CI/CD Automation with Agents – Ep.534

CI/CD in Fabric gets a lot more interesting once agents enter the workflow. In this episode, Mike and Tommy break down how agents can help with Git-based development, merge conflicts, validation, deployment scripting, and repeatable automation patterns that make Power BI and Fabric teams more effective.

News & Announcements

• Microsoft Fabric Task Flows — Microsoft’s open-source Fabric Task Flows project is designed to take a business problem through guided discovery, architecture design, validation, and CI/CD-ready deployment output. That makes it especially relevant to this episode because Mike describes a “creator agent” pattern where agents generate repeatable Fabric solutions and scripts instead of leaving teams with one-off manual setup.

• Get started with deployment pipelines — Microsoft positions deployment pipelines as Fabric’s built-in application lifecycle management tool for promoting items across environments. Mike and Tommy use that built-in experience as the baseline comparison point before arguing that Git-based workflows can offer more control, review, and automation for advanced teams.

• Get started with variable libraries — Variable libraries centralize and share configuration values across Fabric items so teams can customize behavior without rewriting everything in each environment. In the conversation, Mike highlights them as a practical part of a stronger CI/CD story because they help keep dev, QA, and prod deployments organized once automation enters the picture.

• Matt Pocock’s handoff skill — This compact skill focuses on handing work from one agent session to another without duplicating what already exists in plans, issues, or other artifacts. It fits the episode well because both hosts emphasize that agents become more useful when context is preserved deliberately instead of being rebuilt from scratch every time.

• Fabric Jumpstart — Fabric Jumpstart is an open-source, automated, community-driven catalog of Fabric patterns and deployment accelerators. The hosts call it out as another way to move faster because it gives teams proven starting points instead of forcing every project to begin from a blank slate.

Main Discussion

Topic: CI/CD Automation with Agents

Mike and Tommy frame this episode around a practical question: if Fabric teams are already working with Git, deployment pipelines, and automation, where do agents actually make the process easier? Their answer is that agents are most useful when they reduce the hardest parts of source control and deployment work without removing the need for testing, review, and good engineering habits.

• Deployment pipelines still matter as the built-in, click-driven ALM path inside Fabric, especially for teams that are more comfortable in the UI than in code.
• GitHub and Azure DevOps represent a different level of control because they introduce branching, pull requests, commit history, selective file changes, and richer automation options.
• Agents can dramatically lower the barrier for teams learning Git by helping explain concepts, resolve merge conflicts, restore missing changes, and run the commands that would otherwise feel intimidating.
• The hosts spend significant time on merge conflicts because they are one of the most frustrating parts of collaborative PBIP development. Agents are especially useful here because they can inspect intent across multiple changes and then help validate whether the resulting solution still works.
• Mike argues for using agents as “creator agents” that build deterministic tools, scripts, and actions rather than relying on a vague prompt every single time. That approach creates repeatable automation with clearer inputs, outputs, and testing steps.
• The discussion also highlights that Fabric-specific tooling matters. MCP-style model access, API-driven validation, and model-aware workflows let agents do more than just edit files blindly.
• Fabric Task Flows comes up as a strong example of this pattern because it can help generate architecture, deployment artifacts, and CI/CD-ready scripts based on a stated business problem.
• Both hosts stress that teams should not treat the first generated workflow as production-ready. Good agent workflows usually require a few revision cycles before they become reliable enough to trust.

Looking Forward

The next step for most Fabric teams is not to automate everything at once, but to pick one frustrating CI/CD task and make it repeatable. Start small with a merge-conflict helper, a validation script, or a commit-driven documentation workflow, then refine it until the output is dependable enough to become part of your real development process.

Episode Transcript

0:07 measures. It be lighting up the sky. Dance to the day. The laughs in the mix. Fabric and A. I get your feels. Explicit measures. Drop the beat now. Feel the crowd. Explicit measures. Good morning and welcome back to the Explicit Measures podcast with Tommy and Mike. Hello everyone and welcome back. How you doing? Happy summer, Mike. We are swinging into the summer really

0:38 We are swinging into the summer really quickly here. It is moving fast, Tommy. Time is just flying by this year. We’re almost halfway over to the year already. Can you believe it? It’s weird. I feel like I just started 2026. My wife and I were just talking about the same thing. We’re Oh, really? Oh, yeah. Like I I am I am amazed that it’s June. Do you plan big summer vacations at all? Like is that So your kids are a little bit younger. Younger kids are a little bit harder to travel with, which I understand. But what’s what’s your routine look like for like traveling and moving around? Is

1:08 for like traveling and moving around? Is that that a great question. Is that a thing for your family? How does it work in your family? So what we’ve tried to do the last couple years and I think when we started really forming our family is try to rotate because my family is the type of family where we don’t plan things but it’s like we’ll get together. My wife’s family always has an event to the name like so it’s like it’s the summer vacation Laram 2026. Okay. Okay. So it’s more likely you would do something on their side because it has a name. The other one’s like yeah we’ll do

1:38 name. The other one’s like yeah we’ll do a barbecue. So what we’ve tried to do is they don’t listen. but we try to do every every year one family, the other family, then just our family. So like basically like this is going to be the Pulyas, then it will be the Laramies, then it will be just us like the five of us every year. When it comes to the things we do, this is a tough one here because you’re right, my oldest is nine, which is crazy in itself, but we’re going to New York this year, going to going to a wedding, so

2:09 year, going to going to a wedding, so that’s where we’re going to do the poolary thing. on my side of the family and we were talking about like going to Ellis Island. Like that would be cool. You can see the history. You can see,, look at your great-grandfather’s name in the book, but this is tough because I would lose my mind, Mike, if they’re like bored, which of course they would be because they’re not going to get it. No, No, but my brain would blow up. I’m like, it’s your family. No, you don’t understand how cool this is. But I’m nine, so I just want to play video, you

2:39 nine, so I just want to play video, you nine, so I just want to play video,, move on. That’s why we’re not know, move on. That’s why we’re not doing Italy anytime soon because I don’t think they would appreciate it. Mike, your kids are slightly older so or a little older. So, are you when did you find that transition when you do trips when it comes to Oh, I know you may not appreciate it fully, but you’re definitely going to have a moment like have did you think about that when you plan trips?, yes, a little bit., I think now now I think you’re our focus has shifted a bit more towards,

3:10 shifted a bit more towards,, we’re getting closer to high or have two in high school and we’re getting closer to like graduation of high school and we’re like, okay, we don’t have many summers left until they go off to college and they’ll come back hopefully from college to us to visit and things, but someone else else it’s a little different like they’re going to be leaving the nest here pretty soon. So,, we’re changing a bit of our our tune and perspective of like if we’re going to do some trips or travel things, we should be doing them now before they fully leave the house and they’re they’re gone. So,, a lot of our trips now are getting a little bit

