AI Driving Your CoE - Ep.529

AI can strengthen a Center of Excellence, but only when it is applied to clear business goals, known pain points, and a defined maturity baseline. In this episode, Mike and Tommy frame AI as an accelerator for documentation, knowledge sharing, and repeatable processes—not a shortcut for missing strategy, weak sponsorship, or an undefined culture.

News & Announcements

• Resource Profiles in Microsoft Fabric Data Engineering (Preview) — Microsoft introduced Resource Profiles in Fabric Data Engineering as workload-aware Spark presets that let teams choose performance patterns instead of hand-tuning dozens of settings. The preview includes profiles for write-heavy workloads, read-heavy Spark workloads, and read-heavy Power BI or SQL consumption on Delta tables. It matters because it gives Fabric teams better defaults and a more approachable path to performance tuning.

• Why Fabric Data Warehouse is the Modernization Path for Synapse Dedicated SQL Pool Customers — Microsoft positions Fabric Data Warehouse as the forward-looking modernization path for Synapse dedicated SQL pool customers while still supporting existing Synapse investments. The post highlights non-disruptive scaling, fewer operational limits, unified governance, and tighter integration with OneLake, open formats, and AI experiences. For Power BI and Fabric practitioners, it is another clear signal about where Microsoft expects enterprise data warehouse strategy to move next.

Main Discussion

Topic: AI Driving Your CoE

Mike and Tommy use this episode to separate the hype from the helpful work. Their core argument is that AI can absolutely make a COE more effective, but only if the team already understands its mission, maturity, and success criteria.

• AI should enhance a COE, not replace the people, process, and governance that make a Center of Excellence effective in the first place.
• Before applying AI, teams need a clear baseline for maturity and goals, and the hosts repeatedly point listeners back to the Fabric Adoption Roadmap as the right place to start.
• A major theme is that AI can make a COE bigger very quickly without making it better, especially when it produces generic content that no one actually uses.
• Some of the best early use cases are practical and unglamorous: documenting semantic models, describing columns and measures, tracing lineage, and organizing institutional knowledge.
• The hosts also call out meeting summarization and requirements drafting as high-value workflows because agents are good at condensing long conversations into a usable starting point.
• More advanced AI scenarios only work when the agents have real business context, language, and process knowledge rather than generic prompts and abstract best practices.
• They are equally clear about the limits: AI cannot solve missing executive sponsorship, weak ownership, or poor organizational alignment inside the business.
• The right way to judge success is with measurable KPIs such as adoption, support efficiency, self-service effectiveness, and other signals that prove the COE improved outcomes.

Looking Forward

Teams that pair AI with a clear maturity model, focused KPIs, and strong business context will be far more likely to build a COE that is both scalable and genuinely useful.

Episode Transcript

0:01 Explicit measures, pump [music] it up, be a be a hype. Tommy and Mike lighting up the sky. Dance to the day, the laughs in the mix. Fabricating AI, get your fix. [music] Explicit measures, drop the beat now. Pop your skins, feel the crown. [music] Explicit measures [music] [music] Drop it Drop it loud. Tommy, good morning and welcome back to the Explicit Measures podcast. We are back again.

0:31 back again. How are things going for you, Tommy? Well, I We’re on 529 episodes. It’s right near the summer, man, and I’m excited. So, I’m going to give another kid update real quick because I encourage all parents for the summer, I’m writing my summer contract with them. I I’ve mentioned this before. Summer contract, okay, yes. Okay. So, and I highly encourage this. So, what I do is do is and I got actually got it from Derek Jeter, from his parents, but it works. works. Because you want them to

1:01 Because you want them to continue their education, also continue the whole critical thinking, but if they’re like me, if someone just tells them to do something because they have to do it, doesn’t jive, right? They have to own it. They need to have that ownership. That’s at least my personality. Sure. So, what I did yesterday is I have this contract saying they have daily and summer projects, things they have to do in the beginning of the day before they want to play, and then things that will be through the course of the summer. Okay. Okay. So, yesterday I sat them and then if they exceed those expectations, there’s

1:32 they exceed those expectations, there’s like lunch with dad, go to the tar the dollar store, whatever the case may be. Nice. And so, but I want the point, the goal, the real secret is Okay. I sit them down them down and I say, “I want you to review this. You have a day to review this. Write down what you want to revise, you disagree with, what you want to add.” Now, Now, I have the leverage, so I have the final say here. say here. However, I’m open to negotiation. I am open to discussion. So, come back to me, write down, and then we will sign find

2:03 write down, and then we will sign find because once you sign, you own it. That’s awesome. It It works so well. too Do you Do you print them out and put them somewhere visible for everyone like on the fridge or somewhere that’s ever visible for everyone to see? I make a copy. So, they have it in their room and I have it in my room. We both sign it. So And And then they cut their finger and then they have to write their name in blood on the bottom of the We’ll do the Yeah, we do it in the or at the silence in Italian. Yeah, exactly. [laughter] The Godfather starts playing in the background.

2:34 background. [laughter] Here’s a glass of wine. That’s exactly That’s great. I have tonight something in my bed. Yeah, it’s [laughter] chores this morning. You forgot your quick tables. So, there’s a horse your quick tables. That’s that’s funny. That’s good. I’m I’m very encouraged by that. Yeah, it’s It’s also you you find to think your They are a good example of entropy. is what I will say. [clears throat] This They seem to They seem to move

3:05 This They seem to They seem to move to the lowest amount of energy state that they can be in. And sometimes you have to push them a bit more to like get those chores, get the work done,, those things. You can’t do the fun things until some of the chores and other actions get done and accomplished as well. Yeah, and that’s why I keep telling them like my oldest who’s going to be 10 this year I said or And I said, “Listen, you have the science project. You can pick whatever you want. Do whatever you want. You want to do data collection, you want to do something,, biological. Don’t care. It’s up to you. However, I’m

3:35 care. It’s up to you. However, I’m counting on you to keep it up because I’m not going to tell you every day to do the project. You are have to complete it by the course of the summer. But I want you to be able to own that.” And I think it’s that but that’s fair. Like again, a lot of this I feel like is exciting expectations. My my daughter is now helping to babysit. She’s gotten to the age where she’s gotten her her life Oh, is it the Red Cross babysitters certification thing you go to a little class and stuff like that. So you you class and stuff like that. So you, you can babysit things. and she

4:08 know, you can babysit things. and she has been watching little kids and they’re little very, four, five, six age kids and she’s like they just don’t listen. And she’s like I tell them to do something and they don’t do it. And like I don’t know like okay, let’s So this is funny for me because I’m like teaching her like how we like brought her up. Like well, okay, before before they do the bad thing that helps us set expectations and it helps us set okay, well, when they do are doing something you say don’t do set the requirements. And and so make sure they understand that if they continue to act

4:39 understand that if they continue to act this way, there are consequences and here’s what the consequence is. You will have to sit in this chair. We have to take away that toy. We can’t do this thing outside. Like make it understandable and you get there and she’s like I have to keep saying it over and over again. I’m like I feel like you’re getting it back. I feel I feel [laughter] I feel like I’m vindicated. Like I thought your first reaction would have been like don’t listen. [clears throat] [laughter] Isn’t that frustrating? Yeah, yeah, can be maddening. WHITE HAIRS EVEN. [laughter]

5:10 BY the way, you just said the most parenting consulting thing I’ve ever heard. What was that? Well, you have to set the requirements. You might as well just said if it’s on the scope, you can’t do it. It’s can’t it’s not in scope. No, you cannot have another brownie. Not in scope. in scope. [laughter] when when consultants turn dads, I guess it would be. [laughter] That’s hilarious. Awesome. All right, well, enough of the fun stuff here. So our main topic today is talking about where can AI run your center of

