AI Drives Fabric Success – Ep. 493

Mike and Tommy kick off 2026 asking the big question: will AI drive Fabric’s success? They share their real AI toolkits, debate the future of consulting in an agent-powered world, and brainstorm how Claude Code and agent skills could transform their daily workflows.

News & Announcements

Databricks Excel Add-in

Mike highlights a LinkedIn post about a new Databricks Excel add-in that lets users connect directly to the Unity Catalog, select tables, write SQL queries, and build pivot tables — all from within Excel. Mike sees it as a direct shot across Microsoft’s bow: Databricks is skipping the semantic model entirely and going straight to the user’s favorite tool. The add-in includes row limits to prevent Excel from choking on large datasets and is aimed at CFOs and analysts, not developers.

Open Source Projects Going Closed

A developer announced he’s taking all his projects closed-source because AI agents have scraped all the documentation — the agent becomes the expert, destroying the business model of “code is free, docs are premium.” Tailwind is cited as an example, having to significantly cut their team. The prediction: a new marketplace will emerge built for agents, where you pay for your agent to have access to libraries and documentation.

Tommy compares it to how mobile internet disrupted newspapers — paywalls were the first response, and they didn’t work well. The implementation of agent-based access is the real challenge.

Claude Code (Computer Use)

Claude’s new “co-work” feature runs locally on your machine through Claude Code, letting you organize files, build presentations, process receipts, and batch-rename content — all by talking to it. Tommy challenges Mike to try building agent skills: create a skill, give it context about your domain, and let Claude handle the busy work with minimal direction.

This leads to a breakthrough idea: both Mike and Tommy write consulting proposals in Markdown. Years of statements of work, estimates, and project plans now serve as training material for agent skills. Feed in a pre-sales call transcript, point it at your library of previous engagements, and out comes a draft proposal.

AI Is Rewiring the Economy

AI Is Rewiring the Economy — Simon Taylor argues AI will do to services what containerization did to goods: drive a massive cost reduction. In 1956, loading a ship cost $5.86/ton; after containerization, 16 cents. When AI does that to knowledge work — writing DAX, building pipelines, designing reports — the economics of consulting and BI services fundamentally change.

Main Discussion: AI Drives Fabric Success

Where AI Actually Works in Fabric Today

Mike breaks down where AI is landing well versus where it’s still a miss:

Where it shines — creation and code generation:

Building SVG visuals by describing what you want
Using MCP servers to organize semantic models, create measures, write descriptions
Translating between languages (M code → SQL, SQL → Python)
Building color palettes from images via Copilot in Edge
Generating theme files and deploying them through agent skills

Where it’s still mediocre — insights:

Copilot producing text summaries and tables from your data
AI-generated insights (“you’re up 5%”) — still very much an “eye of the beholder” problem
Microsoft keeps focusing on insight generation; Mike thinks the real value is in creation

The 10x Fabric Developer Toolkit

Mike and Tommy share what they’re actually using day-to-day:

Mike’s stack:

VS Code + MCP servers (Fabric MCP, Power BI MCP) — does what Tabular Editor does but through natural language instead of C# scripting
Copilot in Microsoft Edge — free, already available, great for code translation, color palettes, and quick Python functions
Power Designer (in Fabric) — theme generation tool with AI-assisted color palette creation
LLMs in VS Code — experimenting with building SVG and custom visuals

Tommy’s additions:

Claude Code with agent skills — pre-built context projects (e.g., “Fabric Guru”) loaded with Microsoft documentation on notebooks, PySpark, etc.
Cursor + Claude Code — for broader development work
GitHub-based review processes — automated validation of workspace changes via Git integration, catching breaking changes in TMDL files
Browser sidebar agents — ChatGPT/Claude running alongside Fabric in browser sidebars with pre-loaded project context

The Bottleneck Just Moves

Mike shares a key insight from an article about a team that 10x’d their developer output with AI agents: the bottleneck shifted from code production to code review. Developers could produce at the rate of 20-30 people, but the QA and approval process — still gated by humans — became the new constraint. They had to move people from development to QA.