3:40 our trips now are getting a little bit bigger. We’re going a little bit farther., Kelly and I like to,, my wife and I, Kelly,, like to travel and we like to be around different places. We I go to conferences sometimes. She comes with me when we we conferences. So, we want to have that joy of like giving that experience to our kids as well as as much as we can. I will admit though, travel used to be like, let’s go to like somewhere that’s like an hour or two away and go like get like an Airbnb or rent something and just stay there and then come home. Now, it’s gotten expensive, right? So,

4:11 Now, it’s gotten expensive, right? So, the the the older they get, the more like, okay, let’s go to Washington DC. Let’s go see something over there. Let’s go to a different state. let’s go to some place we haven’t been before so that the trips are getting a bit more elaborate., but I think that’s just we want to make sure we get build some lasting moments and memories before everyone goes off and and does the college thing. Are the more extravagant trips by design or is just naturally happening that way? I think they can handle more. Right. Before when they’re little, it’s hard to

4:41 Before when they’re little, it’s hard to travel through an airport with a lot of little kids. Wow. And so you’re either like, “Okay, we’re going to just man up and drive,” which again depends on how your family climate is. I think most of our family climate is doesn’t really if you if my son and I were driving, we can handle drives. we just purchased a car in Florida and so we went from Fort Meyers, drove the car all the way back to Wisconsin. Nice. Nice. Was like one day of like 9h hour day and then another day of like 14 hours driving. So he and I just took it in two

5:12 driving. So he and I just took it in two days. We just zoom all the way through and it was easy. Like we not a problem. No one had a problem sitting in a car for that long. If you had brought my girls, they would have not been able to make it that long in a car for that long. I just I got to get out. I’m antsy. I got to stretch my legs. I’m like, we’re just driving. Just Yeah. Yeah. Hunker down. The best thing in the world. We’re sitting. sitting. So maybe it’s just the simplicity of the man brain versus the the the complexity of the woman brain. Like I was just like, “Oh, just drive myself home.”

5:43 like, “Oh, just drive myself home.” It’s the same thing like when you have a fire. Like guys can just like stare at it. Just have the conversation. Even if you’re in a different situation, guys can handle it. You put a fire and we’re like, “Cool. I messmerizing. You can just stare at it for hours.” Meme. Yeah. it the we’ll be driving to New York, Jersey, and all that. And that. And so they’re generally pretty good. We have stuff for them to do. I have they have their but we’re not the What’s the drive time from from Chicago to New York? So since we’re going to Long Island

6:14 So since we’re going to Long Island first is probably about 14. So it’s not terrible. terrible. You do that one shot. No, stay in the middle somewhere. We’ll we’ll stay in the the middle like somewhere in maybe we’re going to go to Cleveland. stay there for a night and then just drive the rest because to your point everyone will go crazy and then we’re at that age right now where like does anyone need to go to the bathroom? No. Okay, 10 minutes 2 minutes later, I need to go to the bathroom. I’m like, and then you’re like, my vacation’s over. over. That’s it. We’re never going to get there. I don’t know if it was Jerry Seinfeld,

6:44 I don’t know if it was Jerry Seinfeld, but he has this joke. He’s like, when I when we were doing our vacations and you have to pack and all the things are in the car, when I finally close the door on my wife’s side and everyone’s in the car and I walk from her side around the car to the driver’s seat, that’s my vacation. That’s my little piece. But yeah, I I the summer’s good though., when the kids are littleer, I feel like you need a vacation for the vacation. 100%. 100%. Because I think in the beginning stages

7:15 Because I think in the beginning stages when kids were young, I was like, “Oh, this is going to be fun and relaxing and I’m used to going to places and relaxing.” But it was actually more work when you have lots of little ones and you’re doing keeping care of them and they’re hungry and everyone else like, “Wow, this was I I would have rather just stayed home because it was just less work to stay at home.” Let’s be clear. When you do that, that’s called a trip, not a vacation. It was a trip. We did lots of trips in the beginning. Not a trip. Now you’re doing very little vacationing. Yes. 100% agree on that one. All right, let’s get into our main topic today., the main topic today is

7:47 today., the main topic today is talking about continuous integration, continuous deployment automation with agents inside Microsoft Fabric. So, this is interesting, Tommy. Yeah. Yeah., we’ll try and pick this up here. I’m not hearing a lot of social media or people describing or explaining if they’re using agents in their automation experiences. So, let me kick it off there. I think that’s our our main topic today. Anything else, Tony, you want to cover up before we hit the main topic? Honestly, I I want to dive in because I’m I’m Okay. I’m in the same place as you when

8:18 Okay. I’m in the same place as you when it comes to you would think this would be the most easiest simplest way for people to get started with anything agentic around fabric because it’s already utilizing git but to your point I don’t hear a lot there’s not a lot of howtos there’s not a lot of this is what I’ve tried you and I talk about it but outside of that I’m not heard much to be to be frankly honest honest so and I think the reason why we’re talking about it today and I think we continue to talk about it is if you’re

8:50 continue to talk about it is if you’re not aware regardless if it’s a fabric repo or just a normal GitHub repository a GitHub is obviously incredibly utilize agentic solutions like when I commit or there’s an issue have certain triggers that will actually pull utilize an agent to run over your repository to write comments whatever it is again it does not have to be fabric because we have the ability ability to integrate with fabric though there should be some really dead easy like

9:22 should be some really dead easy like tasks automations and just enhancements that we should be able to do. I and I’m thinking about this Mike Bull from me as the individual and again as a team on what can we do to help our lives out when it comes to utilizing CI/CD which hopefully will convince more people that’s the default way you should do it and then more importantly with what agent solutions are available out of the box in GitHub. [snorts] [snorts] Yeah. And I think I think there’s also a

9:53 Yeah. And I think I think there’s also a note here around like I I think people sometimes still get confused around what is an agent and what is a large language model? model? Yeah, 100%. And then even what is co-pilot? Like how are those three different because what is co-pilot and how does that work? And again as I as I’m continue to formulate my mind around how these different things work. So let me let me

10:15 different things work. So let me let me before we go down a maybe a large rabbit hole here, let’s maybe level set on a couple like continuous integration, continuous deployment items. I think we need to define a little bit of what that means to us. to us. Yeah, Yeah, because I think that will help frame out where the agents fit into this as well. there in my mental model, there are two methods to do continuous integration, continuous deployment. There’s the simplified version which is Microsoft gives you a PowerBI premium

10:45 Microsoft gives you a PowerBI premium per user or PowerBI premium and you can use deployment pipelines directly inside fabric. So the idea or the thinking here is you need three workspaces. You’re going to have workspaces or three or more, right? For whatever you’re talking about. There’s a workspace for dev, there’s a workspace for test, there’s a workspace for prod. You could have more or less, but generally there’s more than two. Two two or more. And then you’re trying to move an artifact, some item in a workspace from one environment to the next, right? I