5:41 about where can AI run your center of excellence? Does AI run it? Do you have AI helping you and it supporting you with it? Does this does this work for us? I think with AI being so pervasive now across multiple organization and people getting more experience to it,, I will share some experiences that we’re looking at and how we’re leveraging AI to either assist, support, or continue to aid what’s going on with your center of excellence. So, a a joint of two topics here. We talked a lot about center of excellence in the past. I’ve had some pretty good conversations with some leaders in Microsoft. And one

6:11 with some leaders in Microsoft. And one of the most important things for a Power BI deployment success This is coming from Microsoft when they have worked with medium, small, large customers is the establishment of the center of excellence. Having a center of excellence really cardifies or or solidifies the ability for this the organization to run well with Power BI. And I think maybe even more so, this even becomes more important now when you start talking about data engineering, the fabric experiences that are coming

6:41 the fabric experiences that are coming out as well. So, we need some more organization around this. anyways, good conversation today. Tommy, you’ve got a couple news items for us. Take us through some news. Yeah, the news articles are from Fabric and I Mike, both articles have a theme and it’s the changing of times. The winds are a changing. And I think we’re finally seeing that’s here. So, the first one is around resource profiles in Microsoft Fabric data engineering and this is a new feature, a not nearly a product, but a new feature

7:11 not nearly a product, but a new feature in data engineering that per provides preconfigured workloads around Spark compute profiles. They remove the need for manual Spark tuning, letting teams choose a profile aligned with their workload. Are they doing a lot of write heavy? Is it a lot of read heavy for Spark? is it read heavy for Power BI? And they’re all available in how the data is read, written, consumed. So, they really actually replace a lot of the environment, a dozen of manual Spark configurations. And right now there are

7:42 configurations. And right now there are three pre-configured profiles. Write heavy, read heavy for Spark, read heavy for Power BI. So, this is, also too, one of the things they really focus on here is these profiles are built for the medallion architecture, Mike. So, for ingestion, transformation, and consumption. So, I love these. These are really good. One of the things with Spark that is interesting is there’s [snorts] a lot of knobs you can adjust. adjust. A lot A lot of things that can can work

8:13 A lot A lot of things that can can work on things. And I think this is one I I’ve actually talked to Santosh about this in the past, which is like Microsoft already knows. Like, you guys have have the patterns. You have the usage. You have,, every single job that we should be that I would ever want to run has probably been run in the Microsoft’s system somewhere, and you know your team is studying this., small files need this compute, big files need this compute, you big files need this compute,, know, bronze, silver, let’s optimize for read write or read or writing and and then let’s optimize for read later on

8:44 then let’s optimize for read later on down in the gold layer. Like, these patterns are like we just need to generally get the gist of like what’s the best the best section, where do these compute designs need to be applied? and I think this is really really useful here cuz these profiles, one, we can pick them from Microsoft that has already built them, but then you can tweak them for your resources and your projects as well. I think this is just super amazing. And this goes really back to Mike, this big thing we’ve talked about around who’s Fabric for, right? And this is more to me one of those

9:15 more to me one of those sublime messages around the fact that it’s for the business. And it’s really going to be more for if you don’t have to be an engineer to do data engineering things. And not just as an okay, but as an encouragement. This is more and more of what Microsoft’s pushing for for people and for businesses. I I think this is another one of these good trends., this is probably maybe the same thing that was going on with like Excel and Power Query in the beginning. Right, we don’t have a good way of making it easy for users to just

9:46 way of making it easy for users to just show up and do things. This is like a a couple of these knobs that like again, there’s hundreds, thousands of properties, I don’t know, probably hundreds of properties under the Spark hood that you can go adjust and tweak and adjust and modify. So, one of the examples here that I like to use is when you’re using V-order sorting. So, V-ordering on your Delta tables packs them in a way that’s good for a semantic model. But, that doesn’t necessarily it spends a little bit more compute every time you’re doing the V-ordering cuz it’s you’re you’re you’re shifting and

10:16 you’re you’re you’re shifting and shuffling the data around so it’s actually optimized for how Power BI could read the files. Okay, great. But, if you’re doing bronze, silver, gold architectures, most of my, I would argue all of them, but most will almost always pull your semantic models from the gold layer. The gold layer is what that does. And so, having this general rule around, “Hey, that’s where we do the read-heavy operations. We do the V-order sorting on the gold tables.” And that means the bronze and silver

10:47 And that means the bronze and silver can actually be optimized for read and write or write basically more write optimization, which then makes it go faster. You don’t have to do the the the V-ordering and it will save you compute by not doing that earlier in the in the process. I would say one thing that you mentioned about Power Query and how they had a help for business users. I think the fundamental difference though is Power Query was really built for the business. That wasn’t built for data engineers. That was built for whatever department you’re in.

11:17 That’s a good point, Tommy, and maybe I should rephrase my statement there. Yeah. Yeah. It Power Query was a design for something that we already did, Mhm. which was complicated. Right? It Like there was a [clears throat] Let me say this way. There’s already a

11:31 Let me say this way. There’s already a language under the hood. It was doing data transformations. It was moving and shaping. So, someone said, I think it’s worthwhile for us to find out the most common data transformations and give you a user interface that can do about 80% of your data transformations. Before that step, it was like other programs like Talend or other like, you programs like Talend or other like,, more heavy data engineering know, more heavy data engineering programs. Yeah. So, to your So, to your point, Tommy, like I I get it, but but I also feel like the only reason like Power Query showed up from a very

12:01 Power Query showed up from a very technical world that made it very easy business users to do similar activities to what we’re doing inside data engineering Yeah. Yeah. space. But, we didn’t need to write SQL. I didn’t need to have a SQL server. I didn’t need to go install Access. Right. Right. Right. That’s a lot of what I think I did previously was like, hey, if I need if I need a substantial database, I’m going to I’m dropping it into Access. Yeah. Yeah. Yeah. Right. And that’s that’s a bit more technical than your average business user, I would say. Oh, 100%. But, with Spark,, you’re not falling

12:33 with Spark,, you’re not falling into Spark, right? A normal person before Fabric wasn’t like, hey, I just heard about the Spark thing and I’m in marketing. Power Query was available,, you might discover it. But, you’re not going to discover Spark and then fine-tune yourself. And I think this is a great packaged version for what people need to do. I I’d even go I’d even go farther than that, Tommy. I think even a lot of people now today in this era still don’t understand Power Query is there. I still Yeah, yeah. I still do classes and teach things around like Yeah. Let me Let me teach you about

13:03 Yeah. Let me Let me teach you about Power BI Desktop. Here we go. Let’s go through the whole thing. And I’m saying And I I I I have a specific I think you do this, too, Tommy, as well. We have a specific moment in time we’re saying, I’m going to show you how to bring this data together and shape it inside Power Query inside Power BI. This is the same Power Query that runs inside Excel. And then I show I literally go back to Excel and show them the same thing. And like that I copy some M code. And like, look, this is the language that they use. Let me copy it over. Look, it just works.” And then I’ll go into Excel and say, “All the work that you’re doing in

13:33 “All the work that you’re doing in Excel, that you’re loading data, moving things through, stop touching the raw files you’re manipulating the files, the extracts, the whatever you’re doing. Instead, [clears throat] focus on using Power Query in Excel, and that way if you ever want to go to Power BI, there’s a one-button import, drop the queries right over. So, and now you get the data tables rebuilt inside the the Power BI Desktop application. So, people are like, “Oh, wow, this is so And then but even today, people don’t see it. They don’t see it right now. So, I I will save this for I’ll save this

14:04 I I will save this for I’ll save this for another episode, but I had a ton of training in this winter all around Excel for users, and some of them, based on the on the the course, was ending with Power Query. But we were focusing They were not a Power BI shop. They were still using Excel. So, it raise It does raise a question the way you’re saying that do you teach Excel and Power Query first, and then introduce Power BI, or do you push Power BI first? But I I think we’ll parking lot that Yeah,

14:34 I think we’ll parking lot that Yeah, that’s probably Yeah. I think it’s a good conversation. I would generally say you just meet people where they are. Yeah. You try and get a very decent assessment as to where people are spending their time, and you try to cater that story to what what can make their time more effective. And then I think you get people on the hook and like, “Okay, now we understand.” And then it it helps add value, or save some time at at minimum. Yeah. Yeah. All right. So, the next one, my friend, this is where the times are changing.

