What's Next? Building a 3-Year Power BI Roadmap - Ep. 510

If your team is staring at Power BI, Fabric, Copilot, and a dozen AI announcements wondering what actually belongs on the roadmap, this episode is the reset. Mike and Tommy step back from feature-chasing and talk through how to think in horizons: what to lock in now, what to pilot next, and what bets are worth making over the next three years.

News & Announcements

• The message “AI will replace you” is irresponsible and unhelpful; I’m sick of it. — Kurt Buhler argues that AI messaging has too often dismissed the people who built careers around Power BI, Fabric, and software development. His point is simple and important: AI should be framed as augmentation and enablement, not as a threat narrative that alienates the very practitioners expected to adopt it.

• From Application Silos to a Natural Language-Driven Data Ecosystem — This paper proposes an “AgentOS” model where natural language becomes the primary interface and software is composed through agents, skills, and intent-driven orchestration. For data teams, it sketches a future where finding, shaping, and acting on information becomes less about hopping between apps and more about coordinating intelligent systems over a shared knowledge layer.

• ExtractLabel: Schema-driven unstructured data extraction with Fabric AI Functions — Microsoft walks through how Fabric AI Functions can turn free text into reliable structured outputs by enforcing a schema with ExtractLabel. The practical value for Power BI and Fabric teams is huge: cleaner extraction from tickets, claims, notes, and documents without having to build a bespoke NLP pipeline every time.

• It’s-a Me, Agentic AI — Lee Han Chung uses Mario as the metaphor for explaining agentic AI systems, from the base model to harnesses, tools, environments, and reinforcement learning. It is a genuinely useful framing for technical teams because it makes the hard part obvious: the model is only one piece, while the real engineering lives in the surrounding system design.

• Improving skill-creator: Test, measure, and refine Agent Skills — Anthropic’s update adds evals, benchmarks, and multi-agent testing to skill authoring so teams can verify whether a skill still works as models change. That matters beyond Claude itself because it pushes agent workflows toward a more disciplined software-engineering model: test the behavior, measure the cost, and keep tightening the trigger conditions.

Main Discussion

Topic: Building a practical three-year Power BI roadmap in the age of Fabric and AI

Mike and Tommy use the episode to separate durable strategy from shiny-object chaos. Instead of pretending anyone can predict every product change, they outline a planning model that helps teams decide what should become operational muscle, what belongs in a controlled pilot, and where AI fits without blowing up governance or trust.

• Year 1 is about fundamentals, not fireworks. The most immediate roadmap work is still tenant hygiene, semantic model discipline, deployment pipelines, governance, and reliable reporting patterns. If those basics are shaky, every AI or Fabric experiment just adds more surface area for failure.

• Power BI remains a core business interface. The discussion pushes back on the lazy idea that dashboards are about to disappear because chat exists. Natural-language interfaces may improve access, but curated reports, measures, and semantic modeling still matter because business decisions need consistency, trust, and repeatability.

• Fabric changes the planning conversation. A three-year roadmap can no longer treat reporting, storage, engineering, and data science as separate tracks. Teams need to think more holistically about how lakehouses, notebooks, pipelines, warehouses, semantic models, and reporting experiences fit together into one operating model.

• AI belongs in workflows with guardrails. Mike and Tommy frame AI as a force multiplier when it is attached to real tasks like documentation, extraction, prototyping, code generation, and analyst acceleration. The roadmap implication is that organizations should target bounded use cases first, where human review and measurable value are both easy to keep in the loop.

• Skill development is now part of platform strategy. The future roadmap is not just tool adoption; it is people adoption. Teams need analysts, engineers, and architects who can work with agents, prompts, evaluation loops, and automation patterns without losing sight of modeling fundamentals and business context.

• Standardize where it matters, experiment where it pays. One of the strongest themes is that not every capability deserves enterprise-wide rollout on day one. A sane roadmap defines the non-negotiables that create stability, then carves out room for smaller bets in AI, automation, and new Fabric features without making the whole platform feel unstable.

• Roadmaps should be horizon-based, not feature-based. Rather than promising a laundry list of products, the better approach is to plan for capability outcomes: trusted self-service, scalable data architecture, governed AI enrichment, and faster delivery cycles. That keeps the roadmap useful even when Microsoft changes names, packaging, or release cadence every other Tuesday.

Looking Forward

If you’re building your next planning cycle right now, start by drawing three columns—stabilize, pilot, and prepare—and force every Power BI, Fabric, and AI initiative to earn its place in one of them.

Episode Transcript

0:02 Tommy and Mike lighting up the sky. Dance to the day, the laughs [music] in the mix. Fabric and A. I get your feels. Explicit measures. Drop the [music] beat now. His kings feel the crowd. Explicit measures. [music] Good morning and welcome back to the explicit measures podcast with Tommy and Mike. Good morning, Tommy. [snorts] Good morning, Mike. how you doing? I’m doing well. I’m doing well. I’m

0:33 I’m doing well. I’m doing well. I’m getting excited. we are about ready to get over to the Microsoft Fabric Conference down in Atlanta, Georgia. So, that’s happening. with that, you’ll notice I have a new PowerBI shirt on today. so, pretty swifty one. Pretty swifty one. Tommy would like it. one thing I’ll just note or throw out here, I’m going to do a giveaway for some of these shirts. So, if you like these shirts, if you enjoy them, make sure you find me at the conference. And I’m my thinking right now. And how I’m going to

1:03 thinking right now. And how I’m going to do it is come find me at the conference. If you get a picture with me and post it on your social medias, you’ll be entered to get a shirt like this from the conference. I’ll be actually I’ll have a bunch of them with me. I’m going to give some shirts away like this while I’m at the conference. for you to get entered into the drawing, you got to find me. I’ll be wearing some PowerBI shirt everywhere at the conference. So, make you make sure you find me, come up, ask for a picture. I’d love to get a picture with you. Let’s post it on social media and that will enter you for a drawing. I’m going to give away one or two shirts per day

1:35 give away one or two shirts per day throughout the duration of the conference. So, make sure you find me throughout the the conference and get pictures and stuff and we’ll share some of this fun stuff and you have a chance to win a shirt. Anyways, Anyways, fun stuff. that being said, Tommy, how are things going with you? What’s I guess we should introduce our main topic here. This is an interesting topic today. today. So, Michael, go ahead. You don’t do the topic. So, that’s true. I forgot I knew the topic. [laughter] This this daylight savings time has just

2:05 This this daylight savings time has just messed me up. So, the main topic for today will be talking a little bit around what’s next., Tommy and I have been doing this PowerBI and fabric thing for a long time now. We’ve made some very early predictions in the beginning of the podcast. I would argue a lot of them are coming true from the beginning of the podcast. We’ve also made some predictions in the podcast that didn’t quite come out so well. So,, we’re going to try to take a look here and observe based on where we’re at right now. What’s the trajectory of where we’re going, what does Mike and Tommy

2:37 going, what does Mike and Tommy say? Let’s look ahead three years and,, do a little bit of dreaming [music] [music] and see where we might land in three years with the podcast. Wow. Or or PowerBI or fabric. Let’s So today’s going to be a little bit a question from the from the mailbag here. Someone’s asking about like what’s next? Where is things going? What do you see as experts in the community? Like where are we headed? And so I’ll just maybe spitball some ideas of where we think things are going to land. But in lie of that, we do have some news items. Anything else you want to add to the main topic, Tommy, before we get news?