The same applies to Fabric: you can produce reports faster, but someone still needs to validate them. Does the data look right? Does it answer the business question? AI pushes the bottleneck downstream — it doesn’t eliminate it.

What It Means to Be a Professional Now

Tommy reframes the value proposition: it’s no longer about selling technical services (“I’ll build you 15 reports with 10 DAX measures”). It’s about selling outcomes and domain expertise. The ability to write DAX will be assumed — the differentiator is knowing what the business actually needs, scaling that expertise with agents, and validating the results.

Mike agrees: the human role shifts to being the liaison between IT and business, interpreting requirements, documenting what tables and measures mean, and ensuring the semantic model accurately represents the business. That’s not going away — if anything, it’s more important.

The Semantic Model Is More Important Than Ever

Tommy’s closing argument: as AI agents become the primary consumers of your data (not just humans looking at dashboards), semantic models need to be built for agents. Well-structured models with proper descriptions, relationships, and measures become the foundation that agents reason over. A marketing team’s agent can analyze email subject line performance because the semantic model was designed to expose that data clearly.

The future isn’t raw data + agent = insights. It’s curated semantic model + agent = actionable intelligence.

Episode Transcript

Full verbatim transcript — click any timestamp to jump to that moment:

0:07 [music] Hello everyone and welcome back to the Explicit Measures podcast with Tommy and Mike, welcome to 2026. A world of a new world of AI and AI related things. I feel like we said that last year.

0:39 [clears throat] It’s I feel like it’s accelerating to some degree, but we’ll see. We’ll talk about it more today. Today, we’re going to unpack our main topic around AI drives fabric success. So, we’re going to try to unpack where does AI live in parts of fabric? Where are Tommy and I finding some success with using various things of AI across the fabric landscape? What are we doing? What are we actually using today? The tools are changing. Things are always ev evolving. What are Tommy and I doing? So, that will be our topic for today. Tommy, I got a couple news items. , admittedly,

1:14 Tommy typically is the one that throws down a lot of the news items on the things you’ll find things. He reads a lot of articles on the internet, a lot of feeds. You have a lot of feeds that come through to you. For whatever reason, the articles or parts of articles that were hitting me just hit right this week. And I added a couple articles this week myself through the news. The first one I’d like to go through here is an article or it’s a very short post here that I found out here on LinkedIn. So this is a LinkedIn of a post here. a VP of data Josh

1:50 Tells us how he’s using Excel and now there’s a new data bricks Excel addin which I thought was incredibly interesting. So I’m going to put the post here. You do have to be logged into to LinkedIn in order to see it. But here is the the post from the LinkedIn post here. So I’ll put that LinkedIn post here in the in the chat window as well. So Tommy, did you have a chance to look at this one? Tommy, what do you think about what do you think about Excel and data bricks? Well, we’re getting that overlap too because data bricks focusing we know a

2:23 Lot more in the semantic model. I don’t know how often you tell people about the feature in Excel where you can connect directly to your semantic model and you could call that govern data too. I’ve shown that I do a class that has that feature. I highlight that feature multiple times. , and quite often people come out with this , hey, I have a data table and I’d like to share it with the rest of my organization. Well, yeah, you can do that. You can groom that table of data with a one button click inside the semantic model. You can publish it and it just becomes exposed inside your Excel document. And people are like,

2:55 “Oh, really?” So I Yeah, I’m I’m more and more Tommy I’m touting that feature. Yeah. And that’s something anytime I know a client touches touches Excel which spoiler alert they all do. [laughter] Yes. no we we always tout that feature because it’s one thing to work with data in Excel it’s another thing when especially the feature in the Excel PowerBI plugin is it’s not just a pivot table anymore. If I actually just want to expose the data itself and direct connection I can do that. So the data