15:04 this is where the times are changing. Finally, this is really the first time that we are hearing from Microsoft about the end of Synapse and the end of really Azure, or at least the path forward. We haven’t really heard it explicitly or directly, but this new other article came out is why Fabric Data Warehouse is the modernization path for Synapse dedicated SQL pool customers. So, this is the first time like that we’re actually hearing yeah, if you’re doing Synapse now, the modern

15:36 if you’re doing Synapse now, the modern way is Fabric. Basically saying that Synapse is almost becoming the legacy program, just like SSIS was. And which is really again, something we’ve talked about, about, but it’s never been actually stated. So, some of the things that they’re actually saying here is is positioned as a modernization path for organizations, solving long-standing performance, scaling, architectural challenges. They even brought up the original design like slide from I

16:06 original design like slide from I think it was Microsoft Build, which is coming up from 2023. Is their first slide slide changed a little, saying that yeah, this is the unified platform for your data. And again, compared to Synapse, there is not disruptive, instant scaling, you don’t have to worry about pause or resume, there’s no hard limits. limits. It’s really meant for data distribution and again, autonomous workload management. Everything’s in one lake, Copilot for AI assistance, and then they say centralized governance.

16:37 then they say centralized governance. So, this is to me is a semi-big day to me because I think this is a is a this is big to me because of the fact that they’re saying it. Not because something new came out, but the fact that we’re finally hearing and we can actually really say, not just that we think it’s going that way, but more that yeah, Microsoft’s saying if you’re using Synapse, Synapse, if you want to move forward, it’s through Fabric. I think yes, I agree. And I think the fact that Tommy they’re announcing more of this positioning. I

17:08 announcing more of this positioning. I announcing more of this positioning., mean, we’ve known we can always observe as people ask me all the time, is is is this XYZ product dead? Is Microsoft going to add any more? Do anything more than I do? I’m like, I don’t know anything more than anyone else does. But what I can say is, look at what they’re doing. Look at what types of features are coming out, who’s making the feature, what’s changing. It’s those kinds of activities that I look at and go, hey, let’s observe what Microsoft is building. What what are they creating at

17:38 building. What what are they creating at this moment? Yeah. And where the new features are coming from is likely where their attention will go for creating net new features. And so, I think Synapse was on this evolution of, the the modernization of the data platform. They had the SQL data warehouse thing and it and initially it came out really expensive. And so, Oh, yeah. Oh, yeah. a lot of hesitation I think initially at Synapse because it was the expense part, but then they got that figured out and it became more reasonable and it wasn’t that bad. And I honestly, I built a

18:08 that bad. And I honestly, I built a number of solutions on top of Synapse. it was really really good. The part that I really liked about Synapse for me was the ability to have a blob storage tables tables being able to access them directly through SQL or having some Synapse was able to bring the SQL Server portion of I need to access Delta tables from some system. So, I could do some data engineering with Data Bricks and then I could land it in storage location and then I could then bring that through Synapse over to Power BI through the SQL endpoint. That was

18:40 BI through the SQL endpoint. That was That was the aha moment for me. Once they moved that system over into Fabric with the SQL data warehouse and all these other like SQL analytics endpoint, the One Lake. So, once they tightened the integration between those systems, I didn’t really need Synapse anymore for me. and then also, Synapse was this opportunity you could go buy Spark. And then Spark was a part of your, you could write notebooks. But it always felt behind compared to Data Bricks, this the Synapse world of things. So, if you could do it, you could get it done, but there’s a couple additional pieces

19:10 but there’s a couple additional pieces of friction. So, I think you’re right, Tommy. I think they’re they’re no longer committing here like there’s no net new investments coming through the Synapse world. And I think this article is also trying to highlight most of the features you needed in in Synapse are now equivalent features are now based inside Fabric. Yeah, that’s a huge point here. And it’s funny because you talked we just talked about the first article which was very much that, “Hey, it’s bundled. It’s easy to do.” And this article is now for the data engineers to read, right? Like

19:41 the data engineers to read, right? Like that that the audience here is someone who’s using Synapse to go, “Oh, we, if we want to move forward, the next step is Fabric. It’s not an upgrade in Synapse or in Azure really.” And I think this is the story here. Halfway down the article it talks about actually a little bit more than halfway. It says, “The modern modern innovation story or modernization journey from Fabric to Synapse. Fabric should be viewed as the progressive modernization, not a disruptive exercise.” So, I I think this is really true. a lot of the things that you would be

20:11 lot of the things that you would be doing in Synapse can now be done in Fabric. It’s a lot easier to get things turned on and moving and integrated. maybe people will have some integration systems where you could say,, go look at all the stuff I have in Synapse and then pull them over directly cuz almost everything that you need on the Synapse side is able to be rebuilt again. You have pipelines, you have notebooks, you have the SQL analytics endpoint. All that stuff exists inside Fabric, which is wonderful. I absolutely love it. So, I I like where we’re going here and we’ve built coming up, right? In 2

20:41 we’ve built coming up, right? In 2 weeks? Build will be in 2 weeks and I’ll be at Build. So, if you are going to go to Build, make sure you say hello. I’ll be doing a a a lightning talk and a table talk. And all of my talks will be around building agentic experiences on top of Fabric. So, using how do you create things with agentic experiences right inside Fabric? what does that look like and how is that changing what we develop and build? So, the a lot of my talk will to I’ll also be at the experts area helping out answer questions, talking through people in the community, helping design

21:12 people in the community, helping design or or think through solutions that people are are trying to pull together. Anyways, I’ll be there. Build’s going to be exciting. I’m looking forward to it. There should be some again Microsoft Build’s a big event. There should be some really cool announcements. So, I’m looking forward to hearing those from Microsoft. Well, you have to promise me one thing, Mike, when you give your talk. Mhm. Okay? Or or I’m not going to listen to it. You have to wear that sweater. Oh, filters make it the model? Yes. This is what I call Right for the podcast. I I don’t have a I don’t have a a session that I’m going

21:42 I don’t have a a session that I’m going to be recorded on, so I can wear whatever I want. No one will No one will see it except the people who are in my session. Yeah. [laughter] Well, what does that mean? Funny you should ask. Funny you should ask about that. Have you heard about us semantic model? Yeah, have you heard about the podcast? Yes, exactly. so. so. Oh, it’s true. Awesome. Well, those are our news items for today. Let’s jump into our main topic today. Tommy, frame us up here for the main topic here. AI is driving your COE. So, what should we talk about here? What does this look like?,

22:12 What does this look like?, we have talked about we’ve done a series on implementation planning. We’ve done a series on governance. We’ve done a series on adoption over the course of the podcast. And all that has been such a big focus on three things, Mike, people, process, technology. Very much a calling card of the podcast and I think you and I also the work that we do. The knowledge center, the center of excellence is a huge part of that. It’s integral part of a pillar of anything that’s successful. But it’s hard to do. It’s hard to keep up with. It’s hard to manage it because

22:43 up with. It’s hard to manage it because it’s dedicated You need that dedicated time, resources, and talent. But we are moving to this part where we’ve talked about AI and agentic solutions to be able to write, deliver content content to to Should you? My question, I think, for the conversation here is should you rely

23:02 the conversation here is should you rely or really allow agentic solutions deliver the bulk of the content to deliver the bulk of the architecture and the structure of what you want your center of excellence to be. Should we allow it to? Is that the right question though, Tommy? I’m curious. What what How would you then frame the question here? Well, I’m not maybe maybe I’m thinking the answer isn’t should we, but maybe the answer is more about