11:16 one environment to the next, right? I have a semantic model and I have a report. I need to move the semantic model from the dev environment to the prod environment. How do you move the definition, not the data, how do you move the definition of that item from one to the other? Deployment pipelines give you that ability. There is a little bit of I’m gonna say finagling or or setup or there’s a tricky part here because the report points to a specific semantic

11:46 report points to a specific semantic model and when you move the report from workspace dev to workspace prod. You now have to understand that that work that report was pointing to the dev model. You need to delete that connection and replace that connection with a connection that goes to the prod model. Okay, so that’s how the the pattern works here. A lot of other investment has been made across most of the stack to help you build out a lot of other items such as notebooks and SQL databases and other systems or other

12:17 databases and other systems or other items that are in the workspace. And most of them, I haven’t played with all of them because there’s just so many anymore, but most of them will allow you to rebind or during the deployment pipeline process allow you to make a parameter that says, “Okay, I’m going to parameterize this connection string on this report.” And that way when I deploy the report, I can dynamically replace the string in the deployment pipeline. Okay. Okay. All this to level set to say deployment pipelines are more for the clicky clicky developer that wants to just click things through the system. Yeah, you

12:49 things through the system. Yeah, you agree with that statement there. 100 honestly 100%. And because we have source control and GitHub today a hot take here, my more extreme takes, but I feel like deployment pipelines for me have become more like the poor man’s solution for CI/CD CI/CD because of the breath of features that we have in GitHub or in a real repository. It’s great and I know I noticed a lot of people use deployment pipelines but for me I’ve

13:20 deployment pipelines but for me I’ve transitioned basically everything over to the GitHub repository solutions rather I don’t know how much are you before we get on to the actual repository side of things source control side of things how much are you relying on deployment pipelines today I think it really depends on the organization or how your your company culture is if you’re if your BI team or what if you have a central BI team if you have a federated team if your team’s skill set is coming from the

13:52 your team’s skill set is coming from the business angle if you’ve hired people that are building reports and they’re coming from the PowerBI ecosystem PowerBI has been out for much longer right 2015 we have 11 years of PowerBI work so if you are hiring or have hired a lot of straight up PowerBI developers you’re probably going to want to use more of the interface that’s just seems to be the business way of doing things it’s more interface driven it’s not so much code. So in those organizations, I’m going to probably say it’s good to start with the deployment pipelines. That’s that’s a good place to

14:23 pipelines. That’s that’s a good place to start because most of the easy things you’re trying to do are accomplished there. When you start moving into like lakehouses and notebooks and some other the other items that you’re trying to then distribute across different environments, things start to get more complicated. Yeah. I’d also argue if your team is coming from the IT side of the world, let’s converse that, right? So, let’s go to the other side. In the IT side of the world, there’s a lot more code where we check things in. And to be honest, not all data engineering teams are up to

14:53 all data engineering teams are up to speed. Not all of them know how to use Git. I’m I’m working in companies that have did engineering teams that they just they’re still figuring out what Git is doing or how to do it right or they’ve built their own little pattern around how Git works and reviews are different and pull requests are just not quite figured out yet., in general, general, I think what AI is doing is it’s allowing us to have a lot more access to these traditionally very developer focusing tools. And because I can talk to an AI and say

15:23 And because I can talk to an AI and say make a pull request, resolve this merge, resolve this merge conflict, I can talk to the agent and it just kind I can talk to the agent and it just figures things out for me now. So, of figures things out for me now. So, I’m even things that were like mysterious to me as one who does do a lot of development, not so much anymore. you just hit I think the nerve here on as we get into the topic of okay what can automation and agentics automation do for when you’re in the GitHub repository I think the framework we’ll take today we’re let’s assume organizations are using

15:53 let’s assume organizations are using GitHub in some capacity right well just off the bat Mike I think there’s a ton of angles we can take here but one of the more complicated things people find the first time they start using source control and especially the first time they’re using it in fabric is what you just said a merge conflict what’s that so unlike in a binary file which I can make an update and

16:23 which I can make an update and it’s really a zero one so to speak there’s not really the code behind it if I’m now working in a PowerBI project let’s say two of us are working on it on our own computers well I updated a measure but at the same time someone else on my team updated measure. There’s two different commits. I’m going to get an error or I’m going to get a notification from source control when I try to pull the new changes going, by the way, we don’t know how to resolve the two edits that member made. These two commits can’t work together. You’re not adding to it, you’re updating. So,

16:56 not adding to it, you’re updating. So, yeah. Go ahead. And be I want to add just a little bit of user context here of this one, right? Yeah. Yeah., you’re talking about one I think this is just for again I’m unpacking what you’re saying here just briefly before you keep going on here. PBIP everything’s broken down into smaller files. Yep. And the conflict happens when so the reason you break a PowerBI report or semantic model into the PBIP format is because then you have individual files representing each table, table, each culture page.

17:26 each culture page. Yeah. And so the report size gets individual pages down to the visual. So by having smaller items that are singular files, you can update a single visual and not impact the definition of the page. Right. Right. Right. Right. And so that that what that does is it enables us to work on different parts of the file. the file. Right. Right. When some user modifies the same measure or even different measures in the same table, the table definition with the measures inside it are all included in the same file. Is that right? Is that a

17:57 the same file. Is that right? Is that a fair statement? Yes. And so and also I can open up the pip file the definition and I can move things around. I don’t have to only code it, right? I can actually change everything in a PowerBI desktop which will change those files itself too. Yes. Correct. So you have the concept but the concept is we have two people editing the same file and that’s when we have problems. Right. If I if we are editing two different files if Tommy you’re in table A and I’m in table B, we can edit till we’re blue in the face and then put our two changes together. No, everything’s none the wiser. It’s not.

18:27 everything’s none the wiser. It’s not. Let’s say we’re editing the same page and I moved thing to the top left. You moved it to the top right and GitHub’s going to say, “Well, there’s a problem here. This artifact, [snorts] we don’t know which one to put it in. Which one should we put it?” And this is probably for most people the most it is the most difficult thing to comprehend and more or what am I trying to say? Much less try to solve it, right? Like, okay, what’s going on here? And then next, what do I do about this? So this merge conflict which is a very important

18:58 conflict which is a very important concept to know in a source control again back in the day I say back in the day but back in the day without anything agentic you have to figure it out you have to change the code and it can be very tedious because source controls made for developers so do you go okay what’s xyz what’s this should I change this color and right off the bat this is where most people go I’m not going to do source control. I’m going to do deployment pipelines or I’m going to go

19:28 deployment pipelines or I’m going to go back to SharePoint because I can’t move forward without getting this conflict solved. solved. And this is not an easy concept either e either. Let’s be honest here. Merge conflicts can be very difficult regardless if you’re just getting started in Git or not. So when we think about automation and just applying things that are available to us, Mike, there is no better way in everything I think we’re going to talk about today than a merge conflict. And the fact that

