regular outputs that are exactly what we want. And what’s going to happen is that’s going to it’s still going to make mistakes, but it’s going to make a less of them. And when it does, we’ll be able to to articulate questions, prompts, engineered prompts back to it to help us understand what really went wrong. So, that’s that’s the the new skill is we’re going to let our hands off. We’re going to let it just go. It’s going to continue to build these black boxes, but we’re going to have better ways of

25:24 we’re going to have better ways of understanding and asking the right questions to interrogate that black box and get the answers out that we need. So, it it’s a double-edged sword here, Tommy. I’m not just saying hands off, let run. It’s there’s a new skill being developed here for people Right. Right. where you one trust that it’s doing the right thing, and then two, you’re giving one better prompts up front, and you’re walking it through more of the process with you to make sure that the black box thing is built the way you expect it to be created. And and I I don’t see anything wrong

25:54 And and I I don’t see anything wrong with that. My argument here is just like when I am getting better at application development prompting wise, the context providing what I want, the instructions, the framework of what I’m looking for. However, However, I’m not looking at the rust or the HTML or the JavaScript, and I’m just assuming that it did it the right,, in terms of best practices or as fast as it should. And that’s for,, more or less simpler applications. I can’t validate anything in it. I have no context into that. And I’m going to

26:25 no context into that. And I’m going to I’m I’m going to offer you a challenge for the month of June. Mike, when we get into June into June Let me just Let me Before you give me the challenge, I just want to tease your thought here one more one more step here with what you’re what you’re saying. And we need to get on the main topic here. Yeah, yeah, yeah, yeah. We’re doing a good talk on this one anyways. anyways. Tommy, Tommy, the challenge you described to me of the code being a black box, Mhm. Mhm. it feels very analogous to when a customer gives me a very large model for the first time.

26:57 model for the first time. It’s technically a black box to you, Tommy, Tommy, until until you go spend some time doing something to interrogate that black box to understand what’s going on. So, what do I do now when some client gives me a black box these days? The first thing I do is I go look at like, “Okay, what are the relationships with the tables? Does anything there look wrong? What are What’s in the tables? What’s What’s the column names, right? Are we?” And then also with this, I’ll I’ll even go to my agent

27:28 with this, I’ll I’ll even go to my agent now, so I’ll use the fabric modeling MCP server and I’ll say, “Write up simple documentation for this model. model. Mhm. Mhm. What tables are coming in? What are the data sources for each table? Give me the lineage of all the measures. Just Just do a general like scan of the model, okay? okay? I’m not going in and I’m not going to look at every single DAX statement. I’m going to let the model I’m going to use the agent model to describe the black box to me so I can reason about what’s inside it.

27:59 so I can reason about what’s inside it. And so, And so, that’s not perfect what we’re doing here, but it’s a similarity of like a symmetry of what I think will be happening. [snorts] [snorts] Okay. I I know we need to get on the topic. There’s so many things. There are [laughter] Let Let me just say this and then and we’ll move on. And we’ll Let’s parking lot this because I think this deserves a conversation. This should be an episode. Yeah. Yeah. So, I’m going to remember this. I’m not

28:29 I’m going to remember this. I’m not going I’m not going to make my comment because we’re just going to go off a tangent again. My challenge for you though, for June, when we talk about agentic features and AI solutions, I’m going to challenge you for the month of June to not talk about what it might be like in the future, but talk about what the current situation is today. I don’t talk about that. I haven’t been. Everything I’m describing to you is what I’m doing now. I’m doing this today, Tommy. This is why I’m talking about it. I’m talking about it because this is happening right now for me. for me. I

28:59 I You’re talking You mentioned many times that the fact that Oh,, in the future we’re not going to even have to worry about doing it this. This will be occurring. I get a lot what you’re saying you’re applying, but it’s not all of it. So, it’s a theme. It’s a theme. I You are the most visionary person I know. And that’s not just throwing you flowers here, okay? That’s not pandering. Dude, I’m seeing it. I see it already. Like Like Exactly. You’re seeing the future. But I I see it because I’m doing it in small doses right now. Like it’s The

29:30 small doses right now. Like it’s The full thing is not It’s not fully baked yet. yet. Okay. Okay. I’m already doing portions of this right now. Like now. Like I do not contend that. I’m building full websites on GitHub pages without looking at a line of code. Phenomenal. I’m I’m building full I’m building full blog posts by describing what I want and thinking through the ideas with an agent and having the agent write the entire post and publish it for me. Like so, already I have areas of this experience that are being really big baked out and built today in very small doses. It’s