23:34 but maybe the answer is more about can we afford not to? I think at this Like for me, so maybe maybe the question is Yeah. is Yeah. Well, I don’t know. So, I I don’t know if we’re hitting that tipping point of like if you’re not playing the game, you’re out of the game, right? So, I’m trying to think of like an analogy here that would work well for this one or how I’m how I’m visualizing this one in my mind. the how Tommy, I’m not a runner, but I do know some things about running. My son runs a

24:04 some things about running. My son runs a little bit for the cross country. and one of the things that we found that is a big impact on how well you run whether it’s mental, whether it’s I don’t know what it is, but the shoes you wear make a pretty big difference when you’re running. So, if you’re running cross country, they like to wear spikes. Shoes that have little little spikes at the bottom of them that like so cuz cross country you’re across Okay. So, Okay. So, if you could run in sandals, you could run barefoot, you could run in just

24:34 run barefoot, you could run in just regular sneakers, right? But, if you’re going 5 mi, the weight of the shoe matters, the grip of the shoe matters, is it comfortable? Will it be able to to to flow and bend as you run that that race? So, So, as you get closer to higher performance, you find that the gear you wea- you wear matters. And And this probably translates to any sport that you’re in, right? There’s There’s gear that you’re using that that matters. You could play the sport without the the advanced gear, but with it

25:05 advanced gear, but with it you you perform better. Let me just pause there. You see where I’m going with this? Yeah, and I have a few thoughts. Do you have Do you have anything else cuz I have a few comments right off the bat. Okay. So, let me Let me cap the idea of the thought here. So, So, I don’t think AI is going to just do your job for you and just replace you. I I think of it as a [snorts] [snorts] [laughter] Well, this may not go over so well. It’s like your performance-enhancing drug for business users. [laughter] Maybe Maybe that’s not a good analogy,

25:35 Maybe Maybe that’s not a good analogy, but it’s like buying better shoes because because it’s giving you better performance in areas you’re already strong in, right? You’ve already been training for doing this stuff. You’re already working on these things, but,, if you could shave off another couple minutes off your time, that pushes you higher up in the rankings, right? Well, what if you could use a tool What if I could give you a tool that could aid your thinking, help you think about things deeper, and start shaving off wasted time in your day, giving you more ability to do things. and so, I

26:07 more ability to do things. and so, I I look at AI doing those things. In lieu of that, you now have to apply that to I think the center of excellence. Where What are you What are the things that you’re doing in the center of excellence? What kinds of things are you documenting? How do you want to distribute and share information? Can you leverage AI to support you in maturing, producing If I needed Let’s say I have a larger organization, Tommy, and we feel like we need two people to run the center of excellence. Could we just throw AI at it and just let one person own and run it?

26:39 let one person own and run it? Would that save us some some resources there? So, that’s maybe where I’m thinking here is this is a an enhancement piece to what you’re already doing. doing. What’s your reaction? I I see where you’re going with the analogy, but there’s a few things I feel like I want to caution against that mentality. Right? So, I I see where you’re going with that, and there’s a lot that I agree with, but when you’re talking about a race, you’re talking about competition. It’s you against someone else. Someone has to win.

27:09 has to win. And I think when you think about the center of excellence and the knowledge center, too, Mhm. like that’s usually not the highest priority for an organization or one of the things that they have to innovate on. And again, think about the majority of organizations, that’s not the primary focus. It is not also a urgent need. Go back to your running example where there’s that,, the will to win. We’re going to do whatever we can because we have the race on Saturday or we need to be the best in the business.

27:40 we need to be the best in the business. It’s interesting. know any businesses that think that way about their center of excellence. Like, we need to be best center of excellence around. Most organizations was taking it, though. Okay, yeah, I I understand. I understand. Most organizations, I’m going with the mentality of how organizations perceive their center of excellence. That’s where I’m trying to take this. Not so much the competition side. A perception of the center of excellence is can we just I honestly, we’re happy if we just have one, even if it’s not perfect. Like, because it’s so hard to maintain. And it’s not the one like, we

28:10 maintain. And it’s not the one like, we need to innovate here. How are we going to this year make our center of excellence better? How are we going to,, be better than we were last year with the center of excellence? That question’s not asked a lot. It’s not asked a ton. Again, so let me before I go on, let me ask you, what’s normally the perception of a team’s or an organization’s center of excellence? How do they perceive it? What is the value add that they see? Because I think that goes a lot into how they’re going to invest in technology for it.

28:41 to invest in technology for it. Hm. This is a really interesting comment, Tommy. So, I I don’t know if I went there with what you said about we have to win the race. I don’t think that I don’t think center of excellence is So, winning a race assumes that there’s some end objective. There’s some like,, goal in mind, right? to your point, I think a lot of companies don’t understand what the goals are of the center of excellence. They just have something that they know they need some governance. Someone needs to report to something about what they’re doing, what they’re working on, how they’re producting how they’re producing things. I think a lot of times

29:12 producing things. I think a lot of times the center of excellence is a knowledge library of,, “Hey, we have problems in the business. We have this central place where we’re going to put documents and knowledge there.” So, to some degree, like the center of excellence serves those purposes. But, I think when I think about the race, Tommy, the way you thought about and you described the race tells me you’re not a runner. Because Because [laughter] Clearly, I There’s a lot of other There’s a lot of other things you do. Right, running is not one of them or running cross country is one

29:42 running cross country is one one of them, too. So, yes. Well, that is that is part of cross country. There is a There is a carb crash. Oh, yeah. Every Before every big race, like a day like the day before, they’re like they go to someone’s house and they have like spaghetti like crazy. Like everyone eats a ton of spaghetti. Like as much as you That’s just Sunday. That’s every day for Tommy. It’s every day is a carb crash. so the reason I I say that your your perspective is a bit different here because when I looked at the race, it’s not about it There is a goal. But, in the race, you’re it’s more of a self

30:13 the race, you’re it’s more of a self competition. Mhm. It And a lot of racing or running is what did I run last time and how can I run that race again, but do it faster? How can I best my time before? It’s not necessarily, you always want to like,, push harder, go further, do more with what you have. But, I think it’s this idea of resource allocation. Right? In the race, you’re trying to allocate your energy in a way that lets you get through the race faster. That requires training, it requires exercise, it’s part of the gear that you have. So,

30:44 it’s part of the gear that you have. So, those things compile, they compound, and what it does is it gives you a better output. And I think of the the output isn’t really I need to win this, it’s always about and this is where I think the center of excellence does make sense in the analogy, which is you need to understand what your baseline is. And then you need to say, “What is the next stage of what I can get to?” Because you’re not going to go to a race and knock off 5 minutes in one race. It just doesn’t happen. You’re going to go to a race, you’re going to knock off 30 seconds, you’re going to knock off maybe a minute. Like

31:14 going to knock off maybe a minute. Like if you had a good race versus a not good race. Some days you’ll falter and the race will be like you just felt bad, you had some flu or cold, and you’re just behind on your times, you’re not getting a personal best. Right? So, as you think about this, it’s it’s about pushing yourself to do more. And I want to emphasize that comment. Yeah. And and what you said, Tommy, which is center of excellence don’t know like what’s the baseline, what should they be shooting for. And I think here Correct. This is I think a maturity step

31:44 Correct. This is I think a maturity step inside the center of excellence. And I think to your point here, Tommy, this is really good because this goes back to the the what is it? It’s the Fabric adoption roadmap. It used to be Power BI and then they changed it to Fabric. The Fabric adoption roadmap now has different meters and gauges that say, “Okay, if you’re doing let’s talk about documentation or,, knowledge sharing or security.” Microsoft has written out a really good blueprint of stage one through five. Where do you fit on those stages?