19:58 than a merge conflict. And the fact that you said that because what can AI what can agents do for us with merge conflicts? Well, agents are so this is another area that I’ve gotten some information from Matias. So Matias is doing a lot more of this than I am and so I’m learning a lot from Matias box. Matias and I are doing another podcast called Agentic Thinking. We’re thinking about any anywhere we can put an agent on something. We’re trying to talk about like what how are we using agents in different spaces. Okay. So,

20:26 agents in different spaces. Okay. So, that being said, when we are working with agents, agents are really good at understanding the intent of the change, right? So, Tommy, you made measure A, Mike made measure B, or maybe I updated a measure that Tommy had, but you added another one. So there’s there’s some the the sing the singular file that we’re looking at has some sort file that we’re looking at has some difference there. The interesting of difference there. The interesting thing about this is sometimes you have to look at okay what what is changing were there other changes in the repo that required this function or this whatever file to change. An

20:58 whatever file to change. An agent is really good at looking at like all the interactions between what your change does, understanding the intent of it, and then proposing or recommending back to you. Here’s what I think we should do to in order to rectify the change and everything work. So, I’m going to I’m going to maybe say two things here that I think are really important. The first one with merge conflicts is the one the agent understands how it works in the broader schema of the tool the tool stack also

21:28 schema of the tool the tool stack also it has the ability to go through and search if I make this change what other stuff breaks the second part of this is I do a lot of pattern where and this is more about app building for me so if I get a merge conflict on an app that I’ve built or I’m working on two different things and something’s having a problem a problem I I’m able to make the agent make the change and then I also ask it to run the build run the run the does this thing still compile correctly. Can you still

21:59 still compile correctly. Can you still verify that it works? Does the solution still run after the change? Right? Because you if the change is made and then the the file or the program or the app all just falls apart and blows up. Like that’s not a useful thing anymore. So So it’s not an app. It’s not it’s not fixed, right? that that just means you just made some you just made two changes when you shouldn’t have and now you have to go fix other things potentially from that merge conflict you have to go fix other things so I think the agent’s good in those two areas areas making the change and understand the intent of the change from two people and then two going back and saying okay now

22:30 then two going back and saying okay now automatically test it make sure that it runs is is it validating correctly check those two things and then I have a higher confidence by having the agent just do those works and that saves me hours of time of not having to do that stuff oh my gosh Yeah. Well, and so it’s funny because in in the fabric world too, let’s say you’re working with someone, let’s say they’re an idiot. Let’s they’re not smart or they’re not smart, but they’re dumb. Like they don’t know Git is probably the better way to say that to be more politically correct here. here. You are opening up,, your

23:01 You are opening up,, your GitHub desktop or whatever you’re doing. All right, let’s get the latest changes. And you pull all some of the measures that you built already are gone. and you notice that some of the pages when you open up power desktop are gone. You’re like, “What the heck happened?” Well, now you have this problem of it might not be a merge conflict, Mike, but you may have this issue where the three measures that you created are gone and the things that you already did. But you have options with GitHub. You can either use the command line, you can use GitHub

23:32 use the command line, you can use GitHub desktop. Generally speaking, most things you need to do are going to be in the command line, unfortunately, with Git with Git. with Git. and you only want to pull some of the things. You can revert a commit, meaning, hey, that push that I did two days ago, bring that back. Mhm. Mhm. Here’s the problem. That would take away all the changes the other person did. So, you have these errors and for most people, it’s really hard to say, I’m going to extract single artifacts. It’s possible, but it ain’t easy. This is

24:04 possible, but it ain’t easy. This is another great option especially for those who are just getting started with working with a team in source control and fabric where I can there’s a co-pilot CLI called copilot and I can tell hey copilot I made a commit the previous day you don’t have to give it a lot of information like it had a three measures in this table I need those in the latest commit I need this from the latest pool they’re removed it will go through like all right let me check it will ask you yeah I see three measures in the display folders that’s we’re talking about.

24:34 folders that’s we’re talking about. Great. Great. It will run all the commands for you. Will run the re not reverting but extract those particular items add it to the latest pull. If there’s a merge conflict it will run that for you and and without anything agentic. This was for most people impossible. Like you are breaking things you have to go back. You can’t move forward. The report might not even open in some situations. Yes. And you’re stuck. like the team stuck at a point you have to cancel a commit and

25:05 a point you have to cancel a commit and it for me when I got started Mike it was by far the most soul crushing moments around source control and I say soul crushing we like I have to redo everything because I I could either try to learn CLI commands for the next 20 hours to get the right items and the artifacts and hopefully it works but I can run it now with general general commands a general prompt and be able to merge things together. And I I can’t

25:36 merge things together. And I I can’t imagine imagine I think we’re just at this place, Mike, where I can’t imagine teams not utilizing this. Like I I don’t know if you have available to you source control. control. What are the excuses to not use the options we’re talking about? Because we’re just talking about two amazing amazing features that are available to us. Why would you not do this? I I don’t I I don’t This is where I think I’m a little bit torn, Tommy. Right. Right now, there’s a

26:06 torn, Tommy. Right. Right now, there’s a lot there’s a lot of messaging around organizations pulling back from AI because it’s too expensive. when when AI was so this this is going to go a little bit off the tangent here, but maybe let me write down let me give you the two points real quick at the front here and then we’ll come back to it. One, AI is getting more expensive. Period. Period. It prices are going up. So that that’s one that’s coming already. The second point here I want to make is if you’re doing straight file edits particularly around the PBIP

26:36 particularly around the PBIP Mhm. There is likely better ways now and MCP servers are very useful in order to make changes with an agent but not actually break things because the tools themselves such as such desktop or powerb. com already have validation of whatever change was made built in and one of the advantages of the MCP servers or MCP tools is if you edit a model with an MCP server the PowerBI MPC modeling

27:06 an MCP server the PowerBI MPC modeling server What this gives you is it gives you the ability to say, “Oh, this thing’s changing. changing. I made the change. Something broke.” The MCP server gives visibility back to the error back to the agent. So the agent is then saying, “Oh, I made I added a comment. I did it the wrong way.” The model says, “Nope, this is not right.” And gives me the line and command of the information that’s wrong. So there’s this there’s this constant

27:36 So there’s this there’s this constant loop of like agent stuff right build something with agent or write a prompt hey agent I want you to build something agent understands prompt agent builds things something breaks send error code back to agent agent picks up error code fixes problem and then here we are again where this like loop of like constantly fixing things over and over again so the the MCP allows the agent to self-close that loop you can give it a prompt and then it iterates through the problem and if you work watch the MCP server for PowerBI

28:06 watch the MCP server for PowerBI modeling run, it will make some changes and it’ll say, “Oops, nope., that didn’t work. It broke. I’m going to retry something else.” It’s going to pick a different route and it’ll it’ll use the tool to do specific reads and writes directly from the model. So I think MCP stuff, tools, and maybe a mix of tools and skills is really where the agent becomes much more not clean. The the agent becomes much more predictable in its outputs.