30:00 built today in very small doses. It’s like, okay, website building, static website building. All that, I don’t even look at code. Done. I don’t have all I’m doing is I’m projecting that experience onto what I think will happen inside data engineering and the rest of Fabric. And And Kurt’s article, this is actually a great lead-in. Today is exactly what this is. The this This experience So, what Kurt’s going to So, let’s go to Let’s Let’s actually transition. This is a good transition because this is where we’re at in the point here. Kurt is describing And let me Let me

30:30 Kurt is describing And let me Let me present another analogy here, Tommy, cuz I think this makes sense. It just kind I think this makes sense. It just came to me. The fact that we have a semantic model in Power BI Desktop already we’re The The act The very act of using Power BI Desktop is already an abstraction from the actual code you need to build to make visuals. Yeah. Yeah. I don’t actually write d3. js to make the visual go. [snorts] [snorts] It’s an abstraction. It’s not an AI abstraction, but it is an abstraction away from the actual code that we’re

31:01 away from the actual code that we’re building. I grab a visual, I put it on the page, I drop my fields, boom, it’s done. I didn’t look at a single line of code in that whole experience. Now, Yeah. Yeah. I would even argue making the visual and dropping the fields on the visual are even like too much work anymore. Like I don’t even want to do that. [laughter] I want someone else to automate even more of this and go even faster, but we’re already looking at Desktop as being a substantial abstraction away from the actual code. Right. Right. Kurt is now saying the same thing in his article. He’s saying, “Look, the 5

31:32 article. He’s saying, “Look, the 5 minutes to wow in Power BI Desktop when I originally came out, he’s having the exact same experience now with Claude code and building like websites that use data data to build the website. And instead of the instead of him doing the whole, write d3. js directly to make the website, he just describes what he wants. wants. And out and out pops from the black box something he can look at and vet. Did you ever see The Godfather?

32:04 Did you ever see The Godfather? I have seen The Godfather. So, if you remember when they’re in Sicily Sicily and he sees the person I I believe he married No, no, no, it was actually anyways, but he saw a girl and and the town the Sicilian went mm, lightning. And he’s like, “What?” He’s like, “He’s got to He got hit by the lightning bolt.” Because he saw this beautiful,, girl and they got married and and I think we’re seeing the same thing here where we’re having these moments where we’re getting it feels like being

32:34 where we’re getting it feels like being struck by lightning because you cannot believe believe how well something can be produced or the ability to produce

32:42 the ability to produce what was just a vision. Yep. Yep. and and this is is more than just a bridge was being built. It’s there is always this other island, this other place that we unless you spend years of practice, you were not going to be able to build just in years, years of experience and learning JavaScript, HTML, and then actually applying it that to your point. I I can literally build an app while I’m playing catch with my son, get a cup of coffee, like, “Oh, good, it’s done.” And

33:12 coffee, like, “Oh, good, it’s done.” And then I can just immediately start doing it. For example, I built an app called Skill Vault, which I can download skills from anywhere, and then it will push to all the the folders in my directory, say, “Okay, send this to Claude, send this to Copilot, send this to Codex.” Yep. Yep. And those skills get synced. And then it’s or to get up. It’s awesome. And it’s such a cool idea, but it’s a great app in terms of being able to edit it or modify,

33:42 it or modify, but it was something that I wanted. You but it was something that I wanted., it’s something that I’m like, know, it’s something that I’m like, “This would be useful.” There’s no way in heck that I would have been able to do this do this without it. And we’re seeing this too in the Power BI world. Yep. Yep. Mike, I am feeding from a transcript things in Notion, which is already a cool tool to begin with, what the milestones are from an SOW that I uploaded, uploaded, that aligns with the conversations with we had, that writes Claude agent instructions for an MCP, which it

34:13 instructions for an MCP, which it understands. I let it run, and it takes 25 tables and cleans them, and helps build the measures, and like, ” what? Build a bunch of conditional formatting measures, too.” And automatically these things are done to our earlier point. And I’m not looking at the code. I’m just having this conversation with it. Yesterday, Mike, I built two SOWs with both computers talking to two different agents going, “Okay, yeah, that looks good. No, we should probably need to change that. Oh, by the way.” And it was

34:44 change that. Oh, by the way.” And it was insane to this ability here. So, these are these constant lightning bolts on the first time we saw Power BI. Yep. I I think this this whole concept, Tommy, of Tommy, of agentically coding solutions now, like this just this whole like it’s it’s reduced the technicality of stuff to almost nothing. And it’s increased the level of automation to a high degree. I can automate a whole lot more and how do I how do I get more data around? How do I get data in? What what