32:15 Where do you fit on those stages? And I think the analogy here fits. If you’re a stage two in something on this adoption roadmap you don’t have to get to stage five right away. You may that may be your ultimate goal long term, but the goal is just move one level up. Just look to the next step ahead of you. And when you’re running races, it’s always like, “Ah, this is a sucky race. I can’t stand this.” So, you pick up you pick something in the future that you can run to. Okay, I’m going to run hard to that tree. Okay, turn the corner. I’m going to run hard to the top of that hill. And

32:45 to run hard to the top of that hill. And it’s like these goals that you set just far enough out of the way that you run to them, that you push to them. And you just try and do the best to the next step. And I think that’s a better analogy for me on how I perceive like like the center of excellence. And what I’m trying to convey is that’s awesome, but now add to that what can AI help you do to accelerate you to the next step? By the way, I love this analogy here where we’re going with this and because there’s there are a lot of similarities here, or a lot of crossover. So, one

33:16 here, or a lot of crossover. So, one thing I’ll first mention though Okay. is even take the competition with someone else’s side, they’re still you’re always looking for a competitive advantage, right? Even if the shoes, right? Baseline. Got to have some baseline. Yeah, I agree. If I gave you shoes that were lighter, you’re not going to go I don’t want them because that’s going to defeat, like my time I can’t compare my time. You’re going yeah, my time will definitely be better. Yeah. And this goes back to also to the other point not so much the competition, but the investment. So, if you are a dedicated

33:48 investment. So, if you are a dedicated runner, runner, you are going to get the better shoes. You’re going to get the diet, right? The the whatever the resources because you the whatever the resources because it’s going to actually you can know it’s going to actually you can marginally and actually measurably see an impact. Correct. Right? We both will already assume. We will both agree that AI can definitely help your center of excellence. I don’t think there’s really a an argument there or discussion there. Well, we’ll have to go like where do you apply it? Like what makes sense to help with AI? Like where does it where do you apply the AI in a way that helps it make sense? But also

34:19 way that helps it make sense? But also too, it’s too, it’s the thing that point that’s probably more of the argument here is the investment side, right? So, you can tell anyone yeah, AI can definitely help our center of excellence. But

34:31 center of excellence. But I think the questions we really need to ask are how much investment are we doing? How much cost is this is the identity solution is going to be? And then you have to ask too, what marginal impact is that going to make? Like just because I buy the better shoes, am I actually going to see is it going to be a 2-second difference? Is it going to be a minute difference? Right? So, and I think to your point, Mike, I it’s it’s important part here is where do you apply it? So, I think that we have kind apply it? So, I think that we have a few different lanes here is is

35:01 of a few different lanes here is is where do you cut off in terms of investment? If it’s going to be $15, 000 to do, to do, maybe it’s not worth it. And then the other side is where do you apply it so you actually are going to see an impact. One last thing I’ll say here and I want to get your opinion is let’s also make sure that we understand that just because the AI is creating content does not necessarily mean it’s creating impact. Right? Because AI can create all

35:31 impact. Right? Because AI can create all the documentation, right? It can create all the docs, it can create all the instructions. Cool. But is that actually where the impact is? Right? Just because it created content does not necessarily mean it is making your center of excellence better. So, few avenues here. Resources and investment and where do you apply it? I think is where we really need to dive into here. I your couple of reactions on some of the things you said here.

36:04 Quantifying where you want to go. So, first and foremost, I think it looks at you need to know where you are. I think the first point you made which was like what’s the great point? What’s the baseline? What am I comparing to? Like what does good look like? What does performance like so racing in a run running a race is like very under easy to understand like that’s the this time is faster, lower times are better. That’s that’s the measure of success. When you’re in the the the of excellence like what does a measure of success look like? Do you have regular meetings with your team? is it Do we do quarterly center of excellence deep dives? Are How

36:35 center of excellence deep dives? Are How are we measuring success back inside the center of excellence for what What does success look like? So, I think think first and foremost, it’s something around that. And this is where going back to the Fabric Adoption Roadmap is really useful because that’s the blueprint. Those are Look at individually those sections, right? How you educate with training, how do you roll out Power BI, how you roll out things in Fabric, where is a central location for knowledge and information. If you don’t have these things, you can say, “Well, we’re we’re

37:05 things, you can say, “Well, we’re we’re definitely weak in these areas.” We can then then actively invest to move to the next next portion of this. So, that’s So, one thing is I don’t think enough organizations are taking a self lens, an introspective look as to where they are and where they want to become. Where Where they think they’re weak, where they think they’re strong. Because I think when you take a real look at this, you change your behaviors around I need to stop running through the day-to-day of just making stuff get

37:35 the day-to-day of just making stuff get out the door. We actually start building process and slowing down. I have one customer customer who has who has changed their messaging on how they deliver business intelligence. And they said, “We’re going to be making data products. The commitment to the organization we’re making data products for you. Instead of just serving the needs of giving you a report every single time you need one, we’re going to make a product that is about data. And then we’re going to teach and enable your your team to go use these certified, designated,

38:05 to go use these certified, designated, ready to go items. And when you don’t understand it, we’re here to help you support it.” So, that that strategic goal was aligned up to the upper levels of the business, and it’s helping them win. They’re finding value from the data products. People are getting more of their answers met, and they’re building products that are universal or or flexible, meaning they answer many different questions, so that serves a wider audience. That’s the goal, right? It’s actually distributing business intelligence. I’m picking up

38:35 business intelligence. I’m picking up what you’re putting down, man. For and I I’m pretty sure that you and I we’ve talked about this on the podcast even back in back in when we had Seth on the podcast that we talked about data as a product. I know that was an episode. I’m 99% sure, but I think also, and tell me if I’m wrong here, what I’m hearing you also say is I’m hearing that the biggest impact you can have right off the bat with and using agentic solutions is probably in the beginning. And I’ll keep the I’ll keep the analogy

39:05 And I’ll keep the I’ll keep the analogy going here. Sure. And because I think you can equate it to cycling. When I started cycling, I was using aluminum bike,, normal handle wheel, and the time was what it was. But I knew if I invested in heavily in a more cycling bike, it was going to see a marginal difference. Like, I’m at a point now with cycling that if I were to buy a better bike at the same price, it’s not I’m not going to see that significant difference compared to going from that old rickety bike, Yeah. Correct.

39:35 Correct. Your Huffy isn’t with steel frame isn’t going to cut it anymore, right? Right. Yeah, carbon fiber while the investment is going to see this marginal impact and I know that, but I can’t do that again, right? right? Yes. Yeah. And I think for a lot of organizations when you’re starting out, that investment for agentic is great. Like if you already really have one, so agentic is maybe not great. I would say, “No, you’re completely wrong in that thinking.” If anything, if you’re at a bare,, first step of any

40:05 bare,, first step of any agentic solution, then or or or rather COE, then agentic solutions I and I’m going to say it for two reasons. Mhm. For your documentation, which is huge for that knowledge, but also for the project management of it. Sure. huge part here where you you are going to see immediately your impact of what the COE can be. Mhm. So, I’m I like this analogy, Tommy, and I’m I’m I’m actually I’m pulling up the adoption road map, fabric adoption road

40:35 adoption road map, fabric adoption road map, as well. And there’s a lot of really good things in here that are just good baselines. It talks about like executive sponsorship, it talks about business alignments, it talks about content delivery scope, what are you going to supply, how you’re going to supply it. Ooh. Again, yeah, yeah. and then it goes down to the bottom here when you start looking at the bottom of these articles, it starts talking about consideration key actions, but then it talks about this whole section called the maturity levels. And this is the part, right? For each of these sections, content delivery scope, center of excellence, governance, mentoring

41:05 of excellence, governance, mentoring and user enablement, and all these different things that work well together, it talks about the initial, level 100, level 200, repeatable, right? And it it goes through each single section of these engagement areas, 300 is defined, 400 is level capable, 500 is now we’re an efficient team, right? So, these things basically help you outline where do you sit in the business and where,, evaluating against these criterias. So, let’s go back to where the AI part of

41:35 let’s go back to where the AI part of this sits. I think you you made a really good analogy here, Tommy. anywhere you apply AI, machine learning algorithms, all of that stuff, it seems like the most value is added at the beginning part of that project, right? You’re the the initial engagement with it can make a huge difference right out the gate. Now, I do have one maybe word of caution that I’ve seen some places and maybe I’ve seen it in my own world as well a little bit as well.