3:07 main topic, Tommy, before we get news? So always a fun thing where I AI just comes in the clutch again. yesterday out of nowhere my computer started it just blue screened out. I was like okay happened once initially. Sure. Sure. And an hour or later it happened again. Okay. And it was a different error. So obviously you’re like is it and I’m I’m also very fragile to this because the last time I had a blue screen it said uncorrectable error which is never a

3:40 uncorrectable error which is never a good thing. Like it just basically meant like yeah you got to find a way to install Windows or like clear your C drive and it happened three times within an hour and then it happened during a meeting. Oh boy. Oh boy. I was on Yeah. Not good. Like he’s like what the heck? So Chad TB it it and like just going through I was like took an image on my iPad because I couldn’t use my computer. Yeah. Correct. and Tesh is like, “Yeah, you looks like you got a real tech USB,, like Wi-Fi because I my

4:11 USB,, like Wi-Fi because I my PC doesn’t have its own like Wi-Fi adapter.” adapter.” Oh, okay, Oh, okay, fine. But like I’m looking at the device manager, there is no real tech and I keep diving. I’m like, it’s hallucinating. So, I fed it to Claude. And over time they’re like, “Hey, maybe just to verify because I’m like I have this like TP wireless ones like and I just added the manufacturer into like the code they gave me like yeah shell script and it was made by realte. So apparently that and like yeah

4:43 that and like yeah the drivers something there was just having a heyday affecting the you the GPU drivers like this is the dumbest things. So, I will admit, Tommy, every time I go through like an Nvidia update, I always get like two or three days of like maybe it’ll fail, maybe it won’t, or if I restart the computer right away, it goes to like a a hiccup or something. I don’t know what it’s doing. Something’s weird with the drivers on things. So, that’s awkward in general. But yeah, I agree, Tommy. There’s something it it

5:15 agree, Tommy. There’s something it it runs really well and then it doesn’t and then you’re like, “Ah, when I need you, you’re not you’re not available. which is annoying. Said, “Hey, what’s your driver?” Like, “Yeah, you should revert to like four versions before this.” Wow. Wow. Like, and that’s what I did. I have Yeah. Yeah. So far so good. So far, so far so good. Yeah. Wow. That’s wild, Tommy. Let’s get into some other news items here. I So, interesting story. AI saves the day again., I like some deep thinkers and one of these deep thinkers in our community is Kurt Beller. You may

5:46 in our community is Kurt Beller. You may have heard or read of his articles before. before. who he is. data goblins is kind who he is. data goblins is his brand on the blog side of things of his brand on the blog side of things and he there’s been a lot of chatter around AI is gonna take all your jobs. AI is going to every I don’t agree with this statement and as I’ve been unpacking this a little bit in general I think the message is wrong. I don’t think AI is going to take your jobs. I think AI is just going to shift the work that you do. And I was watching another article Tommy. Do you watch Network

6:16 article Tommy. Do you watch Network Chuck on YouTube? Have you heard of the guy Network Chuck? Who? Who?, I think you’ve told me about him. Network Chuck. He’s like a huge beard and he talks like a lot about like networking and computers. He has like a big big home lab. He came out and he said, “I almost quit YouTube.” And he just really felt like the AI space was just providing a lot of like angst, intensifying his work, and he felt like AI hasn’t really made his life easier. It’s not like he’s doing less work. If anything, he’s bringing on more

6:47 work. If anything, he’s bringing on more projects, doing more things, dividing his attention faster than what it was before. And so now he’s feeling this like heavy weight of like the AI intensifies your work as opposed to like reducing it. One person can now do the work of 10 people. So now how do you operate more efficiently with AI? How can you get more done with AI? That’s like the message here. And everything keeps changing so fast. That’s is one thing that’s also very interesting. the speed in which things have been changing is is rapid here. Anyways, I just really felt like his

7:18 Anyways, I just really felt like his points that he was making here were pretty relevant., another article had just come out from a person who blogged on Microsoft talking about,, co-pilot will replace your reports., experts are no longer needed. AI will be able to optimize all of your PowerBI things for you automatically. You won’t need any people.,, data agents will replace all your dashboards. You’ll you’ll all your personal data. coding agents with tools will replace all your developers. And so,, while I think

7:48 And so,, while I think maybe this is a weird concept, but if we stay right where we are, Tommy, right now, like if I if I put my head in the sand and said, “I’m never going to learn AI and only kept doing what I’m doing right now without moving forward, without evolving, without learning, I think some of these statements around AI would potentially be a bit more true. It will replace the person who I am right now. now. But But yes, yes, I’m not going to stay still. I’m going to figure out how to leverage it. I’m

8:18 to figure out how to leverage it. I’m going to figure out how do I change my business to better leverage AI in the business. It will definitely change what I do, but I’m not anticipating on stop working. I’m anticipating like leveraging the new things as they come out as fast as I possibly can to accelerate my business. Tommy, you’re you’re incorporating AI all over your business and you’re a solo partner. Exact. Yeah, exactly. And, just look at our podcast. We should be calling it the explicit measures. Will AI win podcast. So, and here’s the thing

8:49 AI win podcast. So, and here’s the thing though. It it does think and it it I’m not going to do a hot take here because I do want to say like the whole it will replace your job thing. You it will replace your job thing., there are know, there are How would you change that phrase? like what what what in what way would you change that phrase to make it actually an accurate phrase? I me personally, I’ll lead with this one. Okay. Okay. I think I would change the phrase AI will replace the work that you’re currently doing with a new way of doing work.

9:19 work. It’s for and then the the tagline or the other note on this is it’s up to you to learn how to leverage AI to shift the work you do. Like you you’re going to need to do different work. it will be different than what you’ve been traditionally doing. How do you change yourself? What do you need to learn to evolve into this new thing? And that’s where I’m my stance is landing. landing. I’ve said this before, but and I’ll say it again because I think it’s the thing that probably does hold the most true is

9:50 that probably does hold the most true is AI is not going to take over your job. Someone who’s using AI will much higher likelihood will come over. Yes. Yeah., in the video with from Network Chuck about how he almost quit YouTube, he made an interesting point. He goes, “There there’s not a need like imagine someone coming in who’s never learned the technical side of anything and only has used AI. What happens if the AI really makes a mistake or does something really wrong or you have to actually fix something

10:21 or you have to actually fix something that’s not the AI can’t get it figured out? out? There has to be someone. you have to have some skills to figure out what the AI was working on and be able to fix or pull apart the problem or diagnose exactly what’s going on., I do think the AI is going to get better on everything, but I I think it’s going to be like people who step into a role or position that have a lot of technical skills and can use AI. I think that’s that’s the

10:53 can use AI. I think that’s that’s the new secret sauce, right? Just saying you don’t need any technical skills anymore and you just trust the AI wholeheartedly. I don’t think that’s right. I think there’s still this level of like you need to understand what’s happening. You still have to understand the systems, the frameworks, how it all fits together. I don’t think that’s going away. Oh, 100%. And this has been our biggest thing, right? Where I think we’ve had an episode or it’s on our backlog about could we hire someone with no,, with little to no technical PowerBI

11:24 with little to no technical PowerBI skill. skill. Sure. Sure. And you and could they be successful doing PowerBI? Heck, many people that we know who are MVPs started in PowerBI with little to no technical skill and just ran with it. So the I I yeah

11:41 just ran with it. So the I I yeah honestly I think just whatever you do there is something that our our new kind there is something that our our new wave of the future the technology can of wave of the future the technology can just enhance what you’re doing. I agree with the whole you’re doing more work but some things it can do better than myself and I think for a lot of people is there’s not going to be just the the Tommy bot that’s going to take over. Hopefully not., but it’s really much more on if I if there were three other people like, well, I’ve been using these

12:13 people like, well, I’ve been using these agents to go through all my Power Query, I don’t have to do any of,, little to no the tedious effort. Well, that person, I think, has advantage over me. Yes, I agree. All right, Tommy, I want to throw another article in our mix here. So, this is one I I don’t think we’ve actually landed on this exact issue, Tommy. This this next one here is I think this is a white paper. it’s a it’s a computer science white paper about talking about agent OS. Agent OS.