35:15 around? How do I get data in? What what things am I regularly building? So, a lot of my time now is really kind So, a lot of my time now is really rethinking, reinventing of rethinking, reinventing my workday to try and think about, “Okay, there’s a lot of things in my workday that I do that are not fun.” One of the things that you gave me, Tommy, that I’ve been working heavily on with this automation piece is just my general workflow around statement of works. Yeah. Yeah. Right? I have a contact person. I have an a website. I have a, an address. address. I can actually do a good amount of like just research on front in front of

35:45 like just research on front in front of this thing and have it prep a lot of things. And, Dynamics 365 was telling you that you should be able to get this feature. Like I don’t need to pay $85 a month for a license for this other Dynamics thing where I have all the information I need right at my fingertips. Mhm. Mhm. So, now I’m I’m really rethinking like where does my data live? How do I actually actually I I I think maybe part of this too, Tommy, is Tommy, is by us being able to vibe code things like the way Kurt is describing here, we’re actually able to truly retain our data. data. We don’t have to give it to a

36:15 We don’t have to give it to a third-party product or program or something else that is just difficult to get it back out. One of the things that’s been really well thought of when these current softwares as a solution programs, programs, they’re really good to get the data into them, them, but they really don’t care to let you get the data out. Right. Right. And so a lot of my friction is around whatever application you’re picking, is it easy to get the data out? And typically, the getting data out costs you more money. So then I’m like, well, I don’t want to spend more money to get my own data that I’m giving and using in your product.

36:46 I’m giving and using in your product. I’m just going to give screenshots of your product to my agent and say rebuild it. Now I have my own version. I don’t need you. And then now the data’s stored in my fabric, in my database, where I can go build whatever else I want on it. Let me ask you, why is this so important rather than because we’ve focused a lot on the ability to build things and the ability to accomplish things, right? But But you if you really looked at the whole picture here, when we are building more things or

37:18 when we are building more things or building things on behalf of what we’d normally do, by definition, that means there’s a lot of time, there’s an absence of time. Or there’s there’s a void there in terms of what I would normally do. If I would spend 5 hours on building a report, then now it takes 30 minutes, Mhm. Mhm. then what am I doing with that other 4 and 1/2 hours? And I want to have this conversation with you because we’ve talked, again, in spades about what you can do and how cool this is. But I think

37:49 can do and how cool this is. But I think to me, the most impactful part, and I think the lightning bolt for me, is not so much that I can build it, but what I’m allowed to do in those other 4 and 1/2 hours. And I want you to take this in the realm of as a consultant, as a leader, as,, as a role in a company, why is this so important? Why is those extra 4 and 1/2 hours, the time that you’ve gotten back? Actually, let me I’ll rephrase this. Are you Should we be building more things in

38:20 you Should we be building more things in in in extra time or at,, the time we got back, or is there something else that we should be spending that with? Do Am I clear? Yeah, I think so. Okay. Okay. There There is I don’t know how to I don’t know how to describe it yet, Tommy. Mhm. The analogy that I think of in my mind is is there’s a video game that you’re running. running. Okay. Okay. And the video game rewards you with little nuggets of success. Right? I I go

38:52 little nuggets of success. Right? I I go into something, I go build something. the game,, I’m I’m So, I just I just rediscovered, Tommy. I don’t know if you ever played the game of Warcraft. Did you ever play Warcraft back in the day? StarCraft. I was a StarCraft Okay, you were StarCraft fan, okay. Same same concept, right? I love these games because it was all about like resource management and building things and,, trying to like allocate resources. Someone on one of the social media platforms, I don’t remember where I saw it, was saying something that the people at

39:23 was saying something that the people at the top of the of the of the economic pile, they’re the individuals who’ve understood how to allocate resources better than everyone else. I was like, “Interesting. Neat concept. Warcraft is an allocation of resources game. game. I like thinking about where do you spend your time? So, everything’s a resource. Everyone has the same amount of time, Tommy. Tommy. But yet some people earn millions of dollars a year and some people don’t earn millions of dollars. So, what What’s the difference? Is

39:53 What’s the difference? Is Is it just like I don’t know how to I don’t know how to It’s a very complicated issue. But some of the people that are at that are doing really, really well are just allocating their resources differently than other people. This is a lot of times you’ll hear this with like,, getting out of debt things. Like how do you allocate your resources? What time are you spending? How are you using your efforts to get out of debt in these certain situations? Like Like there’s always needs for things. There’s There’s someone always willing to pay you for some service or something like that. So, how are you honing your skills? This is like the Alex Hormozi thing, right? Work so hard that everyone