42:05 well a little bit as well. You have to be just a bit careful about throwing AI at documentation because you want it to be concise enough that it makes sense to the users. Mhm. So, Mhm. So, But you don’t want it to be too verbose that it makes the users skeptical and like overwhelmed with all that you’ve done. Now, this preserves both It’s like a balancing act. You have to know your audience in this in this space. What I do think is really useful, so I’m going to give some So, we’re talking about AI and and using it in our center center of excellence. Let me give you a practical example of where I think that you should use AI.

42:35 you should use AI. Odds are your organization has data models that have been shared with the organization. Sent out. Sent out. Odds are it’s difficult to work with those. And the larger your model is, the more difficult is to work with that model. The more tables you have, giving it to an un , a new [clears throat] user of Power BI, they will be overwhelmed with the 18 fact tables you give them and the 27 dimension tables you give them. And

43:05 dimension tables you give them. And when they grab a column over, it won’t make sense. Like, why can’t I grab these two things over? It doesn’t It doesn’t make sense. Well, I 100% exactly. Correct. So, there there’s this idea of you can throw too much documentation to people and it overwhelms them. Where I think AI makes a lot of sense right now for center of excellence is start looking at some of the automation tasks you need around documenting the model. model. So, does all your columns have a description? Does all your measures have a description? Do you understand the lineage of those columns? How would you go back through and document from the gold layer through

43:37 document from the gold layer through the semantic model into your application where does that column come from? So, there’s a lot of AI that you can throw at those kinds of problems that would just be very slow for you to go through one at a time to go through every single thing to get down the lineage from source system all the way into your model and actually identifying where these columns come from. I was just talking with a client on this recently. There’s this challenge between when I when as a as a business user, I go into Salesforce and I see columns and

44:07 go into Salesforce and I see columns and fields and information that’s there. Mhm. Mhm. I can see the list of all my customers. I know all the addresses and all the contacts. contacts. All of that exists as like lots of little tiny tables in the operational system. system. My mental model of what I interact with to make the data and then what I see in reports have to match. So basically I’m taking all the Salesforce data out. Yeah. I’m doing a whole bunch of transformations to it, taking lots of small tables, merging them together, building customer tables, building like contact lists, all like shipping supplies, all these things. And so what

44:38 supplies, all these things. And so what I’m doing is I’m building a representation of that Salesforce application in now some of my reports. So what you’re bridging mental gap, you’re trying to bridge the operational system and now the reporting system. And if you’re not doing a clear job identifying what is a customer and labeling it in a way that the user knows when I go to the application and select this customer, the same customer appears in my reporting. As much as we like to think about it, sometimes the that BI team is so far

45:08 sometimes the that BI team is so far removed from like the operational system that we lose that information for the for the business user. So this is where I think AI should be applied. Power BI MCP modeling server is super great. Really like that. Using the Power BI fabric MCP server can create items, look at items, scoot around the Power BI ecosystem. That’s another great place to pull AI from and use that while you’re building things. I think I think first you’re going to love some of the

45:39 first you’re going to love some of the things we’re going to say, but I want to just focus very much on the idea of using using agentic solutions around the MCP just for the documentation or at least for, descriptions around your data. I think it’s a big win. I think it’s a huge win. It’s a huge win. Now so not only is it a huge win, Mike, I think you and I need to develop just like there are levels in the fabric adoption road

46:01 are levels in the fabric adoption road map. map. There’s levels two to I think how you’re applying anything agentic around your COE. I would call that level one. just provided that that’s the baseline that you can do. And hear me out here. out here. Mhm. [clears throat] You mentioned some very solid AI-based systems that you’re building. So you keep going. I like where you’re going with this. But you mentioned something very important here about how much,, we don’t want it to be too verbose. You we don’t want,, too how much content is worth,, worth it. If you just throw a single isolated agent

46:35 you just throw a single isolated agent at your COE to build it, it’s not going to be that helpful. It’s going to provide a lot of fluff and a lot of bloat, bloat, but it’s probably not around the culture of your business because there is a level here. There’s a step up from just doing MCP servers and writing your descriptions and yada yada. That’s nice. So where where really a center of excellence is impactful because it knows the business and it knows the culture. And normal agents, if you’re just chat botting it, it’s not going to do that.

47:05 botting it, it’s not going to do that. And that’s why I think there’s a level here. MCP is the first step to me. That’s a baseline of a foundation of what you’re doing. But if you actually want to start really utilizing anything agentic or using this, you have to have a I’m going to call it a CA. You need your own custom agent here because an AI alone is not going to understand your business, understand the people, and you made another great point. BI is

47:35 made another great point. BI is isolated sometimes, too. So you can if if you are just a BI person saying, “Hey chatbot, write us a,, description of our process.” If you don’t have input from the operations team or the agentic solution can’t get that, it’s not helpful. It it is one of those things, Mike, where we deal with is it worth to have something at all if it’s not helpful? Right? Or we should have nothing. If I just have a bunch of documents that no one reads because it doesn’t mean anything, should I have it at all? And I

48:06 anything, should I have it at all? And I think you can fall in that trap with AI. So, So, let me ask you, when you look at providing AI and driving your C C O E, are there levels here? Do you see that as there’s steps up on what you’re doing, the harness that you’re doing it in, in, and the agent that you’re actually applying it to? Ooh, there’s a lot going on here. My first reaction to what you’re describing around MCP servers, so Yeah. there’s like there’s two things that I

48:36 there’s like there’s two things that I think that are very useful here, and I’m seeing a lot of value from both of them. There’s this concept of the MCP server, and then there’s this concept of making a skill. a skill. These are two things that are very very useful to leverage when you’re talking with AI. So, That’s really good, yeah. I want to make a bit of a distinction here, right? A skill is something that is just text document, it explains explains guardrails to an AI of what you want it to do. to do. The MCP servers are physical tools that you’re getting that you can go use in with your agent, but it handles other

49:07 with your agent, but it handles other things like how do I authenticate, what API calls should I make? When I get information back from something, what describes that information around the MCP server? So, there’s a bit more, I think, with the MCP servers. Also, MCP servers right now are potentially a little bit more heavyweight. They add additional context to your models, the large language models, which which call causes them to cost more. I think MCP servers are not going away. They’re going to continue to stay here. They’re going to get more Evolved.

49:37 Evolved. optimized. A little bit more a little bit more like cleaned up cuz they’re they’re there’s a bit egregious on some of the the token usage on your models. but that being said, So, I just wanted to call off the MP MCP server. I think that’s part of what we’re doing here. But, to your point around different levels of what we’re working with agents, I still think a lot of people are just focusing on I go to this website and I chat to it, and it responds to me with an answer. That that’s probably level one of things.