28:38 more predictable in its outputs. Let me let me focus on this here too with the MCP because it’s interesting you bring this up. when I think about MCP in the context of source control for PowerBI like the one limitation when I use agentic and GitHub repos because you can actually just for anyone who’s listening if you never heard of this you can have a folder in your repository GitHub or and then in that there’s a folder called agents sog Github folder called agents you can have these agent files that run well to your point that’s

29:10 files that run well to your point that’s going to be filebased but it’s not going to be able to query your DAC Right. It’s not it’s not going to be actually run and look at the data in the model. It’s going to look at only files when it comes to the GitHub repo. I think you’re talking about something interesting here where I can create a process with MCP because again just illuminate on what MCP can do the local server in this case. You can actually have it update Power Query, update DAX measures, but it can also validate columns, look

29:40 but it can also validate columns, look at the tables, validate run DAX queries. For example, just the other day, Mike,, we were working on a repository and we had to point everything over to a new data source. I went to the MCP after I got the pull and I have a few, skills that are validation. It’s like hey this this pb file I have open on my on my computer just got a new can you validate the changes and then validate the data.

30:10 the changes and then validate the data. That is what you cannot do without the MCP off the bat. So I actually want to say how many rows are in this data source. what are the what are the results of this measure? If I’m just running the GitHub, I’m going to push back on you slightly around the MCP because you don’t need you don’t need technically the MCP server to do it. It’s easier with the MCP server a lot more straightforward, but the MCP and one of the things that the MCP server wraps up is the MCP wraps up a lot of API calls and handles the

30:40 up a lot of API calls and handles the authentication piece. So that’s that’s why the MCP server is there. You can still call the fabric APIs or call an API. You can absolutely say the DAX queries and get back a a file with the data in it for you. So that still works. I’m just making a note there like no no I think that’s a good point here. You can you can actually have those API calls run and actually send you a kind calls run and actually send you a a result list without using your of a result list without using your local computer at the time. So correct. Yes. Yeah. So but it is if you want to just

31:10 Yeah. So but it is if you want to just get started it’s super straightforward if I’m on my computer just to say what the changes are. But regardless, I think that MCP really does open up a lot from data validation, right? And rather than just looking at files, right? Because I think a lot of people get hung up in our world on the fact that we’re filebased like, oh, I’m not a developer. Like, I’m not building apps. I’m not using MPM or Python here,, to build an application. No one even knows who those things are really most of the time. Yes.

31:40 really most of the time. Yes. Right. Most of the time. So, when you’re thinking about fabric, people think data. and I think there’s that important distinction when I think of source control and when I think of the fabric environment I think of my files but I also think of the data quality here here let me ask you why do you think there is not a lot of buzz in our current unless you wanted to continue on anything on the MCP no no I think the MCP is good for me I what I just maybe would note around

32:10 what I just maybe would note around like just models and like how you’re using agents and things right I think I’m really pushing on this idea. The creator agent is so important here. So, So, find that again. Yeah. So, creator agent is when you use agents to create things. And I’m I’m coining a term here that people don’t use today currently. This is a Michael made up term pushing this. Yeah. Like someone’s going to make up a better idea that’s going to actually be more indicative of like what this is actually doing. But I don’t like using agents to do the work. I like agents to build the

32:42 do the work. I like agents to build the tools that do the work. So they’re very good at the the agent is really good at creating a system, a Python script, a deployment script, a GitHub script. It understands the technology really well. And then I like asking the agent to build something deterministic, meaning something that has a defined input and a defined output. That way it works every single time. Deployment pipelines or deploying something into Git or checking something. These should all be like, hey, write me a really a good script that checks these things. write me a

33:13 that checks these things. write me a script that’s going to do this every single time we go right before a deployment. And then you tell the agent, “Hey, agent, now that you have this specific task that I’m going to do over and over and over again, the agent’s not trying to like learn how to call an API and learn how the model comes back and learn all the I don’t want to teach it that stuff. I want the agent to reason about writing the code. So, I want to spend my time with the agent saying, “Build me something repeatable that I

33:44 “Build me something repeatable that I can measure that has defined inputs and defined outputs.” And then when you need that tool or that task, do it. Again, the reason I bring this up is one of the really rich tools that I really like is the tool made from Alex Powers. Powers. Yeah. Yeah. Yeah. Fabric task flows. He made this task flow assistant, which is basically a skill skill that has inside the skill or a series of skills and inside those skills he has a whole bunch of scripts that do very

34:14 whole bunch of scripts that do very specific things. He has a whole bunch of information that’s been scraped off of the documentation from all these different items and artifacts from PowerBI. It’s basically bringing the knowledge into a single easy to read very lightweight system that you can just go talk to. And so when we when we start talking about agents in the CI/CD pipeline of things, this agent can create for you an architecture from scratch. It gives you everything. It makes the task flow in the in the workspace. It makes the different items. It links the items

34:44 different items. It links the items together. It puts all the items together inside a variable library. So it’s already ready to go to go from dev to test to prod. You can switch out the guids and the information right there directly in those variable libraries in in literally inside this task flow assistant. You say I want three environments. I want to name them dev, QA, and prod. It will do it. It just says oh yeah, no problem. I know exactly. And it’ll build out the three system architecture for you automatically. So the the nice part though is the agent is doing the item

35:15 though is the agent is doing the item building, building, right? right? The agent’s doing the creation of those things. Now you have all that created and then you can step back and say okay now that we have a system the creator agent built the stuff we can then manage it right it runs on fabric APIs alone and they’re very deterministic whatever I send in I get the right answer out every single time it’s interesting you bring up the creator side too because I think for people to get started you’re talking about creating actual artifacts for

35:46 about creating actual artifacts for fabric right so like you said like it’s going to create the lakehouse going to create the environment. Correct. Correct. And it’s funny because I took that differently when you said that because I find like almost a creator creator. and again, there’s a better name here for a lot of people getting started. Let’s say we we’ve talked about skills and MCPs and scripts that can run a DAX query and a lot of people go, I I don’t know what to write. I don’t know how to set that up. Well, to your point, what else? Agents are great at creating

36:16 Agents are great at creating agents. agents. And so one of the things so and I think for a lot of people getting started even if you wanted to take a basic step here when I’m creating a repo and and I have something in fabric or whatever it may be if I want to say I would love for example take your example you did earlier earlier you want every time you do a commit for that GitHub repo to pull call the API from fabric validate some data every single time and create a markdown