40:24 thing, right? Work so hard that everyone else can’t can’t ignore you anymore, right? Just overwork everything. Get be better not just like two or three times better, but like 10 times, 100 times better. If everyone’s putting out YouTube videos or certain level things, you’re going to put out 100 times that field. Like we’re doing it, Tommy. We’re putting out 100 piece No, sorry. We’re putting out about 100 pieces of content a day. I don’t see any other MVPs in in the community putting out 100 pieces of content across social media platforms per day. That means over the course of a

40:55 per day. That means over the course of a week we’re doing 700 pieces of content a week. Microsoft, are you listening? MICROSOFT, SPONSOR US. [laughter] BUY THE PODCAST. NO, DON’T. SO, WHEN I LOOK at this, I’m going like we’re doing so much volume that people have to take notice. So, Okay. Okay. then What this does though Yeah, because I’m not seeing the the go back to let me let me bring it back home. All of these little nuggets and things that were I’m I’m pointing on is

41:26 things that were I’m I’m pointing on is these are now reward successes in my brain. brain. And so, I thrive on these quick wins,, I have an idea. I’m going to build this Warcraft base. I’m going to add this thing. Oh, no, it’s attacking me. Okay, I got to get this other thing built up. And there’s these little quick hits of like success and rewards and the game provides to me back those successes. And it encourages me and it it energizes me to go further in the game and do the next thing. The same thing, Tommy, is happening with this

41:56 thing, Tommy, is happening with this vibe coding. It’s the same experience. It’s like I’m playing a video game and allocating my resources. So, instead of to your point, the message has been use AI, it will help you make your life more efficient. Yes, it will. But what I have found is all that extra time and now it’s rewarding me more frequently with my ideas and my thoughts. Hey, I had this cool I thought I have this software idea. Let me see if I can noodle it out. Okay, hour and a half later, boom, app’s done. It

42:26 half later, boom, app’s done. It works. Let’s deploy it. 30 minutes later, boom, deployed. So, the time to reward for me with AI has greatly shrunk. shrunk. So, now So, now I look at this and going, well, I want to keep playing the game more. I’m going to just keep playing. So, now I play all day. I hang out with my family in the evening. And then when everyone else goes to bed, I’m back on my phone. I’m I’m programming more. Like I’m actually doing more work than I I’m allocating more of my time and resources to education and learning and doing more of

42:56 education and learning and doing more of this because it’s so rewarding with these faster cyclical cycles around development. My ideas can go from idea to reality much faster. Now, let’s pull this back this back to specifically the Data Goblin’s article cuz I want to This is actually what Kurt talks about in the very beginning here. In the in the top portion of the article, he says, “I’ve always been impressed with like charts and visuals.” And he was always,, intrigued and he was in the Kurt’s a genius, by the way, just so. He’s got a PhD. He’s been in the medical

43:27 He’s got a PhD. He’s been in the medical field for a number of years and then he’s gone into this data science and data engineering realm, which is awesome. I absolutely Everything I Every time I talk to him, I always learn multiple things from him.

43:36 I always learn multiple things from him. Just a great voice to follow. But, that being said, he said, “I used to make visuals with R, Python, or JavaScript.” I did the same thing. I was learning those languages. I was using GGplot and trying to build them like, “Man, this is really hard. I can I can barely stitch things together, right? Okay.” RStudio, baby. RStudio, I remember using RStudio a lot. Yeah, 100% Tommy. So, what those are those were abstractions on top of like D3 or graphics visualization things. It just made it a little bit easier. But

44:06 just made it a little bit easier. But they were still heavily code. Then Kurt put points out, hey then someone gave him Tableau. Oh, let’s check that out. Oh wow, the time for me to build a bar chart or a line chart or visuals has greatly dropped because the technology is now click and drag and just have things work. So, he’s then says, wow this this made my brain warm with new ideas. I agree. I had the same experience. It wasn’t Tableau for me. I did a little Tableau. But I did Power BI. I jumped in and I was like Oh yeah. Oh yeah, this is this is gold. So, you

44:39 Oh yeah, this is this is gold. So, you Oh yeah, this is this is gold. So,, moving forward this is just know, moving forward this is just another evolution on that step which was, okay was, okay R Python scripts,, writing code to build visuals, then moving into like graphical editors, initially Tableau cuz it was earlier to the market, then eventually Power BI. And here we are on the third stage of this which is, okay we don’t really care what any of the language is. And now the medium by which I communicate through to get visuals built is now talking to an agent.