12:44 about talking about agent OS. Agent OS. And I think Tommy, you would really enjoy this topic just in general. And I just want to throw this out there. Tommy, remember we talked an episode a while ago about how the languages of the computer are usually human written. They’re made for us, right? Everything everything around computers is designed for people with hands, people with eyeballs, and you move this mouse around, you click and you interact in the computer in this like visual clicking way of the world. I believe this paper from

13:14 of the world. I believe this paper from agent OS, which I I stumbled across this across Twitter. Someone was talking about it. I don’t actually have the actual Twitter article about it, but I I just have the paper here, the white paper. This is really interesting. It’s basically talking about large language models operate on a different level than what people do just in general. They’re just different. What would happen if you just reimagined everything about a computer and thought let’s just not even build an operating system. system. Let’s just build everything that is

13:45 Let’s just build everything that is agent first. What would that look like? And I think I really enjoy this this thought experiment and someone will get to it eventually and someone will actually do like what’s in inside this article that’s being written this this white paper here someone’s going to come out with an an agent OS only thing right what does this mean Tommy for us like does this change how we build circuit boards does this change how we build the human interactions to computers this to me feels very much like let me give you another thought a

14:15 let me give you another thought a thought here along this one is I’m traditionally used to talking or typing on my computer and using my mouse. Every every communication to the computer that’s just talking to the computer, it’s okay at best, I’d argue. what, Tommy? Like talking to the computer, there’s like a lot of afterthoughts. Like if I want to transcribe my my voice to the computer and ask it to do something, it’s not as fluid because everything computer has

14:46 because everything computer has been built around keyboards and mouse. There’s some programs like Whisper. There’s some other applications you can put on your computer that make that a little bit easier or smoother where I could just talk to the computer and it responds back with information. But I think this paper is alluding to like we need to rethink all that now. Like we have to rethink fundamentally how everything’s written. The code that we run, how the agents can do things. Does the agent even live on the same type of computer? Do we even need screens and and mice? Does it

15:16 need screens and and mice? Does it change? Is it is it something different now that we interact with it? This feels to me to me like a lot of like the Tony Stark stuff,, in Iron Man. He’s got this Jarvis thing that he talks to, right? This feels to me like we’re now talking about Star Trek type the ship is the computer, right? Hey computer and you just talk to it and do what you want. We don’t we don’t really have that level of connection or that level of like integration. We haven’t rethought what does this look like? And I just find this interesting

15:47 like? And I just find this interesting Tommy in the future will we be buying agentonly devices? Will you buy a 500 agent only thing that just is becomes a computer that’s just the agent and that’s what you do work with? I don’t know Tommy. This is where things get a little bit interest like if you really this is only a very much a proof of concept though. Totally. Totally. It just it’s a to me right now it’s a thought experience like no one has this built. It’s not out yet but as the speed and the rate at which we’re going this is probably

16:17 at which we’re going this is probably not very far away. And I just want to maybe point out Tommy are thinking around like agents getting their own programming language more efficient to them., computers changing how they are built because of agents. Like I think this is starting to come through the system. And I maybe also liken this to another analogy that I’ve seen, Tommy, which was cryptocurrencies and like the hyperpersonalization of how do you mine a cryp cryptocurrency, right? We started with graphics cards everyone’s buying graphics card to mine cryptocurrency and make some bitcoin

16:47 cryptocurrency and make some bitcoin here and there. And now we have like these AS6 miners. We have custom hardware. We have entire data centers dedicated to just working on mining these different coins. I feel like this is the same thing. We haven’t quite hit that moment yet, but the technology is so revolutionary. It’s going to fundamentally change how we build computers. Sure. Sure., this isn’t nice to think about, but I’m not as concerned here. So,

17:17 So, I’m not concerned, but how far away are we from something like this? Months, a year? year? No. I I think for Mike the because it’s basically saying that we’re going to replace your OS because we still have to use bash and python and everything. There needs to be a language developed then yeah buy that set of OS. I am sure there’s going to be things agents are running on but I don’t think you and I are going to be buying that at all. Oh I think you’re going to buy it Tommy

17:48 Oh I think you’re going to buy it Tommy 100. The first company that comes out with this Tommy’s going to be the first guy in line. I need this AI. [laughter] Listen, I talked to Microsoft at Build in Chicago when the Surface Pro or the came out with the AI version. Yeah. Yeah. Like like can I try like what does it So I’m like does it run AI models better? Like well I was like okay never mind then. So I I think we’re getting there Tommy. Like this is this is part of like where

18:18 Like this is this is part of like where I think we’re going. I think I think the company that designs a bit more agent first. I also want to maybe in lie of this article about agent OS Tommy maybe I want to propose another idea here as well. I think we’re going to get to a place where the models some of the models you use right now the premier ones the really big ones are going to be difficult to run on your machine on your computer. However, there is probably a mix of lighter weight models that could be run on your computer. Maybe doing some image image inferencing. Maybe doing some lightweight tasks around what

18:49 doing some lightweight tasks around what an agent is doing. I’m waiting for the moment, Tommy, where agents can pick where they want to run and what models they want to run. I I don’t think we’re there yet, Tommy, but I think in lie of like this agent OS, I don’t me personally, I don’t feel like I want to rely solely on only the big fang companies to produce tokens for me. I feel like there needs to be another third alternative or a different way of doing this because I I just that just

19:19 doing this because I I just that just the ethos of it just rubs me the wrong way. I I feel like there’s another way of thinking about this, right? If I need to do a task and I need to generate images and translate scripts and and do some coding or not every single one of those tasks need to go to a premier model all the time. So, how do you have a router, a decision tree at the front of whatever your requests are and make the intelligent choices based on what it’s doing, right? These long running tasks, lots of agents running and building things over time. It just feels to me like that’s an

19:49 It just feels to me like that’s an interesting concept. How does that change what your computer looks like? Is there a smaller inferencing piece of it that you keep locally and you lift a lot of the lightweight tasks away from the heavy thinkers and you reduce the token usage at the heavy level? So I think I think there’s something here to this. There’s something that’s going to say you’re always going to need some very heavy very smart models to do something reasoning thinking planning big topic information. But then once that’s gets distilled down you can break that down

20:19 distilled down you can break that down into small tasks. And I think maybe those small tasks could be handled by less intelligent, easier to manage, you less intelligent, easier to manage,, lighter weight models that just do know, lighter weight models that just do very specific things. I don’t think we’ve gotten there yet, but I see the market moving towards hypertuned, hyper customized models that help you build better stuff with less tokens. I think that’s going to be the name of the game moving forward. Yeah. Well, very theoretical. Let’s dive into something a little more tangible., this is from Microsoft

20:51 tangible., this is from Microsoft Fabric again and not until FabCon are going to start seeing all the articles. It’s the calm before the storm, but there was an update. It’s called extract label schemadriven unstructured data extraction in Microsoft Fabric. This is a mouthful. What What the heck is this? is this? Yeah, I know, right? Yeah. So, speaking OS and we’re talking operating systems, fabric fabric said, “Hold my cup. Hold my drink. my drink. Hold my drink.” [laughter] Yes. What this actually solves is most enterprise data lives in unstructured text that especially what

21:21 unstructured text that especially what people want to use for NP agentic. Sure that usual extraction method regular expressions and rules are not great for especially scalability lms can help but raw raw models are usually inconsistent and slow. Yes. Yes. So we need a little more intelligence. So this is actually something that they develop which is a schemadriven extraction in Microsoft fabric AI functions. Nice. Nice. So that just enforces a strict output

21:51 So that just enforces a strict output the types and basically extends by the type of the fields the vocabulary etc and you simply define a schema you apply to whatever text and you always get consistent structured output. It’s basically the idea when you like if you’re writing to a chat like hey provide me the answer in a table you provide me the answer in a table pro providing this format just know pro providing this format just doing this now over your data extraction. So