36:46 every single time and create a markdown on here’s the latest numbers, use the data data quality. How do you get started? Do you need to know the API? Well, that what a great example of, hey Claude, hey co-pilot, hey GitHub copilot, this is what I want my agent to do in this in this repository. Let’s create it. And to your point, it’s also really great at other creating other agents and other scripts and skills. Claude has an awesome skill called skill creator which is

37:17 skill called skill creator which is awesome at guess what creating other skills skills. So and I think this is for a lot of people where I and this is I think we’ll move to why there’s not a lot of buzz about this. I don’t think people realize this. I think people real think that they have to create these agentic automations and these skills from scratch and it’s like I don’t know how to get started. I don’t know what it needs. not realizing that the agent is great at creating agents. I’ll give you a brief example of just

37:47 I’ll give you a brief example of just what you can do and people don’t realize using the MPC MCP server or the local model server with cloud desktop there are there you do get to a point where it just gets a lot of tokens and it gets a lot of context in it. So there’s a different points I’ll pause and say, “Hey, we’re going to conclude this chat. Write me instructions for a new chat in this project to get started. It’s connect to the local model. Here’s where we’re currently situated because these three

38:18 currently situated because these three tables,, are still connected to a different source. I need to validate this. Write me the prompt for the new chat.” chat.” Yep. Yep. And that way you can get this fresh branch. But yeah, I could write that myself. But to your point, the ability that Claude goes, I know what an agent needs. why? Because I am an agent. I know what another chat needs. So, I’m going to give her those proper instructions. Mike, most things that I do again in my project in notion,

38:48 that I do again in my project in notion, every single milestone or model that I have has a file or page called cloud instructions. So when I start and I go to not I go to claude I go hey read this notion page about XYZ model and let’s begin it has the semantic definition to your point it already has the context that’s needed once it’s done I tell claude okay update the notion page what changes did you made did we run into any issues so that feedback loop where I

39:18 issues so that feedback loop where I don’t have to go in every single time to go here’s the idea here’s what you’re dealing with there’s already this thing and I think there’s this creator creator side outside of creating things around like creating a notebook for me create the thing that creates the thing I agree and and what you’re saying there that the creator agent I think is an underrated term here also tell me what you were noting here about like hey I’m I’m at the end of my contact window I’m about ready to run out a gentleman

39:48 I’m about ready to run out a gentleman that I like to follow his name is Matt PCO PCO and he has a a couple skills that I use regularly. One of them is called Grill Me, which is like an interviewing system like, “Hey, I’m going to build this system. Grill me about it.” Right. It asks, It asks, “Have you used that in your statements at work yet?”, I have. Yep. Yeah. Pretty fun. So, I don’t I don’t Yes. And sometimes I also I don’t want it to like ask me like a a ton of questions because I don’t have a quite Yeah. Yeah. Well, I don’t have I don’t have a good enough system in place to to actually do

40:18 enough system in place to to actually do it like with talking to the agent, right? If I could have an agent just talk to me, say, “Hey, all right. All right, I have a couple questions and just candidly talk back and forth. I think I would probably use it a bit more, but I I use it more in coding sessions for me like for writing like I’m going to build an application or something. Do you understand what I’m making here? Grill me there. Mhm. Mhm. So, that one works really well. But he also has another skill called handoff. And handoff is it he has some of these skills that are just incredibly short. There’s only like five lines to this skill. skill. It’s doing what you say, Tommy. Make a

40:49 It’s doing what you say, Tommy. Make a handoff between this session and a fresh

40:52 handoff between this session and a fresh agent. agent. Suggest any skills you think we need to know and give me any context that’s not already written out in your product requirement documents, plans or issues or commits. So, so you already have access to all the information you need. Give me the context for things that are not quite there. And that’s that’s really I really like this. This one I have been using a little bit and think, oh wow, handoff is pretty slick. But describes a lot of what you’re saying as well, well, right? right? Yeah. And these are awesome like and go continue because I have a question for

41:22 continue because I have a question for you. you. Well, I I want to push a bit more on like where are we using agents inside this continuous integration, continuous deployment pipelines. Like there’s like what I talked about right now with Alex Powers, which was this is an agent where I’m applying this agent to help me build the items and build the different environments al together. Another area that I I see myself using agents are there are there are times so we’re going to move away from the deployment pipeline piece. So deployment pipelines again is that that’s the business user

41:52 again is that that’s the business user centric version of this. There is a level of knowledge that you have to bring your team up to especially when you’re doing development when you were talking about git and git integration. There is a learning curve there. there is some time needing to be spent to educate the team around how does this work? How do we do development? What can the workspace modify? What can the branch modify? how do we pull code down? What’s our what is our what is our design process? Do I use MCP servers for

42:22 design process? Do I use MCP servers for making all the model changes moving forward? Do we pull down a repo on my local machine and build from there? Do we allow changes in the service or not? There’s a lot of like requirements I think that go with go with this. The reason I’m going down this route is after you get over that hump, once you’re beyond the Git integration piece, then there’s a new world opened up with tons of flexibility around GitHub and DevOps actions. Mhm. Mhm. And these actions are Oh, yeah.

42:52 Oh, yeah. Okay. So, so the actions piece to me is really what makes a lot of this code on Git powerful. And I’m going to speak a little bit from an appdev standpoint, but then also from a PowerBI standpoint as well. In the appdev world, actions are the things that kick off the deployment, right? I’ve made some changes and there are automatic I’m going to use a term here like real-time triggers, right? A real-time trigger in in my mental model is when

43:24 trigger in in my mental model is when something changes in code, when someone goes back to a particular branch, let’s call it the main branch. When the main branch is updated, the GitHub system or the Azure DevOps system can watch. Okay, we’ve made a commitment to a pull request has been made on main code is changing. Now what? And these are like eventdriven actions that show up and okay, Tommy made a change. I don’t need to get a build going. And then in here, you can make

43:54 going. And then in here, you can make these actions do certain things, deploy these artifacts, go over here, do these things. And so this is where I think the agent also shines really well or using a a coding agent or like a GitHub a VS code when you’re working on the code library that get is on your computer you can say make an action that deploys these artifacts into this workspace based on this information. Right. Right. Right. So now now you can automate and this is and this is where my world a lot is right now with companies because I do really big deployments or my

44:26 really big deployments or my customers are building hundreds of workspaces for their customers to to build inside part so I’m an organization and I’m trying to share reports externally or build workspaces for lots of people or I have multiple like this doesn’t fit the normal CI/CD pattern that deployment pipelines gives. So these are really complex potentially highly automated systems where you need a lot of other things. And this in the actions area makes a lot of sense

44:57 in the actions area makes a lot of sense because now the barrier to create more complex actions after something is done. You now have the ability with more control there. And I and this is and why I think this is super powerful is I I need to understand the principle. When I commit to main I’m gonna push code here, right? I’m going to use a tagging system. I’m going to add a tag to this. And when I when it when I commit to main, don’t automatically build production, but only build production