3:28 Bricks plugin here it’s interesting because they’re t taking the coattails from what Microsoft’s doing a bit. Yes, I would agree. And this was interesting. So one first off I’ll say I like the feature. Second feature here is I look at this going this is a way for data bricks to skip the semantic model. This this this fully can cut out. So think about it. Everything you’ve done up to this point to get data in Excel, even if you were doing it in data bricks, the lakehouse, whatever that may be, right? All that data needs to get landed in some some lakehouse and

4:03 Eventually, not even a lakehouse, but it has to be landed eventually in a semantic model and then from the semantic model, you can then check the box or write the query or connect to the semantic model and the data shows up inside Excel. Right? I think data bricks, they’re not they’re not stupid. They know Excel is one of the largest I’m going to call it computing languages in the world because it is it’s a language. You write formulas like it is its own system, its own ecosystem. And there’s I don’t know how many users are out there using Excel. A billion. I think more than outlook actually.

4:35 I think it’s a ton. I don’t know. Maybe more than I’m not sure about Outlook. I know a few companies who they use Excel but they use Gmail for the other things. So Sure. I think that’s that’s one there. To your point though, Tommy, like everyone at some level uses Excel. And I’d even argue a basic skill that you need to hire for any person going into organization, they better be like go through some Excel training. Like if you just don’t know how to use Excel, it’s that’s not an excuse. You have to know how to use that product. And it may not be the exact same thing as like Google Sheets or the other products that are out

5:07 There. There’s I think Apple’s got their own version. There’s some open versions of Excel as well. But the the the concepts transfer I think very well between any of them. Anyways, regardless having data bricks do this, this allows you to connect directly to the unique catalog. You can select data directly which is like select a table. You can select a single query and you could write SQL to go select the data for you from that table. Now would be really cool here if if data bricks would like let you have a little bit of AI in this and like describe a SQL query you want to go get and it would just do

5:40 It and write the write the SQL for you and you could check it thing but maybe they’ll get there but regardless this is really interesting and you have a limit set. So in the images that are sent here in the post article you’ll see there’s a limit to number of rows of data. So that’s also extremely important because that way you could say limit my data to a th00and 10,000 records of data. Even though you’re grabbing a table that will likely not fit in Excel, you can at least get some of the data back out. And it’s interesting too because it’s one of those another example of one of the products coming to somewhere else

6:13 Rather than than that old focus that used to be on no you have to live in our product, right? Especially something like data bricks like you had to go to data bricks. Well, obviously there’s this big focus now. We’re coming to you. And I I like it. I like it. It’s especially for the companies who are using data bricks. , and if I read it correctly too, Mike, the focus is not for the developers using this as much as correct. As you said, your CFO and the analysts, so it’s already that productionalized data. Yes. Correct. Okay. Interesting. Yeah. I’m liking this one. I think this is a really good thing. the more the more

6:46 Microsoft can or sorry more data bricks or anyone can make it easier to get data into Excel I think it’s just going to make people happy and so I think I’ve really I thought this so one I I like the article two I love Excel I’m a fan and then three I looked at this going wow this is a direct shot across the bow at Microsoft what’s going on here so this is just loading in I think just a data table directly into the tables here it looks Like there’s also some UI here. Again, I haven’t played with all this new UI here.

7:19 But you can select an asset. It looks like it’s also doing some some form of a pivot table as well. So, looks like it’s a pivot table option. So, you can get the table in and you can do some pivot table, which again, I would argue if you can’t do that, right, if you’re going to build an addin for Excel and you can’t bring in a table or a pivot table of data when you aggregate on some columns, it’s not worth my time to play with it. like so this this feels like a really good alternative solution to what Microsoft is providing with their integration. So highly recommend the read. I know you use Excel. Everyone

7:53 On this call, everyone on this live session automatically. Go check it out. You you may like this one. Not everyone may use data bricks, but if you do, this might be interesting alternative for you. I really want to go to this. Yeah, I want to go to your next one. I thought that was a bang. Okay. So, , next item agenda here. This is a bit more philosophical maybe a bit in nature here. , this is from Mark also on LinkedIn. Again, you can tell I can spend some time here on LinkedIn as well. But on this post here, he starts discussing something I find very