45:09 I have So, that’s that’s my progression I’m going through. I I understand that I do have some concerns with what you’re saying because I have a I have a person in my head. And they’re my boss. And they’re very hard on me. [laughter] Sure. Sure. And I always think about the time I’m spending and if what value am I providing. And it’s my it’s my driving force. Like a client can never be hard on me because I’m harder on myself., but I I I look at this and what I’m hearing from you or

45:42 at this and what I’m hearing from you or parts of what I’m hearing from you, I’m not saying this is your blanket statement is you have the ability to create more. Like the volume increases, right? Which is totally fair. And I I really consider this lightning effect and what it’s been for me as well. well. And I I consider two avenues. Should we be building five x of what we’ve been doing? Like if I would build one report, I can now build five reports. Is that

46:12 I can now build five reports. Is that the best value of that extra four and a half hours? Mhm. Mhm. On the other side of the coin, yeah. Not 100%. I’m saying my initial reaction No, yeah, yeah. That’s why I said it was not it’s not a blanket statement. Well, like if I was building one report before, so talking to your your boss about the value add, right? Okay, well, I could build one report. Why So, yes, maybe I should build one report, but but maybe I should have spend like it should take me less time to build that one. What would I

46:42 to build that one. What would I sacrifice that other time to do, right? I think that would be spending more time with the business, asking more questions around specific details,, , the the time Money favors speed. Is this Is it statement I’ll just make. Money favors speed. Mhm. Mhm. The people that can move into the market faster, faster, people who can make reports quicker, gets more cycles on feedback and

47:12 gets more cycles on feedback and refinement. So, if we have a topic if we If on Monday, Tommy, you and I can talk and say, “Hey, I got this requirement. I need this this data da da da da.” If I can then feed those requirements back into the AI and immediately have that turned around by the afternoon, literally the same day, but like, “Okay, thanks for your input. I’ve done some work. Here’s Here’s a prototype, right?” If you can give that to them by the afternoon, then the bottleneck’s not you, it’s now them, and they can go react to it. And while the idea is still fresh in their mind about what they think they’re going to be getting, they can say, “Okay, yeah, 90% of this works.

47:43 can say, “Okay, yeah, 90% of this works. I have a couple more issues. There’s a couple more refinement problems here with the report or something.” And then they give you that feedback, and then you cycle it again. So, in the time it took you to build two or three cycles before, your old self would have built one cycle. That’s a value add. So like so you’re not maybe building three reports, but you’re saying in the span of the time of building one, you can now do three iterations on the same report. I think that’s a really good value add. I think that’s a much better way to say what you’re saying. And I’m I’m going to emphasize this because that wasn’t what I was initially going to say, but I

48:13 I was initially going to say, but I think you’re really touching on something. You’re touching a good nerve here. here. I currently have I have a client right now where we’re meeting twice a week and we’re doing the sandbox phase, which is really aggressive when you think about it from building soup to nuts. And the client’s great, but they have they have a lot of good ideas. There’s so many things on the twice that we we the we we the twice that we meet in a week where they go, “Oh, what would be great? I actually need to see this as X, Y, and Z.” Z.” Normally in your head you’d go,

48:43 Normally in your head you’d go, like I guess I’m staying till p. m. tonight if I’m going to get that done. Or I tell them I can do that in next week. week. Most of the time, Mike, I’m thinking about what the person says. I’m like, I can I’ll have that tomorrow. And it’s because I have the MCP server to your point with that speed. So I’m allowed now to not just speed up a project for the sake of it, but to your point, people get antsy with especially that sandbox phase or that building phase. So I can meet on Mondays and Thursdays or whatever it is. I can say, “Yeah, you want all these

49:14 I can say, “Yeah, you want all these changes that you haven’t talked about till today? Well, I’m already try This is a notion da da da goes through the MCP and I can get that like I’ll have a screenshot ready for you by Wednesday.” And so I’m not working like a dog every day. So that’s I think a huge part of this is we can attack the business and this almost goes back to the beginning of what you mentioned about our our intro about the expert in the room. the room. Yep. Because this is really important.