22:21 extraction. So this is interesting Tommy maybe I should incorporate this into all of our hundreds of blog posts that we wrote on the website. All right so as a as a shameless pitch in addition to this article which I’ve just put in the chat window. So extracting key topics, it let you talk about like,, you can extract, I’m looking for a name. I’m looking for a profession. I’m looking for a city. You call out like key terms and then it uses those key terms to scan through large chunks of text. Something along those lines. Mhm. Mhm. Tommy, I’m looking at the podcast and

22:51 Tommy, I’m looking at the podcast and things that we’ve done on the podcast. One of the things we talk about, we talk a lot about topics and computers and technologies and various things. This would be interesting to add this. We’ve also added 300 episodes full transcripts on our website. So PowerBI tips has on the podcast. So PowerBI tipsodcast you can go through and scan through and search all 300 the last 300 episodes until now all the text and code that is there. We have some topic extraction from that. AI has helped us take the

23:20 from that. AI has helped us take the transcripts down, format them, and take our thoughts and consolidate them down to like a summary of the of the episode. So, we’re it’s still Tommy and I saying what we’re saying, but it’s it basically it’s using that text as a reference to what how we’re describing the episode and what was talked about. All that’s available now on the website. Tommy, this seems like a great opportunity. We have tons of text laying around the website. We should be getting some key extraction of terms and terminology and adding that to our corpus of text. Yeah, Mike. And this is actually a

23:51 Yeah, Mike. And this is actually a perfect example of something I’ve been trying to think about where I I told you I have this like GitHub starred application I built that I want to be able to have basically intelligent search on it like hey what are all the repos that I starred that have a cloud skill in it right oh yeah oh yeah scanning all the text that’s inside the thing thing the read me yeah really just the read me I extracts the read me and it’s a better way for me to like manage my stars, but I can’t get the AI tooling

24:23 my stars, but I can’t get the AI tooling to work to do like an intelligent search because I think this exact reason here. Let me ask a question around this, Tommy. When when you’re doing your star, so I’ve seen your little application, this little side one. So, basically, it reads your GitHub page. It looks through all your stars. You’ve got hundreds of stars at this point. You keep finding really good stuff, but you need a searchable way of going back through and the search on GitHub for repos with stars is okay at best. It’s not the greatest. greatest. Yeah, exactly. Exactly. So let me ask a question Tommy to that end. Where do you store the data in your app? Where is that going?

24:54 that going? So right now I believe it goes to a SQLite database SQLite. Is there a reason why you chose that over something in fabric? because I I had something I wanted just to test locally first. Test it. Just fast build. It knew how to do it. do it. Fast build. Exactly. So I’m going to propose an idea here or concept. Tommy, I think in these fast builds. I’ve done a couple projects now. I have been shifting where I put the data. You you can build with agents. Agents can build things on open

25:25 Agents can build things on open platforms. you can do things like open table and all these other tools that like air tableable are these are all solutions where you can store your data. SQLite is another good example, right? You can have that you can you can spin that up and run it and put data into that SQLite database. Tommy, I want to propose to you the idea that in when you do these kinds of projects, your first thought should be how can I build the app to land the data in fabric. I have built an app similar

25:56 in fabric. I have built an app similar Tommy, you’ve heard about this one, too. It’s it’s basically helping us Tommy and I create content around the podcast. How do we collect? We have a lot of sources of data where things are coming from. We have to prepare a lot. We have to prepare for the episode. I think Tommy there’s a really strong story to build your app directly on fabric. This is I’m exploring this now. I find it to be incredibly valuable. And when something like this comes out, right, when you have something like the extract label AI function that appears,

26:27 extract label AI function that appears, I can tell my agent, go create a notebook. in the notebook, use this function and take all the descriptions, all the the text, all the corpus of text, take all these things out and go run this AI on top of it. Tommy, that’s so valuable to be able to have the data already in a world where you can here’s a SQL database and now I can just run a notebook. Now I can just start using AI. Now I can do other things. I need to build new tables. the notebooks

26:57 need to build new tables. the notebooks and other supplemental data tools will help me engineer the data right in next to my actual support data. So I have found me personally Tommy I’m finding a shorter innovation cycle occurring by moving the data directly to fabric for me me now I haven’t done full testing on like really big scale applications I haven’t done full scale testing on a lot of these things yet I haven’t done dev test prod of this I’m just starting like the side projects right these are the the

27:27 side projects right these are the the little things Tommy like you have the star project I have a little web app a front end and I have a SQL database behind the scenes and I’m starting to now build notebooks and some data transformations and some UDFs that are doing some supplemental things in the middle of stuff. So, I think that’s really interesting here, Tommy. Really like that article., I just want to give a shout out to two more articles and I I want we’ll talk about it I think in the future even though we’re not going to talk for another like three months after this., is that worth the read? It’s a It’s

27:58 , is that worth the read? It’s a It’s called It’s a Megentic AI. Tommy, you just left it because it was it’s Italian. Italian. A Mario. A Mario. It’s a Mario. Well, it’s actually the person who wrote it. Hanley did a great job of basically saying, “Hey, how do agents work?” basically like and what what is the best practice to set up an agent and to make it as most efficient as possible. And it went through was like, “Hey, imagine Mario.” Mario.” Mario, small Mario when he start.

28:29 Mario, small Mario when he start. That’s the base model. Yeah. Base. Yeah. And he’s like, he can walk, he can jump, he has some basic things. Move left and right. Those are his base capabilities. but he’s fragile., it’s ready for pre-training. Yeah. Yeah. And then there’s basically things where you can harness the model. So, like the super mushroom gives them instruct instruction tuning,, guardrails, memory, and then there’s powerups, agent skills, and tools. What would that be in Mario? The fireflower, the raccoon leaf,, the the

29:00 the raccoon leaf,, the the star. And each one of those little Yeah. Give him an ability. And,, he also goes through like the equivalent of that, an agent. Then the mushroom kingdom. Here’s your environment thing,, and here are all the tasks you have to do. What things you should you avoid? What things should you interact with? The objectives. Cute. And Yeah. And then finally, even the most boring evaluations and reward modeling. So know it’s doing a good job. The flag pole.

29:30 The flag pole. Yeah. The end of the the end of the level. Yeah. Yeah. So, it’s actually a great way of like if you’re setting up an agent, one of the best ways think of it in the terms of Mario. I love this article. This is a very fun article and I love the analogy. This is my childhood, Tommy, playing Mario when on that little square controller that would just dig the corners into the corners of your hands so bad. This I the memories here of this those early levels. How many times did I play that Mario level one jumping around on those those those initial levels? Amazing

30:02 those those initial levels? Amazing stuff here. So, I would agree, Tommy. This is a great experience here. One thing I’ll this article I think alludes to it is as you’re getting better with agents and figuring out how to use these things in your daily workload. I think there’s this idea of unpacking solving problems. Be the problem solution fixer, right? A lot of these agent things are helping you don’t you’re not writing code as much, but you’re trying to solve the problem. And I think for me, I’m finding myself switching a lot of what I do and moving

30:32 switching a lot of what I do and moving directly into what problems can I solve today as opposed to what code of language should I be writing, which is I think a great shift. All right, Tommy, last last article you’ve got going on. Last one. Everyone needs to take a look at this. Claude has simply updated their skill creator skill, which is basically if you were trying to create a skill, didn’t know where to start, Claude had its own skill to do that, but they’ve really updated it now to your 2. 0., which has a lot more capabilities., and finally, it even adds evaluation. So, okay, create a

31:04 even adds evaluation. So, okay, create a skill. Does it work as you do as you want it to? And you can actually go through and test it out to find you what’s good output, run test to confirm that the skill behaves as expected. So, I love this. This is great. I like this note in the middle here. There is faster, more consistent evaluation with multi-agent support. AB testing. Why not run your skill or derivations of your skill multiple times? Write different versions of it. Test it. An executor with the