50:07 things. When you start using other tools, like to your point, Tommy, you said the word harness, just understanding what a harness is, it it and being able to just stand back and say, I can point to where a harness is being used in the system. Like now, I’ve been so equipped with this this or thinking through this myself, any I can really point out where the harness begins and where the large language model starts and ends now. I I now feel like I I know a little bit more around that. And there’s a lot of white papers around that says your AI model

50:38 papers around that says your AI model things, whatever you’re doing, the harness is like 90% of how effective the AI model will be. Yeah. Yeah. So, a harness is is the code that wraps it’s a it’s a deterministic bit of code that wraps around this non-deterministic large language model. That’s how I describe it. And people don’t understand deterministic, non-deterministic, literally software that you can say input A gives me input output B every single time. It’s super super reliable. That’s what the harness is doing. The harness is a programming language around the large language model to give

51:09 the large language model to give it guardrails. The MCP tools and the skills are just ways to customize the harness because the harness is only so flexible. It it’s rigid. So, in order to extend the harness to make it more easy for people to leverage, you use these other tools. The further you go into this this world here, right? I see level one being talk to the agent, level two starting to use like a harness like a VS code, not using the standard fabric harness, which is Copilot, like little chat windows you

51:39 is Copilot, like little chat windows you see in there. I think that’s a good basic starting point. but using more complicated harnesses and understanding how they work really adds a lot of capability to you as a user. And now I’m starting to see a new world, Tommy, here, where it’s more about automation with agents, like a team of agents. Yeah. And how do you get them all to work together? And so, a lot of what I’ve been doing is I work with N agent Yeah. to build a specific task. Hey, I want an agent to help me

52:10 task. Hey, I want an agent to help me take this YouTube video down and digest it and give me a summary of what’s going on there. Like, I don’t watch podcast or YouTube videos anymore. I send YouTube videos to my agent, my agent downloads the transcript, it summarizes the information, and then materializes that information for me back to how this is relevant for my business. Yeah. I found this cool video about video editing audio clips or video editing clips with an agent. Interesting. Hey, I call it I I’ve I’ve renamed my bot recently, Dax. So, Dax is my bot. So,

52:40 recently, Dax. So, Dax is my bot. So, hey, Dax, hey, Dax, what research this? And it knows exactly in the skill, it says it’s good on the translate, it does all these things. So, for me, I can throw more articles at this thing, and now it’s creating a knowledge library for me. So, how I’m using this in my center of excellence is I keep throwing in articles, it’s writing up the article, it has weekly summaries of like what’s important, what should I work on, what should I focus on. And there’s so much noise in the system, I’m now using my agent to like tamp down all the noise in the system

53:10 tamp down all the noise in the system and help me focus on things that are really important and can move our business forward., that’s another area. I have other agents building content aggregators. There’s a lot of content being generated across the internet, through Reddit, through LinkedIn, through X. How do you see all that content? Are there things that you want to be aware of that would be useful for your team? Who’s finding value with different things? I’m using agents to bring that information together and aggregate it. I have a Yeah. We’ve talked about content nudge in the past. That’s That was all built by an agent. So, I’m I’m now building tools to

53:40 agent. So, I’m I’m now building tools to aid what I need to do in the day-to-day business. business. Now, it’s still the the tool it produces is very it’s code. It’s it’s software, it’s an app, it runs very regularly, but I use the agent to assist me with that app. What does that look like for your organization? Does that do you build apps with agents to help you with your center of excellence? How do you field calls? How do you field things? Like you can start looking at what are the daily tasks you do in your center of excellence and starting to automate them with agents. you’re talking about we we had a

54:11 you’re talking about we we had a episode about the maturity curve of using AI with fabric as a developer, and that’s exactly what you’re speaking to. I I want to challenge you a bit here, and I I think it’s different when you have a mature that same maturity curve is different when you’re applying that to a center of excellence. And here’s why. why. The higher up you go in the maturity curve for a fabric developer is much more around the technical side, right? It’s around the your technical expertise

54:42 It’s around the your technical expertise and your and your com-, how comfortable you are with using anything agentic, be it MCPs and skills. The way I look at this with a center of excellence, at least the maturity curve, I think it’s the effectiveness of the COE. COE. So, and this is where I’ll make the distinction here. The reason why for me level one is the MCP when it comes to using your AI with your center of excellence is twofold.

55:14 your center of excellence is twofold. First off, no one should ever say you’re using AI in COE because that doesn’t mean anything. When I am actually going to build my center of excellence, again, the output is for a user. Second, it’s a general staple you can do to any semantic model. I’m leveling up with my COE and using anything agentic when more people are involved, and I need more input from the organization. Your COE is part of your culture, which is very different from a

55:44 culture, which is very different from a fabric developer leveling up and using just more technical skills. The higher up you go, the more mature your COE is, the more governance and the more human culture is actually involved in building it and making it effective. The way I look at this, Mike, is I see using skills, but you can apply those skills and have resources from the operations team. Hey operations, we want to make sure that your process is, you to make sure that your process is,, part of the COE. Can you provide

56:14 know, part of the COE. Can you provide us your,, your confluence or your SharePoint page? We want to see how your things are done. We build a skill off of that, right? Or sales, what’s the sales process?, can you show us what’s an opportunity? Give us your definitions. We’re going to feed that in. You need other input outside of yourself, right? You need the organization’s culture and that to influence the agent. Your agent is not just building apps. It needs to know the language, the lingo, the process of the

56:45 language, the lingo, the process of the people at your organization to be effective. And that’s the big difference because again, you can have your agent build a bunch of apps for your COE, but where is it actually effective? Because you want to build, create rather, is probably the better word I’m going to use here with your COE. your COE. You have to create things that are going to be effective for users. Let’s go back to the word culture. We we had a had a an entire episode around this. What

57:15 an entire episode around this. What makes up a culture in general? It’s a common language. It’s a common place, right? I know, for example, good Italian culture, I know where the good pizza places are. Everyone says the the words, there’s a lingo, there’s a commonality. Same thing applies with the data

57:33 Same thing applies with the data culture. culture. And you’re not going to have a great data culture if your COE’s is building general content and building general things. It doesn’t work. I don’t care who’s building it. If it’s general, it’s not going to be something people rely on. So, maybe a bit of a hot take here and and I know I there are a few things said, but I think the main points here is your agentic solutions with the COE has a different maturity than from the technical side. And then two, your agent is needs to rely on the

58:04 your agent is needs to rely on the business more than ever. Do you agree with that? Where are you going with Where are you going with that? I like what you’re going with There’s I think what you’re speaking to Tommy is the AI can also generate a lot of slop. Just very generic repeated things. I’m going to go back to some more real real examples that I like to use around like what where I’m finding it effective and valuable. One of the one of the effective valuable pieces I’m finding with AI is summarizing meetings. So, I’m finding an immense amount of value with So,

58:34 value with So, I think you have to break these tasks down for where do you apply people and where are the people best served to have like that human interaction between me and the business user, me and getting the requirements, me and asking the kind the requirements, me and asking the questions I need to ask, right? of questions I need to ask, right? I think that is something that is extremely required and useful to users. It’s a base level though, right? Like that’s just But the But distilling that conversation down into a message systems requirements, that’s a really good place

59:04 requirements, that’s a really good place where the agents are are very good. So, I have found a lot of use in having conversations with teams, letting the AI record them and make its own notes and summarize them, and then from there I say, “Okay, great. Now, take these notes from this conversation.” I know you’ve done this too, Tommy. Take these notes, build requirements, and then put those requirements back into the semantic model. So, now what we’re doing is we’re putting putting your all your tasks need to be rethought in where does the people get applied and where does the agent get applied? And

59:35 where does the agent get applied? And what are agents good at and what are people good at? Agents are good at summarizing, reasoning about conversations, figuring out what generally was said across a long hour-long conversation around requirements. That’s what agents are good at. Yeah. It gives you a decent answer out, but then you, the person, has to come back in and evaluate, okay, how does that conversation turn into now a new measure, a new model, where were the real pain points, right? The AI is going to pick out some things that are important to you, but you have to materialize that into like real

60:05 materialize that into like real artifacts that the team can use. So, that’s another area where I would say AI is going to assist you. Can I ask you a question there? Yeah, go ahead. ahead. If you see a team who’s not doing that today, today, is there a level of disappointment or an assumption that they don’t know what they’re doing? No, I don’t think so. I think this is also a a function of how much central IT is allowing them to see this. What I’m seeing more than I think teams want to use this. I think teams want to have these tools in place.