8:26 Interesting. He is an open- source developer producing open-source content. And I think he pointed out a couple things that I maybe I wasn’t aware of. Maybe I wasn’t really paying attention to Tommy, but he goes he’s he’s taking all of his nude code projects and they’re now going to be closed source from now on. He’s not going to allow them be open source. The projects are not open. He’s not going to let everyone see it because he feels AI agents have scraped all the documentation, all the open source

9:00 Things. And what’s happening is the agent has now become the expert. And so organizations that did open- source projects, the way they got money, and he uses this example of Tailwind. Tailwind has been hit really hard. It’s a it’s a web based stylization CSS styling of a website open source library. Well, what he said is in Tailwind, they had to cut their developers by a high a high amount. They’re firing people. They’re really shrinking their team because they’re not seeing Tailwind the company that had the

9:34 Open source code. Here’s the code. It’s all open source. Oh, if you need help, if you need documentation or access to docs, that’s where you pay for premium. Well, once the agent goes through all that, the agent can understand everything and it doesn’t you don’t need that anymore. So, now you’ve just given away all your IP and everything and all you’re going after is an an attention mindset. You need people’s eyes on the code. You need people having to use it. you need them to to use your code and that that’s what gains it popularity. So at the very end he gives a prediction which I thought this is really the point I want to hang on here. My prediction he

10:07 Says a new marketplace will emerge. It will be built for agents. You want your agent to use Tailwind Prisma whatever other open- source library or now closed source library you wanted to your agent to use it pay for access. So your agent, you pay for your agent to have access to all the information and then you can utilize everything just as you would. It’s now a closed project and agents are the only thing that can then get gain access and do all the things on top of it. So I think he’s very right on this.

10:40 Oh, really? I think I think I think he’s right. I think these companies are not going to people aren’t going to build just a lot of software and support it and fix bugs for their goodwill of their own heart. They’re in it to make money somehow. So, how do you do that? He’s talking about libraries with APIs and meters in order to pay for things. Pay for what you use, right? If you use it heavily, you pay a lot, a little bit more. If you use it for not a lot, fine. But there’s got to be some way for these companies to make money. Anyways, I’m just going to pause there. What do you think, Tommy? what it reminds me of? It reminds me when we gave birth to the

11:14 Internet on mobile phones and the iPad and how that completely disrupted newspapers. , very true. So, this is I see a very big parallel here where everyone had they had to adapt big time because they realized that no one wanted to buy a newspaper when they could access the New York Times or whatever it was on my iPad and it was a better experience than newspaper. So, immediately the first thing that they did was try to was a payw wall but they put a payw wall up. Yep. They put it up really like you could read two articles. Well, guess what? That didn’t work well.

11:48 So most or and I think this is one of those things that we do have to adapt. Sure, but in theory I can see what he’s saying, but in practice, well, you got to pay for every source and who gets paid first, like if you’re doing that API call. So now I have to set up my credit card or my debit card for each service. Am I doing this through if I’m using Claude, do I pay Claude or do I now have to sign in at Tailwind for every source? That’s not going to work. That is not I I I I don’t see that as the way it’s going to

12:21 Go. I think organizations, you’re right, they’re going to have to adapt. What the answer is? I don’t know yet. So, , someone’s going to need to pay for this stuff to stay open and usable 100%. Like, that’s a given. I’d also argue, Tommy, I think you bring up a really good example around like New York Times or other these news articles that are like putting pay walls in front of things. Yeah, they don’t like it. But also, it’s been a very poor experience. They’ll give you like the first paragraph or two and then they’ll put up a little payw wall thing behind it, but then sometimes people work around it. Sometimes it’s