31:34 of it. Test it. An executor with the skill, an executor without the skill. Which one performs better? Evaluate the output and then decide if the skill is required or not. Or you have a skill that you wrote three weeks ago, three months ago, whatever the old skill is. Okay. Do you need to update it? Is it still holding on to old instructions? How do you revise skills over time? This is a new like Tommy. This is the same stuff as Git, [laughter] right?, this is, hey, I’ve wr I’ve written some code. I wrote it three weeks ago. Is there a better way to rewrite it? How do

32:05 there a better way to rewrite it? How do we optimize it? What are we measuring? This is really relevant. I think this is so exciting to see this again. Remember when we talked earlier in the podcast at the beginning of this, Tommy? We’re not going to be doing the same things. We’re just going to be doing different things in a different way. Now I’m looking at this article going, I think my new task now is to figure out how to be the best skill builder for my agent. Right? That’s the secret sauce. Yeah. Yeah. We’re moving away from me pushing data around in Excel sheets. We’re moving away from me writing Python and and specific scripts in SQL, but now I’m

32:37 specific scripts in SQL, but now I’m moving to what skill do I need? What what do I need to write to the agent? How do I build its knowledge so it knows how to do the thing that I want to do? That’s that’s the new skill I’m moving towards. I’m moving more towards this direction to make the agents and elements more effective for me. This is interesting. I guess great article, Tommy. Really good share. Awesome. All right, so Mike, let’s get to our mailbag. Yeah, do the mailbag. Tommy, read us away for us. Wait, don’t we have a sound for that, too? too? I don’t I don’t think we have we have a

33:07 I don’t I don’t think we have we have a sound for mailbag. You got mail. There we go. Perfect. All right, here we go. So two years down the PowerBI implementation in my company with 200 prousers create creatively exploiting about 20 semantic models. models. Nice Nice standard and self-service reports with 300 insightful KPIs. Love it. Love it. Talk about that. 300 300 management. Yeah, I know. Management is asking me what is the next big thing we

33:37 asking me what is the next big thing we are to do in PowerBI. I don’t have any clue what is next. What should we do? Can you guide me on the road map for the next three years at this juncture? This is interesting. Good question. Well, the next three years, you should build everything in an agent. You’re not going to be needed. Just replace yourself. [laughter] That’s fine. That’s what you would hear if you just read the the the random LinkedIn or posts or or media posts these days. It feels like you’re just going to be replaced. I don’t think that’s true. That was totally a joke

34:08 that’s true. That was totally a joke there at that point. Tommy, where do we take this one? What do we want to talk about here? what are the key aspects you would want to think about? So I want to really dive in here in first off what is the next big thing for an organization because I don’t know if the organization’s thinking like a developer and they’re really looking like really I think there’s two things we need to talk about what is the road map for PowerBI and what should I know as director developer and utilizing the tooling but

34:40 developer and utilizing the tooling but also organizations once you have your models. What are the capabilities that PowerBI can do once you’ve hit some milestones? Sure. I want to unpack this original question around management is asking me what’s the next big thing we should do with PowerBI? Where do we go next? What’s what is the next area of this?

35:01 What’s what is the next area of this? I’m I’m going to make some assumptions here like I I do. I read between the lines here a little bit of what the what the individual’s asking. One, love that you’ve got 200 Procreator users in the system doing things. looks awesome. It looks like to me, if I look at this world, it’s a lot of inward focus reporting. So, I feel like a lot of the reporting here is your team internally with your own prousers. Love it. One area I would maybe recommend is looking at usage, what’s being used,

35:31 looking at usage, what’s being used, Tommy. And I know this one. There’s this idea of like drift a little bit or or report drift or people just doing custom things. things. If your company is small enough, you probably can get away a lot with just everything in Pro. Not a problem. If your company gets larger and you have bigger data sets, you have to start thinking about smarter ways of loading data. And so I’m I’m going to,, at this moment where you’re at in the company, I love it. I think this is all good. I

36:02 love it. I think this is all good. I love that you have a lot of creator users here. I think you would want to focus on the process of what is new content being created. What content is being highlighted across the organization? It doesn’t sound like we have like certified or promoted data sets occurring inside the organization. So when you have a lot of businessled creators, you’re going to get a lot of stuff that’s made. There’s there’s a lot of effort around create, read, and update. I think there’s very little

36:33 update. I think there’s very little effort around how do we mark something for deletion? How do we sunset a report? How do we move people onto a new report as opposed to just half the team using the old report because it just worked for them and half the team using a new report. So, I would focus on some of those things. I would start incorporating some more like monitoring governance controls starting to really establish has anyone in the team gone outside the rails of what’s reasonable? Is someone building a whole bunch of reports with tons of bookmarks?

37:05 reports with tons of bookmarks? Can you maintain it? Right? That’s my question right there. Right? And and those are guard rails that you’re looking at. Okay, we’ve given we’ve given the ability for the organization to build what they want to build. How are we now using that to help us become more efficient? Yeah. Yeah. So, that would that would maybe be one aspect I would look at for what’s the next thing? Start looking at monitoring. Start looking at performance. What’s being used? How often is it being used? which teams are creating the most. I think you want to provide more analytics back up to leadership that says here’s

37:36 back up to leadership that says here’s how we’re using our investment. Let me pause there. Tommy, what do you think about that? You just said for me a trigger word and something that I’ve realized more and more with any implementation of PowerBI and especially fabric is what pressure before I answer Mike what pressures are CTO’s under or what pressure do they feel when they think about PowerBI Microsoft Fabric? Fabric? Do they feel any pressure? Yeah, I I’m sure they feel pressure about

38:06 I I’m sure they feel pressure about something. I’m not sure I would answer just around PowerBI and fabric., I think there’s like this,, everyone’s trying to save money somewhere along the lines. I think CTO’s get this level of I need more access to things. Give me what I need to do better data stuff and and the CTOs are constantly inundated with like the next newest coolest technology piece, right? So, five years ago, Tommy, all the CTOs were like, “What is this PowerBI thing? Why do [clears throat] we like do we really need it? we already

38:36 like do we really need it? we already have like Tableau, we already have other tools, you already have Excel, do we really require PowerBI? And the business keeps asking for, I think I need it, Microsoft’s pushing it, we should really get it. Let’s start using it. There’s this push for this, right? I think the CTO’s job is they’re constantly being inundated with like innovations and technology and the business is constantly coming to leadership and saying we need this, we need this and and the the the CTO’s of the world have to say we need to pick a direction. We

39:06 to say we need to pick a direction. We need to align our business down a path of technology stack that we think will be good and a good long-term investment. Right? I can’t make a decision for I’m going to just throw out some software here, Tommy, that may or may not be good. I’m just throwing things out like Tommy, you’re using notion or you’re using perplexity or you’re using something and you’re finding value from that single tool. Tommy, maybe I’m not using it. Maybe I’m using loop or some other note-taking or I’m using oneote. What do we use? What do we pick? What

39:37 What do we use? What do we pick? What what is our company’s stance on how do we do this? And what tool do we pick? Is it going to enable us to do more in the future or is it going to hold us back? Right? if we if we stay with one note [laughter] right if for Microsoft do we hold oursel back is that program going to continue to innovate where we want to go directionally with the company and so I think this is like the challenge of CTO’s so I wouldn’t necessarily hone in on the PowerBI story exactly I would say CTO’s are trying to make

40:07 would say CTO’s are trying to make long-term decisions to pick the right investment in software so that they’re not held back over time they can continue to innovate and move forward. Is that a fair statement, Tommy? Is that how you would perceive it? I would I’m going to add to that because I think for a lot of CTOs are really who’s in charge of the integration is really how do you ensure that the company sees PowerBI and fabric as an investment and not a cost because that’s usually where things change and sometimes is a perceived value. But