49:45 Yep. Because this is really important. And for me, the calling card, the real aha, the real lightning, the threshold for me is actually the probably the best way to say it. The threshold for me with AI is it allows me as a human to attack the business, the problems head on. head on. And I’m not spending 80% of my time doing the development or doing the work or doing the tediousness of let me format a measure by a currency. Let me build,, all these measures. I

50:15 build,, all these measures. I can now attack multiple parts of the business business which is the value for me as a human. And I think this is really important for people listening. Where does the value come in? What my head says to myself, what I tell myself is, where am I providing the value that someone else can’t? can’t? Mhm. Mhm. Right? Right? And for me it goes to this ability that I have my assistants, but me as the leader leader me as the expert can go, oh, what are the other problems that we have? Let’s really prioritize this and it doesn’t it

50:46 really prioritize this and it doesn’t it to your point it’s not going to be a month from now that be able to attack these things. Mhm. Mhm. You mentioned, and this is the last thing I’ll say here you mentioned when you were debugging a black box in a semantic model Mhm. Mhm. that you where to go, right? what to look for, the problems that the common causes, which is really only someone with experience can do. In the same fashion when I go to a client or I go to a different department I’m not focused on whether or not I can

51:17 I’m not focused on whether or not I can build that in a time frame, I’m focusing on their priorities. And I know what to tackle, I know what to ask. to ask. More importantly I know what my Agentic Solutions can do. So if someone says, we have pipeline issues or,, it’d be really great to have all these different measures measures there’s no longer a backlog or a bottleneck to say, we don’t have the resources for that. So I’m let’s get that done. What are the more pressing issues? issues? That to me is the real critical junction

51:50 That to me is the real critical junction of where we’re at here. Mhm. Mhm. Is really and and and Kurt mentions this too is the soft skills are that ability to understand empathy something we’ve talked about have never been more critical in this era that we’re in. we’re in. 100% Tommy. And I and I’m This People are worried about the agents taking away our jobs. I don’t think it is. It’s just drastically shifting what we work on and where we spend our time.

52:21 we work on and where we spend our time. To back to my point earlier about resource allocation. Yeah, yeah, yeah, yeah. All we’re doing is shifting and so so my my effort my effort now is mainly focused more on how do I automate more? How do I integrate agents more? What are the like I’m focusing more on building solutions to big problems. And I think now with agents I’m able to attack attack bigger problems than I was before. So, the solutions are getting grander. The amount of code and lines I’m writing is getting larger. It’s It’s expanding

52:53 is getting larger. It’s It’s expanding the capabilities that I have. And it’s only being done by the addition of agents on top of what I already know how to do. to do. Right. Right. I’m looking at this going like oh yeah this this this is what I want to be working on. Like this makes sense to me. Like I like this part of the the experience. And also not only am I able to tackle bigger problems. problems. The cycle time to get a win in those bigger projects are now getting shorter. Right? I can rebuild

53:24 Right? I can rebuild Right. Right. things. The infrastructure on the back end can be ripped out and replaced in months, not years. companies that were looking to do big migrations from one system to another probably can go faster now. Yeah. Yeah. There’s an elephant in the room though and I’m I’m ask you the hard question here. That that I’ll ask myself here, but but when you think about how much we have available to us in our new kind we have available to us in our new era, of era, how does one, a listener or yourself, answer the question, “Why should I hire

53:55 answer the question, “Why should I hire you then? What value add Why What can justify the normal pay that you would get get as a developer if 80% of what you can do, you’re going to automate, right? Because to your point, if you don’t need the same skill set. And I think this is important because I I agree it’s not going to going to, Bill Gates just came out and said there’s only going to be five types of jobs in the future because of AI. We don’t know. Nobody know Nobody knows the future. Don’t listen on Twitter on people saying, “Oh, it’s going to replace all your jobs.” Or no, you’re

54:25 replace all your jobs.” Or no, you’re going to be better at your job. No, we have no idea right now. We’re in 1999 of

54:31 have no idea right now. We’re in 1999 of the internet era. So,, did Did we Did we ever think there would be like a We had marketing engines agencies before the internet. internet. Did we ever think we’re going to have like Google Analytics and all the things that come from like that? Like a whole new Like that’s a huge brand new industry that just showed up with like digital marketing. Didn’t exist before that. that. So, in the data world, for those if if you were to tell people if someone were to ask, “Why should I hire you?” in your normal role, what’s what’s the response?