12:54 Just if it’s on the internet, it’s out there and the way you monetize on it is not done. And now you see like, okay, there’s other news sites that said, look, we’re going to make money off of all the advertisement that shows up on the page. So you scroll through the article and like every two seconds, every paragraph, there’s like another ad, another something that shows. So they make their money that way. So it it’s going to be one of two ways, right? And AI pieces don’t really care about ads. They can skip right over them. It adds no value. AI agents aren’t going to accidentally buy an ad on one of these open source projects. So I don’t think

13:26 That’s the way to go. I I really do think there’s going to be some form of like pulling things back and very effective stuff is going to have some payw wall. Now, to your point, Tommy, how it gets implemented, I think it’s debatable, right? I I don’t think the way that the news companies or the publishing companies have done it has done a good job of putting it the proper payw wall in there. Now, what I could see, , you’ve heard of agent skills, right? How that became like an open source for every product. It was just cloud initially. So, little background, if you’re using cloud code or really any IDE, they have a thing

13:59 Called a skill, which is basically marked down in other files that you can add in. Claude developed it, but it hit so well that this is now in really every AI tool. Every AI tool is doing the same, right? So I can use the same skill. Well, make us an agent store that goes for every product. So you want the Tailwind agent that’s more than just accessing it, but knows the context. That’s an idea, but yeah, I’m I’m not setting that up. I have other things to do, but yeah, I think this is disrupting like Google’s disruptive right now because no, not as many people are doing search. It’s not just a individual

14:33 Company problem. I want to call that I want to call that one out, Tommy, because I think and interestingly enough, you say Google’s disrupted, but yet it’s like hitting all-time highs on on So, and I think Google very quickly is pivoting themselves over to be more of an AI shop now and getting people to pay for things over there on that side of the front. And I do think to your point, Tommy, like every Google search I do now has a whole AI generated something summary at the top of it. So Google has fully adopted, look, if we don’t put some AI summary on top of our search, our web search, no one’s even going to go to

15:07 These websites anyways. So more and more I’m using more of that AI summarized feature, expanding that and using that and then potentially clicking into different links or testing a couple things. Have you even seen a paid search ad or noticed one, much less click on one in the last six months? I think I’ve noticed them a little bit, but again, have I clicked on anything there? probably not. I probably haven’t done as as much there. , yeah, it’s it’s definitely much more in the I use a lot more the AI to do searching

15:40 For things. So, I would agree with you. It’s definitely it definitely has changed their business, but I think I’m looking at the stock market price of Google. it’s like passing other companies now and continuing to grow and having huge upside to things. So, yes, it’s hurting, but I think they’re pivoting to the right things. Yeah. all centered around AI. 10 years ago or actually 15 years ago now, Mike, when I was marketing, digital marketing analyst, 70% was like 65 70% of all Google’s revenue came from paid search. That’s not the case anymore. Now you say I use nano banana,

16:14 Which no one knows what that is, which is Google’s image gen more with fun names. But yeah, no, I I agree. I disagree with the article or the the post because [snorts] the implementation of it, I think, is going to be the barrier to entry of that. So, let’s let’s keep going down this route of the AI and things that are coming out right now. , one one thing that just I think really hit the market hard here and again this is very recent. , and I also saw some other interesting articles on this one as well. Have you heard of Tommy Claude Co? Have you heard about this one now?

16:46 I think that just came out yesterday or today. Oh my goodness. people are already just getting their hands on this thing and already they’re saying it’s changing how they’ve done work and they’re it’s providing immense amount of busy work reduction on their on their workflow whatever they’re doing. , it’s so it basically it’s it’s this idea of Claude has been this system that’s been doing a lot of programming and AI things, but Claude Co-work is trying to get Claude to do things by talking to it, but allowing it to to not have to have a

17:19 Prompt. It it’s a lot it’s easier for you to get stuff done around your computer. You can still talk to with a prompt, but you’re you don’t have to write command line. It’s it’s not lines of code. , it’s taking more responsibility and showing you progress on it as it’s building things. I’m really excited about this one. I think this is amazing. I There’s a couple demos on their website that you’ll see here. , someone has a whole bunch of icons scattered across their home on their computer. They say organize my homepage and it builds folders. It it organizes and sorts things. It knows how to interact with your machine, your local machine. It integrates directly into Chrome, Google