40:39 sometimes is a perceived value. But I know Seth has said this. Yeah, Seth would always said this like P data is not important at all at a company until it’s the most important thing. thing. That’s true. And the thing is we can you can do all of the integration to lakeouses and have all these models but if everyone it’s like we’re spending how much on PowerBI not realizing again no one says any of that for using LinkedIn sales navigator for your sales team right or re

41:09 for your sales team right or re investing in good applications and services for the operations and finance to operate and for whatever reason PowerBI is always looked at as a cost like it’s this like fancy features like no no no all the money we’re putting in as investment not just to show your data in a visual a visual right and I think that’s a huge part of changing that story so an organization asking this question is really looking at what else can we do with our investment and you you

41:39 we do with our investment and you you said that as well the the monitoring the the governance side of thing I think is great But I would say that’s not necessarily I don’t how do I say this is that something that the organization as a whole if they’re asking this person who wrote the mailbag basically like how can we better see return on our investment that’s to me that’s what I hear. Okay I agree with that. I’m going to add another thought here to this one as well

42:10 another thought here to this one as well Tommy. What other things that management would want to push you forward? I I think after this two-year mark, I’m also thinking about Tommy. Remember the models you built two years ago, the data models you built some stuff. It was okay if you if your future self two years in the future came back and looked at what you were building. What do you think you would say? I don’t think you would be too happy. So be you would been like [laughter]

42:40 be you would been like [laughter] what was I doing? What were Yeah, exactly. What was I thinking with this? So I’m I’m going to also say another what’s next for the team here. I think another one here is I’m going to use a term here like domainbased models. models. Sometimes you build models that are too small and doesn’t have enough information in them. So then you get like three or four models. So you have too many models. In other situations you have one monolithic model which is very difficult to manage and update and use and you need to like cut it apart. So it

43:14 and you need to like cut it apart. So it says right here they’re creatively deploying around 20 semantic models. I think one of the next things is just to really go back and look at analytics and usage. Going back to that monitoring thing to a little bit degree, right? Turn on workspace monitoring. Look at how those models are being used. What data is in those models? Open up a feedback form in the reports or somewhere nearby that says what data are we missing? What other data should be brought to these reports? But other how can you be more actionable in your day-to-day business if we gave you a bit more data. So I would probably another

43:45 more data. So I would probably another effort here I would probably focus on focusing on what domains runs your business doing a good alignment between the reports people are building and what’s inside the semantic models. That would be an area where I would invest in and figure out do we need more models than 20 than 20 or is 20 too many and we could actually consolidate things. Do we have any duplication of work? Are we loading the same data multiple times? So I’m I think this is ma Matthew Roach’s Roach’s

44:16 this is ma Matthew Roach’s Roach’s maxim here, right? Transform the data as far upstream as possible, right? So along with this concept, I would also look at look at any of your on-prem data sources. Do you have data sources that already exist somewhere inside your on-prem world that would be easier to get them inside fabric or your analytics platform by moving some of those solutions to the cloud? So, I think there’s also potentially if I look at this some modernization you’d want to look across your company and say how could I modernize where data is going,

44:47 could I modernize where data is going, how it’s being stored. Do we have any very slow on premise hard to connect to things? What could we do to reimagine that in a new,, PowerBI cloud focused vision? What does that look like? like? Yeah. Yeah. And and when you do this, when you start transforming things farther upstream, inherently, you’re going to start trying to use things that are fabric- based. You’re going to want to think about lakehouses. You’re going to want to think about data storage. You’re going to want to think about pipelines. You’re

45:18 to want to think about pipelines. You’re going to want to think about notebooks. I think the progression is to start looking at more of those internal data engineering processes and what can you do to centralize those. Do you have any SSIS? I just saw a really interesting article from someone talking about how SSIS is a good tool but it’s difficult to manage. It’s difficult to change. It’s difficult to track changes. You can do it but it takes a lot more rigor to use that tool to do your data engineering. Is there anything like that that’s just old and hanging

45:48 that’s just old and hanging around that you could modernize? Could I get rid of SSIS and instead use a a pipeline to do the data engineering? What if I could move more data to my lakehouse? Do we have SQL databases we can just start mirroring that increase the speed of the data without adding a lot of extra infrastructure? Right? Those are the questions I would be asking., I also saw a really interesting post from Megan Longoria talking about SQL Server 2016 through

46:19 captures data using mirroring using change data capture. SQL 2025 uses a new method called change data. So the new SQL servers are being more efficient on how they bring data to your lakehouse which I think is the right approach. So these newer technologies are coming out. are you incorporating

46:38 are coming out. are you incorporating them? Does that save you more time and money? So, that’s another another aspect I would push towards. Focus on domain based models. Do you have any? What does that analysis look like? And then start figuring out, can you bring more data s sources closer to your analytical system? Move to Azure, move to fabric, fabric SQL databases. Does that help you move things along quicker? Can you get more data into your world? Are you building to your point, Tommy, your stars app? Can you build the app on top

47:08 stars app? Can you build the app on top of a database that lives in fabric? Does that aid what you’re trying to do? Maybe it does. Maybe there’s something for the field you need to put in there. Let’s put that stuff inside fabric. Maybe. So, that’s that’s where I think for me that would be the next logical step for me. That’s what I would push more towards., once you have that done, then you can start incorporating AI or learning or or this new AI function, right? Hey, we have a whole bunch of data from a website that’s talking about people talking about our products. Are we using

47:38 talking about our products. Are we using that? Is that being leveraged? Do we have topics that we’re looking for? Put that through the AI. That’s where we want to incorporate the AI stuff. What are your thoughts? This is this is a huge opportunity for a lot of organizations because, if you’re already using PowerBI,, and they’re you’re already in a sense half fabric. Yeah. And I think that we we’ve talked about this before about what’s called the Tiger Teams. Do you remember that? Yes. It’s I think of it like as the the secret ops team,, the secret stealth team. Exactly. But you’re it’s

48:08 stealth team. Exactly. But you’re it’s actually a real name called Tiger Teams. Yeah. Exactly. Exactly. So what how whatever the name is is a great opportunity because there’s already a willingness to further invest in, you willingness to further invest in,, further put resources to making, know, further put resources to making,, PowerBI be what it’s capable of. And I think this is a great opportunity to say like okay we have our models we have our reports let’s start integrating our data into lakeouses and really driving that efficiency in terms of would you push right now if an organization came to you

48:40 right now if an organization came to you the anything agentic around PowerBI not in the consumer space Tommy so I want to be very clear I think agentic stuff right now is really good in the, Agentic stuff is really good in the creator space. So, if you’re creating things like making reports, making pipelines, making new data tables, I think that’s a better place to throw agents at helping you create. That’s

49:11 agents at helping you create. That’s where I would put it. Tommy, you and I would totally agree. We’re going to use agents inside VS Code. We’re going to use that to assist us with managing the models, creating reports. This stuff’s there’s things that are coming out right now that are very powerful in that regard. I would use agents for that. Am I going to throw an agent at a semantic model and then just ask it random business questions about my data? Right now, the way stances are, I’m not a big fan of that. It doesn’t it it just seems to be more to me right now. I look

49:42 seems to be more to me right now. I look at it, it looks more like a novelty than it actually does to help me solve problems, right? and let me be really clear here too, Tommy. Microsoft. A couple episodes ago, we gave Microsoft a really hard time around not having clawed co-work inside their ecosystem. Good on them. I have to give them credit and props where it’s due. They have very quickly in a short amount of time absorbed what co-pilot co-work is doing and have added that now to co-pilot. So

50:13 and have added that now to co-pilot. So now Satia has now made an announcement. I’ve got to give them credit for this because I said at the head at the top of Microsoft heads were rolling about like this whole co-pilot and co-work thing. Microsoft now has made an agreement with Claude or Enthropic to bring Claude co-pilot or co Claude man so many words here. I’m going to try to say this right. Claude co-work to co-pilot coowork. It sounds like it’s an agreement to have it all built together and that I think is very smart that will

50:44 and that I think is very smart that will remain to keep Microsoft competitive. Microsoft continually has this story around security around AI. Bringing claude co-work as co-work makes a lot of sense to me. This this is what we want, right? We just need co-pilot to be more effective. And there’s other technology stacks that are doing it better. Microsoft make the agreement. Bring me that tool in a tool that I have access to. I could actually use co-pilot co-work if it functions very similar to what Claude co-working.