45:27 production, but only build production when the tagging or or there’s specific things you can add to there that that build to production when you need to. Let me give a really simplistic kind Let me give a really simplistic look at this here because I think of look at this here because I think there’s a really great one. If you’re like okay I have so many ideas but what’s a or what what is a good idea around what you’re talking about actions. The simple way to think about is if this then that if this occurs a trigger then have these things happen. Yes. Yes. What I do in all my fabric repos is I

45:57 What I do in all my fabric repos is I have an action that says, “Hey, if in this folder there is a commit that updated, run my what’s new agent. The what new agent has read me files that is a what’s new file. It’s a re just a markdown on all the changes in fabric for that workspace or the particular folder that happened including links to a change log.” So, we’ve talked about this with our knowledge center and the center of excellence and how difficult that was. Like, hey, we made updates to

46:29 that was. Like, hey, we made updates to the model, but people have no idea like what those changes were and it’s so hard to keep on track of that. I have an agent Mike that says, “Okay, if I make a change to this,, this folder of models, then add to this what’s new file, add to the change log, and add to the documentation.” and an agent will scan everything go oh I noticed that the latest three commits I don’t have updates for let me write hey we added three measures we added a new page we have a new summary page that describes x

47:00 have a new summary page that describes x y and z and it can scan all these things and provide not just a change but descriptive change so this is helpful for consumers but man is it helpful for the developers too to go what’s new oh look Tommy updated and he added the dash dashboard page. What does he need? X, Y, and Z. I don’t have to write any of this. I have a commit description, but that’s an easy, great, straightforward way of getting started here., Mike, there are a ton of

47:32 here., Mike, there are a ton of different ways here. And I’m so curious to hear from you because I’m I’m flumxed and thank you. Great choice of word here., just or you’re welcome because I thought that was a good choice. I I I’m vexed at the same time. So really I think I’ve been reading lately. Maybe that’s what it is. you’ve been getting all these big What blogs are you reading these down? Big word Tommy today. I’m getting a little pretentious I think actually. actually. Oh my goodness. Oh yeah. So three for three. Yeah. No, I but I I’m honestly confused

48:04 Yeah. No, I but I I’m honestly confused because I don’t know what with all these different things we’re talking about, all these different opportunities in again from very advanced to pretty dang basic. basic. Mhm. Mhm. Why aren’t we hearing more about this? I I I’m so curious and I would love to hear from listeners in our mailbag. Are people who are using source control taking advantage of this? Why is this not a lot of buzz where I feel like this should be like top three buzz in the fabric world now? So, what’s going on?

48:34 fabric world now? So, what’s going on? What’s your take around no no talk around this? Well, I I think this is also fairly new, Tommy. I think I think a lot of people are still trying to figure out how do I build things with agents in general? And so, I think the the CI/CD story with agents or leveraging agents to help you create things is still a bit immature at this point. I’d also maybe argue too, Tommy, we’re probably ahead of the curve on the building with Git in general. I think most organizations are not comfortable with it., most organizations are not thinking

49:04 organizations are not thinking through like this is traditionally an area that was fully controlled by it, right? If you want to get into the Git integration area or GitHub or you want to go to Azure DevOps, these are typically solutions that were like, well, it builds software and they’re the ones that are going to manage that. And so in order for you to get access to it, there’s not a good control mechanism to distribute multiple projects of these git systems to broader teams. You’re I’m not as an organ as an IT leader, I’m not going directly to the business team and

49:34 going directly to the business team and say, “Okay, great. You have SharePoint and now you have Git like go use it.” They’re going to they’re going to not have any clue about that, right? So I don’t like that excuse because how long has GitHub been available? I think it’s just what’s actually there. It’s it’s not an excuse. It’s more of like what do organizations actually understand and again I think this is another trend I’m seeing inside AI is complex hard to do things are becoming more accessible to users across the spans. So you’re going to in the same

50:04 spans. So you’re going to in the same way we gave data engineering tools to PowerBI users by giving them PowerBI desktop and power query we’re giving the same flexibility to well now the business user can understand git can use it can track things can then use code to manage their deployments of things right right what we’re doing is we’re we’re taking these agents and it’s becoming for lack of a better analogy the agents are becoming the power query for data engineering back in the day if you went 10 years 11 years ago and that that’s what’s happening. So

50:36 and that that’s what’s happening. So that that inter layer of those agents are now becoming removing the complexity away from things and I’ve been seeing this for months now. This is I saw this move back in January this year. I said look these agents are getting so good. I will not out I will not be able to outthink them. it will know more about Git and Git integration than I ever will and I can just ask it questions around one helping me skill up to it. But then two, if I needed to build an action, I just say, “Build the action. Here’s what I want.” And I describe the nature of the action. I

51:06 describe the nature of the action. I describe the intent of the action. And the agent says, “Yeah, I know what you’re you’re trying to do. I’ll write the code for you to get that intent done.” done.” So that’s I think there’s a a huge wave of net new work activities building. I’m not sure how you want to call it right now, Tommy, but I think there’s a a very large net new ex work accomplishments needing to be done that are going to the how do I say

51:36 that are going to the how do I say this? there’s a lot of testing and confirming and making sure the agent’s building what it is says it’s building. Yeah. Yeah. So, we’re we’re in the space right now where everyone’s like, “Wow, this is so exciting. We can create lots of things. Creator agents, I’m on board. Let’s go do it. All these things. There’s going to be there’s going to be a time a point in time where you’re going to pay the pay the piper. Oh, shoot. Well, you’re going to pay the piper bit. You call it tech debt, right? When your developers No, the debt comes due. Yeah.

52:07 No, the debt comes due. Yeah. Your your developers built apps and things like,, we could make it the right way, but that would take us another two months. Let’s just build it now and get it out the door. Okay, great. So, there’s a lot of tech. I’m literally dealing with two situations like that right now. Okay, so we’re we’re at the phase where the agents are creating more things faster than ever before. It’s helping me build like totally agree with this. See it on the I see it happening other day. I see it in my own company. I see it in my own other companies as well. It’s helping you build more than ever before. We are also acrewing a lot of agent

52:39 We are also acrewing a lot of agent debt. instead of tech debt, it’s agent debt that we’re not we’re not fully building out the right exact system the way we want to and we’re cutting some corners or it’s doing what it needs to be doing. There’s no testing. We’re just shipping things out the door. There’s some level of rigor and now that the agents can create 10 times faster than what people could create. We’re now creating a lot more potentially agent tech debt. Yeah. And so at some point in the future, we’re going to have to pay our dues and we’re going to have to come back with those agents and clean up and

53:09 back with those agents and clean up and fix and add testing and add more rigor because the if we allow the agents to only do their process without having them build tools and and creator agents, it becomes less testable. It becomes more more more of a black box, ambiguous. It has more soft edges and then you have a harder time testing. Did it actually do what I said it to do? And 3 months from now, did we corrupt something in how we talked to the agent in a way that is now not doing what we expected? expected?, it this is a really good

53:39 , it this is a really good point. This is kind going to be close to my closing thought because that makes me think of how do I get started, right? How does someone begin this journey if they’re at that barebone stage or they’ve been testing it out? And honestly, to your point, the biggest thing I would recommend is test out an agent. I would not start deploying things to your point right away. Go, wow, I’m just gonna ask a bunch of agents to create other agents. Let’s let’s give that an example of that what’s new agent that I have, right?