55:02 No. Yeah. Yeah. I make you AI first. Your company doesn’t know how to AI. I’m way ahead of you. You hire me to teach and educate your team and build samples and prototypes. Like Like Yeah, no, that’s so good. Yeah. The story here is I’ll bring I will bring you to the future. future. I’m the future showing up to your doorstep showing you how to use agents. I’m doing things that you’re not doing. I’m going faster than you can work. The pitch here is That’s so good. Yeah. The pitch here is I’m the I’m the

55:32 The pitch here is I’m the I’m the accelerator. This is the And this is the same pitch, Tommy. This pitch has not changed for me. me. I am the accelerator. I am the accelerator. That [laughter] was like the Terminator. Get to the chopper. That’s a t-shirt. That’s a t-shirt. Yeah. I’ll be back, accelerator. [laughter] So, but that that’s the pitch, and the pitch was always that, right? I’m I’m the fuel to the fire. I’m I’m I’m the one who’s done this stuff before. I’m helping you dodge the potholes. I’m

56:02 helping you dodge the potholes. I’m pit the, the the the the the issues you would make. You can learn all this on your own, for sure. You don’t need me. But, if you want to get to the end result result with the Power BI. If you want to get same thing with the same Power BI. If you want to get to the end result faster, that’s why you hire Tommy and I, cuz we’re the ones that are going to be able to bring AI into your systems, tell you where is effective, tell you where it’s not effective, and where you’re wasting your time. Already, I’m having consultings with like companies that are like, “Well, we think we can do this, this, and this.” I’m like, “Well, what you’re describing is really like a

56:33 what you’re describing is really like a chat window. Like, I need I need you to think about the the conversation. the the creator agent, right? Creator agents is where it’s at. Let’s think about what can the agent do to help your team create. And they’re like, “What?” I’m like, “This is why I’m here. I’m here to give you the surface area of like the vision of where this stuff is going, and then where do we apply it in your business? Mhm. Mhm. And so, what is easy to implement and get going, what is difficult to implement and get going? That’s that’s why you’re hiring us. Like, I just has not changed.

57:03 has not changed. The same It was Fabric, it was Power BI first, then it was Fabric, and now it’s just agentic experiences on top of Power BI and Fabric, and now websites and apps. All of this is like why you Yeah. Yeah. talk to us, cuz we know what we’re doing. doing. Just had a conversation yesterday with a client, they’re like, “Hey, we had someone come in and said that if we optimize, we’re just going to run our LLM over our data,, but we need to optimize LLM. Is that what you recommend?” I’m like, “First off,” I was like I’m like, what are you trying to

57:34 like I’m like, what are you trying to do? What is it trying to do? Just because you have a model Like,, running over data first off, consultants are saying this and they’re stupid, okay? That’s ridiculous. That’s just trying to make money. I’m sorry. Where where is your semantic modeling? How have you prepped your your data for what the agent can even understand? you trying to actually accomplish? Yeah, what’s the end goal here? You just want to ask it So, is is the expectation and this is what I hear a lot. Is the expectation I want to talk to something and say,

58:04 I want to talk to something and say, what were my sales last week? That’s a very You just made the bar so low. Right, right, right. I And the that’s a miss, in my opinion. If you If you’re asking your agent what was what was this thing, this defined,, query that I’m going to run all over over and again. That’s not a good place to use agents. What is a good place to use agents is describe the type of questions you want to answer to the agent and then say,

58:34 to answer to the agent and then say, agent, build me a dashboard that answers those kinds of questions. Then you get a very efficient Power BI report, semantic model, SQL database, you get something that is deterministic and can give you the answers and we can test it. So, back to your point earlier, Tommy, right? I don’t want the agent giving me answers about my data back. That’s not where I think it’s the best applied. 100%. 100%. Because I can’t test that. It’s very difficult to test those things. What I rather have it do is, hey agent,

59:06 What I rather have it do is, hey agent, look at my model, I’m trying to calculate this DAX statement, write me the DAX three different ways. Let me test it and then here’s the output and then I get the code and I can go run it. Or Or how does this measure relate to the rest of the data model? What columns can I use with it? Ask questions about the model and get details back about how your data works. Like that’s where you should be using agents. I think if you’re FTE right now or or even in our shoes, too, you want to position yourself and I don’t want to

59:37 position yourself and I don’t want to use left behind because nice nice very nice. nice. Oh, there you go. thing. Because I don’t want to sound like those people on Twitter who I hate very very much who said like I made $15, 000 today by building 18 agents. So, I’m going to try to refrain from that or I’m definitely refraining from that. But if you have a feeling of not wanting to be left behind or like oh, I I I I don’t know which way to go. to go. I would recommend if you’re in the space all these prep to data for AI futures and all the data features the AI features that are available in Fabric right now,

60:07 right now, position yourself as and again focusing on what they’re trying to accomplish, not just trying to run AI on your data. Yeah. Because there are so many ways you can do that, but it’s going to work based on what you’re trying to do. I I talked to someone the other day and they’re like, “Hey, they said it’s going to take a year for us to clean our data for AI.” I said, “Sounds like you’re trying to do everything and boil the ocean.” Let’s just start with the wins right now. Right. Like what are you actually trying to do? And I think this goes back to being strategic. Right. Like the