51:14 very similar to what Claude co-working. So So yeah, yeah, let’s Does that answer any of your question around aentic things or I just literally just run run through a rabbit trail here? I think that as close as we’re going to get to the agent OS too. So there’s a lot lot Oh [laughter] boy. Oh boy. Yeah. Don’t get me started. I think where if I want a company to see the most bang for their buck right now too would you ever push people to the fabric roadmap roadmap because that’s the other thing I also wanted to talk about too like what what is coming around fabric like what should

51:44 is coming around fabric like what should I be aware of because maybe they do want to see just also the features let’s let’s talk about the trend that we’ve been seeing for the last year or so Tommy fabric has been getting more robust fabric has been adding let’s let’s talk about just the data integration space, right? In data integration, we had initially pipelines, notebooks, and data flows gen 2. That was like all we had. Look at what other products have shown up. We now have copy job showing up. We now have new, you

52:16 job showing up. We now have new, you job showing up. We now have new,,, dataf flow experiences. Dataf know,, dataf flow experiences. Dataf flows gen 2 has now gone multicomputee processing, right? It’s not just single thread anymore. It’s now multi-threaded, right? There’s there’s a lot of changes that have been her happening in the data integration space. I really like those changes. So if we look at that trend, Tommy, they’ve also added like Cosmos DB. They’re adding SQL Server. They’re adding all the other data structure platforms into fabric. I would imagine most of your major data structures

52:46 most of your major data structures should now exist in fabric. Why not have SQLite inside fabric? Why not have Postgress DB in in fabric, right? Where are these other database tools and how can they all just live inside fabric? That’s where I think things are going. I think you’re seeing this absorption sound of all these other data systems getting incorporated into the security model, the workspace security, the, the workspace identity provider, the workspace

53:16 identity provider, the workspace identity. these things are becoming built into the product and that’s really helpful. So I think that is the trend we’re seeing. We’re seeing more tools come to fabric. So whatever technology you like to use, you can just use it and it all works together. That’s I think a bit of the secret sauce. Would you agree with that direction, Tommy? Is that what Microsoft’s doing? I I think so. I I would also point to the fabric road map that’s available on the web, too. I I do really enjoy looking at that. So do you ever take a look at that? I do and I refer a lot

53:47 a look at that? I do and I refer a lot of customers to it because this is one of the parts of my class that I run which is I used to be doing data strategies with Melissa Coats. I don’t know if anyone remembers Melissa Coats back in the day. but data strategies was a company that she ran. She has since retired and is now doing fun things in retirement which is awesome. But I I have taken over the class of that and a lot of that content was originally built by Melissa Coats. she wrote a lot of the original PowerBI adoption roadmap and then Kurt Buer came in and updated a

54:17 Kurt Buer came in and updated a lot of that content and has more modernized it for fabric side. so the fabric adoption roadmap is really useful and you can always score yourself in every area. What areas are you scoring yourself in and are you is your business leading in those areas or not? Where can you do better work on this? So I think that’s a great measurement to understand where your company fits in the PowerBI adoption roadmap. I I would also look too at the

54:47 I I would also look too at the roadmap. fabric. microsoft. com. Oh, you’re talking about purely features upcoming features. Sorry, that’s a feature road map. Gotcha. I was on the actual adoption road map for fabric. Yeah. So that’s what I was referring to where and if you have not been to it. it’s if you want to wait just for conferences that’s fine but if you really want to see what is at least being planned for the next year this is a great thing to break everything out based on like PowerBI one

55:17 everything out based on like PowerBI one lake fabric ecosystem and you can see all the things that are currently planned for PowerBI table visual custom totals that is planned. task flows are still planned even though that’s been around for a little tindle on the web. co-pilot author feedback or apps that are generally available and you can keep going through all these and see basically what also can you try now. So the I like this for saying like here’s what we can utilize. Here’s what’s like

55:48 what we can utilize. Here’s what’s like Microsoft’s working on and maybe we should reinvest in that. Is there anything in here tell me that you’re really excited about that that that you’re seeing in here? I I wish I could say narrative visual, but but all the things for IQ that are all planned right now. So they they actually broke out the IQ as its own category here, which I find telling. I think this is that’s so this is also really really useful. Tommy, what you see appear in this list is what is important to Microsoft, right? To some

56:19 important to Microsoft, right? To some degree, right? Look at how many items in that list are fabric graph related. I think there’s this idea or concept around an ontology and a graph of storage of business logic across your company is going to be incredibly useful for agents to understand how your business works. So that’s going to be another area here as well. can I can I go back to the question here like back to the user’s question around I don’t have any clue what to do next. What should we be doing next? I want to add just one other seed in

56:49 I want to add just one other seed in your brain of how I’ve seen a lot of organizations grow with their PowerBI usage. usage. I think a lot of companies start with a small team if a little bit businessled somewhere it gets involved and they start helping out a little bit with the PowerBI deployment and getting things working. I think you start seeing a trend where from that small team a seed grows and more teams start adopting PowerBI finding it useful learning it and as the as the company in general as

57:19 and as the as the company in general as it’s like a living organism right as the organization starts learning more about PowerBI it becomes proliferated across the entire organization different levels of skills some people are new some people are advanced you have data modelers you have report builders all that though is a focused internally on the business. I would argue once you start getting some foundational stability around reporting in your organization, the next place you should be looking is external

57:49 place you should be looking is external users. How do you delight your customers with data you already have? And look at processes where you have sales and marketing sending vendors Excel files or automating things from it where just flat files and FTP site. Should we really be doing that? I would argue one of the next big movements for your company should be okay, we have the internal need potentially met. Let’s

58:16 internal need potentially met. Let’s focus externally. How do we build portals and pages and delight our customers with custom data for them to work well with our business? I find that’s really useful and it automates a lot of manual wit labor out of your team and makes it very easy for people to be like I call it make make your customer sticky to you, right? Make them love the information you provide to them. I can’t tell you them at time this is just me personally venting Tommy. I can’t tell

58:47 personally venting Tommy. I can’t tell you how frustrated I am whenever I use any online program and there’s like zero API to go get access the data, zero information I can get out of it. I have to use their tool and they lock me in to my data is now stuck in their system. system. I’m fully rebuilding everything, Tommy, in my company to get away from that mentality. every every piece of software that we’re using, we’re we’re relooking at it and saying, “Does this piece of software store data, can I just rebuild it with

59:17 store data, can I just rebuild it with this agent world that we’re in right now? Can I do it? Can I build it on top of a fabric backbone? Right? Can I just have it run and live inside Fabric where I need to do the analytics anyways? That’s ultimately where I’m going to put the data. I’m I’m going to build pipelines. I’m going to load it into Fabric regardless. So, why am I not just doing that now as opposed to later? So, Tommy, that’s one thing I’m really thinking about here as well is, hey, look, I’ve got a lot of this stuff going on. How do I do a better job of

59:50 going on. How do I do a better job of leveraging my initial investment with PowerBI and focusing it externally? I’m going to shamelessly pitch here something because I’ve seen this happen in many companies. You want you will want to have a custombranded your company portal for embedded data. You just will. It’s it’s a good thing to have. We build at Carlos Solutions, we build intlexos, which is an embedded accelerator to help you get this portal in less than an hour. Land it, drop it in your environment, and now you have a

60:20 in your environment, and now you have a custom reporting portal with your own branding, with your own cards. You simplify the whole experience around sharing data externally and you can delight your customers with custom reports, custom table builder experiences, building their own pageionated reports. You can even do some self-service directly to your customers. They can come into the application and build their own PowerBI reports, save them to your environment, but the customer can use them to build and shape their own data structures. So you can build a semantic model and share