54:09 what’s new agent that I have, right? Do not deploy that immediately in terms of like we’re going to use this and rely on this is see how it works. Ask an agent to write a what’s new thing. Have a few commits. See what the output is. And then there’s a refining process. And I think that’s where people run into the most mistakes around anything agentic is they just write one prompt or one instructions, one re,, one phase and they go, “We’re done. It works. It’s fine.” There’s a revision that I always do before it becomes something that becomes

54:40 before it becomes something that becomes part of my reliance or my workflow task flow. flow. And like that what’s new updater probably took like three or four because sometimes it’ be an error. It was not what I wanted the output to be. So I’m like, “All right, let’s go back to the drawing board. Let’s go back. What are the instructions here?” And if you want to do something more complex, like if again you want to pull from the fabric API to,, your data validation sheet, just do some start small. It’s like what we did with DAX actually. If you had a complex DAX query

55:11 actually. If you had a complex DAX query or formula that you needed to write, you broke it down first. You broke into the variables. You said, “Hey, what’s the output of this? What’s the context here?” that helps you build it. And I’m not finding a lot of differences when it comes to building agents. Again, a thing to create the thing., I think that’s a great way for people to get started. Any idea that you have probably can be accomplished, but it’s not going to be on your first at bat. It’s going to take a few swings to get it right. And I

55:41 a few swings to get it right. And I think this is hopefully we’ll start seeing more buzz. Again, I would love people who listen, the two people who listen to actually let us know if I am doing this and this is how I’m doing this because I feel sometimes like you and I are the only dang people who are talking about source control and get automation. Like I know Kurt’s doing an awesome job with the model development, but I think there’s this other side, the more codebased, the more developer and again the more assistantbased that we are still uncovering. So yeah

56:12 that we are still uncovering. So yeah man I I think revisions get a few swings in there and you can really see what you are desiring your what you have envisioned to come to life. I agree. Yeah. A lot of this is I think you just need to get started. I also find that find that I learn a lot from my social platforms of people doing things that are similar or in veins or software development. So following a lot of people., one of the gentlemen you should definitely follow is Matt PCO on GitHub and

56:42 follow is Matt PCO on GitHub and YouTube. YouTube., he’s definitely building a lot of good stuff like what they’re building., definitely check out the Agentic Thinking podcast. Matias and I are going through and showing you what we’re building. Like you you don’t know what you don’t know, right? So, I’m trying to learn from other people. I’m trying to share what I’ve learned around agents and how they work for me. So, it’s really worth your time to step into this, figure out that this is going to be the future and start investing some time learning and playing and using this., and also, I would argue,, I can’t tell you the number of companies.

57:13 can’t tell you the number of companies. I’ve even heard some super scary stuff from leadership and companies say, “We’re not going to let you use it. Go use your personal computer and go buy your own subscription somewhere else and go use it.” That is a in my mind that is a massive huge security risk for companies to be saying that to their leader from their leadership down to their people. If nothing else you take away from this conversation, make sure that you are doing this the right way. Make sure you were trying to go through the proper channels, trying to get your company to pay for the license or the enterprise

57:43 pay for the license or the enterprise license for you because anytime you’re using these nonenterprise-based licensing, your prompts go right to that company and whoever is using them, they can see them, they can train on them, they can use them as information. That is not secure. And so, right, I would highly recommend do not use those personal licensing things unless you absolutely know and have gone through tooth and nail on the the contract that they have a zero data retention policy. This is one of the reasons why I have put my eggs in the Microsoft basket because I feel like I can trust

58:13 can trust they’re saying at the zero data retention policy. I use Foundry for my models and agents and I use GitHub Copilot. like that’s those are the only two systems I use and I get access to as many models as I need in those systems from those from those teams. So that’s a lot of what I do. All right. Anyways, enough said there. we’re going to wrap it here. Tommy, any other final thoughts before we wrap here? here? start using it. And I think like Shaq said, the reason they lost is you got to score more points. Just you got to get

58:44 score more points. Just you got to get started. So that’s brilliant. I think in order for them to win here, they need to score more points. Yes. Exactly. So jump in, get started, start building things, score more points,, which will help you eventually,, win better with AI. I totally agree with that one. That makes a lot of sense. Well, thank you all very much for listening to the explicit measures podcast. We talked a little bit around CI/CD automation. The links for the things that we talked about, so talking about fabric task flows from Alex Powers, that will be in the description down below. If you want to learn more about variable libraries, those

59:14 about variable libraries, those description will also be down below as well. And if you want the conversation around Matt PCO and how he talked about handoff skills, that’s also down below as well. So, highly recommend these reads, go use these tools. Worth your time. Agents will save you a ton of time building out CI/CD or patterns around working with Git, deploying artifacts. All of that can be much more automated now. Oh, the other one I didn’t point out, Tommy, was the the new release from the startup. what’s that one?

59:44 from the startup. what’s that one? the fabric startup fabric. Oh, we talked about that earlier this week or last week. Well, I just want to call off that. That’s a link that we should put in here as well. as well. Yeah. Yeah. What’s the name of what’s the name of the project again? It’s called jump start. Fabric jump start. So, another area is that is not really aic, but it definitely helps you get going with patterns of fabric systems to build. Fabric jump start is another one. We’ll also include that in the link down below., it’s again it’s a it’s an not an agentic way of doing this, but it does have a more

60:14 of doing this, but it does have a more I guarantee you agents were made to build this. Now, there’s a Python library that lets you install all these different patterns and systems directly into Fabric as well. So, Fabric Jumpstart is another one that’s really useful. I’ll also include that in the show notes as well. All right. Thank you, Tommy. Appreciate it. Where else can you find the podcast? You can find us in Apple, Spotify, wherever you get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. you have a question, idea, or topic that you want us to talk about in a future episode, head over to PowerIt Tipsodcast. Leave your name and

60:45 PowerIt Tipsodcast. Leave your name and a great question. And finally, join us live every Tuesday and Thursday, a. m. Central on all Power Tips social media channels. Thank you all so much, and we’ll see you next time. In the mix, fabric and I get your fix. Explicit measures. Drop the beat now. Pockets can’t steal the crowd. Explicit measures.

61:17 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.