60:37 being strategic. Right. Like the greatest skill we can have is that empathy. Is okay, well, what are the problems right now? And I think for us in our space, again, whether you’re a data analyst, director of BI, consultant, it doesn’t matter. Yep. Yep. Our questioning positioning here is what are you trying to accomplish with AI? What data are you trying to look at? Let’s start there because I think I have a I have a hot take. I have a hot take. A real hot take. I think in 2 years all the things we’re doing with data right now are going to be gone. I I think it’s going to

61:07 be gone. I I think it’s going to completely change. I think the framework with how it is in Fabric to data agents is going to be completely different. Does that mean we do nothing? No. But we are such in the early stages here like of what the technology can it can accomplish. Yes, we are. And no one can say they’re an expert on Stop talking about the future. Just talk about what you’re DOING RIGHT NOW. [laughter] SEE WHAT I DID THERE?, THAT WAS ACTUALLY EXCELLENT EXCELLENT. HEY, guess what?, it’s a win for me because you’re acknowledging it. So,

61:38 acknowledging it. So, [laughter] I thought my wife was stubborn. No, okay, fair. That is very fair. In the here and now, Yes. Yes. on May 21st, 2026, what you can do is focus on what people are trying to accomplish and guess what? Dirty little secret, that’s the same thing we’ve been doing all this time anyways. anyways. Mhm. Mhm. So, So, I love it. Absolutely amazing. Always do, baby.

62:09 Always do, baby. [laughter] Yes. Yes, that is funny, Tommy. I love it. Absolutely amazing., that’s incredible. All right, so let’s let’s jump in. We’re going to go right here and we’re going to wrap for today. today. This has been another great episode. Again, we’re trying to keep it close to as you can to an hour at max., , Yes, exactly. It was it’s This is a great conversation. Really like this one. A lot of things going on. This This article from Kurt is absolutely amazing. Definitely worth a listen and a read through. Think through what he’s doing.

62:39 through. Think through what he’s doing. He’s producing some really good content right now these days. So, really really excited about that., and I’m going to put that in the chat window as well. So, this is Kurt’s article Data Goblin’s Power BI 5 Minutes to Wow is the article we were reviewing here., awesome article. Really good things here and Kurt goes through his experience, how he started from code and now he’s moving into Power BI and now his new move, which is now going to be moving into agents. I think this is huge. I think this is going to be a massive win for developers., and this is the stuff that we should be focusing on.,

63:09 we should be focusing on., if we stand still and ignore agents and just don’t do anything, the agents will pass us by and your competitors, the other people in your space will learn how to use it. And they will outwork you and they will outproduce you. So, we definitely need to get that figured out and start learning this today because that way we are part of the leadership moving forward helping companies integrate AI, build AI in your

63:39 companies integrate AI, build AI in your company, building the AI solutions. This will not will not in the same way this is the same moment Tommy I’m having when I had to learn desktop. desktop. Power [snorts] BI Desktop When I started learning Power BI Desktop I thought this is going to change the world. world. And and this this is probably 2x, maybe 3x more impactful than when I found Power BI Desktop, I think. think. Mhm. I I I completely agree. We are we are such at that point like I don’t know what this is going to be, but this is I

64:09 what this is going to be, but this is I I this is awesome. Awesome. But that being said, Tommy, where else can you find the podcast? You can find us on Apple, Spotify, wherever you get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. Do you have any questions or any other topics that you want us to talk about in the future episode? Head over to powerbi. tips/podcast, leave your name, and a great question. That being said, Tommy, thank you so much. please consider subscribing to our channel. you can subscribe for free. If you want to become a member and get some additional benefits, you’re more than welcome to. There’s a QR code

64:41 more than welcome to. There’s a QR code right over I can’t really get my angles right. There. There. That you can go get it and then you can also subscribe down below. We’d really consider subscribing. It really help us out a ton. yeah, we’re actually figuring out how to to tool and have technology. So, how about that? awesome. Thank you so much. We appreciate you our community. This has been a lot of fun. We really enjoy this. And again, like Tommy always says, always blue. Always blue. Always blue. ALWAYS BLUE. IT’S LISTENING EARS, just pump it [music] up, be high. Tommy and Mike lighting up the sky. Dance to the day,

65:12 lighting up the sky. Dance to the day, the laughs in the mix. Fabricate a lie, I get your fix. [music] Excessive measures, drop the beat now. [ __ ] this kids, feel the crowd. Excessive 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.