60:51 you can build a semantic model and share semantic models directly to a customer. That’s awesome. And and let me tell you, when you start giving customers like this level of like capability, they’re sticky. They’re not going to leave. They have the data they need directly from you, you can get done what you need to get done. So I’m going to really push on,, the next evolution for your company. You don’t have to use my solution. I have something out there that makes it very easy. I’ve been building embedded solutions pretty much since embedded came out with PowerBI Premium. So, I’ve been doing this from day one., with over well over a 100 plus embedded

61:23 over well over a 100 plus embedded accelerators out there now, but you should really consider the concept of we’ve done data internally. What does it look like externally? How do we delight our customers? Let me just pause there, Tommy. What are your thoughts? Honestly, I love that because these are things that were that were near unattainable without an entire development team that we can start building and also really utilize. I think I’m finally joining in with you where I’m realizing more and more that

61:54 where I’m realizing more and more that the embedded solutions and also just external sharings not just for the you external sharings not just for the enterprise or because there are know enterprise or because there are a lot of different use cases that organizations have whether they have their own vendors like that makes it that easy to begin to communicate with data data 100%. So, I think it was only really only shown with developers before because it was such a tough road to set up. So, it wasn’t the lemon wasn’t worth the squeeze. I would agree with that. Yes. Agree. And I think now with with solutions that are

62:24 I think now with with solutions that are out there, you can literally one button press install them infrastructure installed. So, if you go to the Azure data marketplace and go look up Intelos, I’ll put the the link here in the chat window just in case you want to check it out here. But, Entexos is our solution. we offer many different ways of purchasing. You can license the software directly. You can purchase the code directly with a code share, a code license, or you can download it directly from the Azure marketplace. It’s actually a direct link you can go get the software from. And that’s where I think organizations will want to spend more time is great. You you’ve made the

62:54 more time is great. You you’ve made the investment. You have the models. Your team is using what works. Give it to your customers. Sell it to your customers. Now you already have the the data is the product. Use that to your advantage. I think would be very useful for a lot of customers to go to. That’s your next step. All right, Tommy, guide for the next three years. Final question. Let’s wrap up this conversation here. The the the question from the mailbag here is, can you guide me on your road map? What would you say are the next three years of roadmap stuff? Where do you see this going

63:25 stuff? Where do you see this going in the next three three years? Maybe like a short thought around that. Well, I think the biggest thing is collaboration and really utilizing more and more of what Microsoft is solidifying in the fabric environment. Mhm. Obviously, we know the semantic models and obviously we know the basic infrastructure, but I think we’re going to be finding more and more ways to get data in the hands of people, not just for consumption, but really like self-service, but

63:57 but really like self-service, but throughout all facets rather than just only building reports, but really making it easier for different teams and departments to own their data. And that’s where I would be focusing. Great. Great. I think I’m going to point on an area where where Microsoft is continually making data more useful and I also think the speed to deliver insights is increasing. I would probably look at my organization

64:27 I would probably look at my organization and say let’s look at where do we have where’s what is the gap between get data and make decision right where is that long does it need to be long Tommy like one thing I have found as a business owner if you want to sell if you want to have a good sales thing you need to be very responsive Not the early bird gets the worm, but like I don’t know, there’s probably

64:57 like I don’t know, there’s probably another phrase in here something like that that I don’t know off the top of my head. But if you have a request from a website and you can respond back immediately within minutes or you have someone interested in what you’re talking about, how can you get a decision or some information in front of them as fast as possible? So I would look at your organization and the next three years Microsoft is going to continue to add things into fabric. I think Microsoft is investing heavily in a lot of real-time things that are efficient to use, easy to set

65:29 that are efficient to use, easy to set up. Traditionally, real time was very difficult to get going., we had Chris Schmidt on the podcast a while ago and he was talking about a lot of real-time things. Awesome. Great conversation., so I think real time is something you may want to really invest and look at and and I don’t want to just say, “Oh yeah, just go add real time because I don’t think it makes sense for everyone.” What makes sense though is think about when you can get the data and think about what decisions you need to make on that data. And when you have

65:59 to make on that data. And when you have a really long lag or lead time between data and decision- making, think about do I need to make the decision faster? Do I need to automate the decision? Do I need to inform people about a decision faster? faster? That’s where I think you can add value. And I let’s let’s take all of fabric and PowerBI away. Let’s remove all this stuff. Tommy, if we just focused on that part. I have information and I need to make

66:31 I have information and I need to make decision. The only thing I care about is the time it takes to go from data to getting someone to make the decision about whatever the the solution is. Right? This is this is decision-making 101. This is this is why business intelligence exists. So look at your business intelligence and figure out how do you shorten where some stuff is fine. Some things take a long time and it doesn’t matter and making a decision faster isn’t going to help your business. In other situations,

67:01 help your business. In other situations, that’s going to be a huge difference. That’s going to be the difference between a sale and a non-sale. What can you do to shorten that decision point? And I think that’s where I would invest my time. That’s where we’re going to see the next three years, three years move is we’re going to start seeing the ability for you to shrink the time between get data and make decision. Focus your attention there. That could apply to both PowerBI alone. That could also apply to PowerBI and fabric as a combined product. And I think that’s where we’re going, dude. I love that. I love that. So hopefully we’ve answered your

67:32 hopefully we’ve answered your question. You can start implementing it. Also, do you really need 300 KPIs? That was one thing I wanted to mention too. [laughter] A little Yeah. is can we maybe start there may maybe maybe consolidation of K maybe there’s a couple KPIs that drive your business and you could focus on a little a little bit less. Yeah, I would agree that one too, Tommy. Microsoft had a very large number and they whittleled it down to like a lower number. I can’t remember the numbers, Tommy. What was it? it? 35 35 35 core KPI. But what do they have like over 100 150 KPIs? KPIs? They had a lot. And this is for all of Microsoft, not just for

68:03 Microsoft, not just for correct. They’re trying to distill down their entire business to 35 or less. That was that’s probably really hard. Honestly, Tommy, it’s very difficult to measure your business with less stuff, but if you measure it right, you’ll drive the correct actions. So, I I agree with that, Tommy. That comment, Tommy. Having less metrics would probably be a good thing. All right, that being said, thank you very much for listening to the podcast. I hope you found this conversation informative., also, I’m gonna throw this out here again. If you are coming to Microsoft Fabric Conference, make sure you find me. I’ll

68:33 Conference, make sure you find me. I’ll be wearing one of these PowerBI shirts throughout the conference. Make sure you find me. Come get a picture that will enter you in for a raffle. I will every day, every evening, I will go pick someone from the social media thing. You have to tag it., we’ll put a hashtag on it, something like that., but make sure you find me, get a picture, and if you get a picture, I will be giving away these shirts at the conference. Come find me, say hello. I’d love to meet you., and then we’ll post some of these on social media and we’ll give out some shirts at the conference. So, Tommy, thank you so

69:03 conference. So, Tommy, thank you so much. I appreciate this episode. I thought this was a really good thinking one., great question today. And that being said, Tommy, where else can you find the podcast? You can find the podcast on Apple, Spotify, or wherever you get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. Do you have a question, idea, or topic that you want us to talk about a future episode? Head over to powerbi. mpodcast. Leave your name. Enter a great question. And finally, join us live every Tuesday and Thursday, a. m. Central on all

69:33 and Thursday, a. m. Central on all PowerBI. tips tips, social media channels. channels. Thank you all so much and I’ll see you in Atlanta. Measures pump it up. Tommy [music] and Mike lighting up the sky. Dance to the day to laugh in the mix. Fabric and I get your feels. Explicit measures. Drop the beat now. Pumpkins [music] feel the crowd. 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.