60:35 teams want to have these tools in place. so, I had made a couple notes here earlier,, where am I using some MCP servers, right? I’ve been using the devops MCP server. That’s been very useful for me in a project. now, fabric or sorry, not fabric, but there’s now this thing called the work IQ that now shows up. So, there’s a work IQ MCP server, which allows you to go access your emails and go access your mailbox and go access Loop and SharePoint and other documents that are across your experience. So, one of the things that I felt like was really lacking from Microsoft in January,

61:05 lacking from Microsoft in January, February was I have all these agents, but none of them can access any of my Microsoft information. There’s there’s zero story for that. Now, I’m starting to use work IQ to get access to those documents, right? I don’t I want to record the meeting in Teams. Teams. I want it summarized. And I want to just send the Teams link to my agent and say or co-pilot or whatever it is and say, “Take this information, get the transcript, summarize it, put it here.” And then I can work on it. to do from this, yeah. Correct. What Let’s go task this thing out into what we got to go do and build. So, to me, I’m looking at this going,

61:37 So, to me, I’m looking at this going, “Where do I apply the agents in my existing workload?” And it’s not taking away the people, but it’s shifting the work that the people do. And I think there’s this idea of I don’t know I I don’t know if this is going to actually turn out to be a real thing or not, but it it feels like there’s this middle management layer of organizations that are going to be attacked pretty heavily from these agents and AI pieces. AI is going to really affect the it’s going to really impact the

62:08 it’s going to really impact the individual contributor. If you already are an individual contributor, meaning you build models, you do things, you make things go out the door, like you’re directly impacting the business, I think AIs and agents are going to make you just that much more effective and capable. So, the individual contributor of information or or stuff is going to get a really big boost from AI. Middle management layer is if you’re not contributing, I think a lot of organizations are cutting that middle layer and flattening the organization and going down to, “Okay, we just need more individual contributors

62:38 more individual contributors because we can set goals high level and we don’t need that middle management layer as much because the contributors are middle managing their own agents, getting more stuff done.” article, it’s all jazz now. It’s all jazz now. Yeah. I have I have a question final question for you and this is going to be my closing thought. probably wrap. Yeah. How can you prove when you use anything agent if you were to apply an AI solution to your center of excellence, how can you prove it got better and not just bigger?

63:14 Well, I think that’s a great question. I would have to probably maybe focus on initially, Tommy, what are your initial KPIs? What are you measuring? Again, let’s go back to what success looks like. Yeah. Yep. Let’s go back to the fabric adoption roadmap. And I think this is this is a your your your department needs to be a bit of self-aware. Where are we? Who are we? What are we doing? What’s our main objectives? Do we have a real executive sponsor that can support us? If we don’t, then we’ve got other problems to solve. Like no amount of AI is going to not no

63:44 Like no amount of AI is going to not no amount of AI is going to solve a we don’t have an executive a sponsor for our projects, right? You you can’t throw AI at that. That that is an internal politic issue inside the department, right? If you can’t get that centered first, first, don’t don’t even bother with AI. You’ve got other fish to fry first. So, I you’re taking the words out. Re- read read the fabric adoption roadmap. I’ll make sure I I’ll put the link here in the in the chat box here as well. Just that way we have it here. But make sure you read through this. Go through each section.

64:14 through each section. Think about it. Really reflect on what your company is doing, where you fit inside maturity levels, what your your culture looks like, do you have executive sponsorship, what is business alignment, all these things. Content ownership, distribution, community of practice, user support. Think through these things. And once you go through these these, assessments, you’ll have a better picture as to where your business is strong, you don’t you fo- keep that strong, but then where your business is weak in these different areas. And I think if

64:44 these different areas. And I think if you focus on the areas that are weak, you will naturally get better KPIs and measurements out of this, right? for example, right? We’re rolling out Power BI, replacing old systems. We want all the users over by a certain period of time. That’s very measurable. You can get people over to the new system. we’re handling X number of tickets today. What’s the average length of time to to resolve those tickets? How do we measure that? What does that look like? Do we want it to re- be reduced? Do we want more self-service? Think about those things and then put AI and practices in place to help you meet some

65:15 practices in place to help you meet some more of those goals and objectives. Yeah. Yeah. And I’m going to really re-emphasize some of the things that you said or really just focus on it because I think you’re absolutely right. If you are trying to apply AI right now and you do not have the baseline, you do not know where you’re at and more importantly you actually that’s probably the most important part. If you don’t have that and you don’t have goals for the future outside of AI, just in general, don’t do AI yet because you’re not going to be able to see anything impactful and you’re probably just going to make it bigger and you’re going to blow it.

65:47 bigger and you’re going to blow it. You’re not going to provide the impact. So, you have to start that point. It’s almost a trick question. You need to know where you’re at. You need to know how fast you’re running the race before you can say how you’re going to improve that race time. And very much so, if we know we’re at this baseline, what are the goals for your center of excellence already, right? And if you cannot say we’re going to do AI and we’re going to create a goal for it. No, you need to know what are the goals that you already have in place. That is where you going to marginally see it and you

66:17 you going to marginally see it and you you going to marginally see it and that you’re going to make sure that know that you’re going to make sure that you are focused. I was just listening to a Steve Jobs and one of the things he said, “Focusing and project management is about saying no.” no.” And I think a lot of times we’re seeing this never more prevalent in the world of AI, but I think also very hyper to this conversation today. You could throw AI at everything with a center of excellence and you’re going to see results, but whether or not it’s just,, junk and there’s some things you’re having to put a bullet to the head for because it doesn’t do

66:47 the head for because it doesn’t do anything. anything. You have to look at your COE. You have to say, “Where do we need to focus? Where are our biggest pain points? Or are where do we need to focus where are the processes or goals that we have based on our company and our culture?” That’s where we can see AI actually really become a win. I agree with that one. And again, it goes back a lot down to I think this this pattern of applying it where it’s needed, right? And And spending time with it you There’s things that you need to start. This is the same

67:18 that you need to start. This is the same pattern I think we find Tommy in report building, right? It’s easy to create a report. It’s easy to update report, but often we don’t delete them. What When When are they no longer serve their purpose in time, how can we delete the reports and move them off of the system? Let’s take a clear look at our center of excellence or even if we have one or don’t have one, are we doing things effective Are we actually being effective with it? Are we actually using our time well and getting real value from it? If not, then we need to step back and say what of this can be removed? Can we delete some things? And let us focus on new things that’ll be

67:48 let us focus on new things that’ll be more effective for your team. So, awesome. Very good stuff. I love this conversation. This is actually a lot more about center of excellence communication parts than it was AI at this point, but I think there’s some definitely good places to look at. That being said, go check out the Fabric Adoption Roadmap. The link is in the chat as well if you want to go see that as well. If you don’t mind, please make sure you like and subscribe to our channel. We do this every week, Tuesdays and Thursdays at a. m. Central Standard Time. It really helps the algorithms and lets more people know that this content was good and viable to

68:19 that this content was good and viable to you. you. Tommy, where else can you find the podcast? You can find us on Apple, Spotify, wherever you get podcast. Make sure to subscribe and leave a rating. It helps out a ton. Do you have more to say on this topic? Do you want us to say more on it? Head over to powerbi. tips/podcast. Leave your name and a great question. And finally, join us live every Tuesday and Thursday a. m. Central. Join the conversation on all of powerbi. tips social media channels. Thank you all so much and we’ll see you next time. Hey!

68:49 Explicit Measures pump [music] it up PA high. Tommy and Mike light it up the sky. Dance to the data, laugh in the mix. Fabric and AI get [music] your fix. Explicit Measures, drop the beat now. [music] [music] Rock his kicks, 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.