17:52 Chrome, and does things there. It can do automatic searches. , I really want to start looking at this one, Tommy. I know you get a lot of LinkedIn messages. , and tons of them are some of them are spam. Some of them are selling me things. Every other day someone’s trying to, hey, you want to sell your company? Like, no, no, I don’t want to sell my company. Like, get out of here. , [snorts] but I don’t want to deal with all that. I want some agent to go through and clean that stuff up. And I think could do more of these things. , so I’m really interested to see where I could apply this and how I can get

18:23 Well it to help me out with my daily workflow. You’re talking about two different things. either the Chrome extension or the cloud extension on your browser. Yes, cloud co-work is more it is actually runs in cloud code which is a terminal thing but you can do things such as or like you said organize all your files, process receipts batch rename everything but I want you to try something today for me. I want you to give Claude because this is the birth of cla co the co-work feature. Go to Claude. Yeah. Fill some of your templates of the design you like in a presentation and

18:56 Even some other helper files and say, “I want a presentation. It should cover this is the agenda.” 20 minutes, maybe less than, and you’ll have a a pretty nifty presentation, which we can now do with cloud code. I can have a folder of all my demos, of all the things I’ve done. Say, “Hey, I need a full module from all this. Build a presentation.” and you can walk away and let it run. So, and it runs Yeah, runs on your computer. It can help with spreadsheets, with formulas, it can help with presentations, do analysis, and or again just organize your file or just

19:31 Help us, put together a project plan based on all the PDFs that you may have from a scoping thing. It can so it’s pretty incredible what it can do. I would agree with this. And so this is one of the things my complaint about agents has been for this longest time a lot of these AI things I’ve been producing and they’ve always been about like providing information to you but you’re still really required to do the work to do something right. I really just want to give something minimal effort and really this to me there’s this balance between like how much description and detail do I must must I

20:05 Give the agent for it to understand what I want it to do to get it to do something right. So to your point, Tommy, like I’d like to sit back and and I’ve done this to your point. I’ve made some presentations. I’ve done some stuff and from that I’ve said, “Okay, I need to build a slide deck. I need to build some information, but I know what I want to talk about. Here’s my outline. From this outline, make a six slide presentation about this or do whatever.” I I really think PowerPoint is dead, honestly. It’s [snorts] picking good images. It’s

20:38 Taking the bullet points that I had. I can do the thinking about what I need to talk about. What’s the orientation of like the communication I want to produce. I don’t want to pick icons that make sense for this slide. I want to I want the agent to figure that stuff out. Like it it just needs to be pretty to look at for someone else to see and have a an interest around it. So, with minimal direction, I should be able to pump out tons of slide decks and just focus on the communication, the human part of it, right? A lot of this other stuff is just I would call busy work. I don’t want to do it. I have some homework for you. This is

21:10 Going to change your life. I just I need need you to spend a little time today or this week to try it and then come back to me. Sure. I want you to go to Claude and you I’ll send you some links, but you turn on this capability called skills and you’re going to ask Claude for something that you do that you would wish you didn’t have to put so much effort in. Ask it to write a skill. you’re going to write a skill for me and [clears throat] it uses something called a skill creator. Let it do that for you. Set it up and then just give minimal effort and you say you tell Claude, “Hey, use the

21:43 Fabric skill that I’ve created around Timol. go through this folder and give a comment or whatever you need it to do or ask to write a presentation off that skill and you it has the context.” No matter what, Mike, all these AI agents need the context if it’s going to run efficiently. Now, that being said, you can set up something that is in sense you can pull from so you don’t have to rewrite everything. And that’s exactly what a skill is. Let me let me So, I I I’m with you, Tommy. I’m on board.

22:15 I’ll definitely test this out and I’ll see if I can get some skill things going. Remember Tommy got a long time ago, and this is this is something Tommy and I were talking about as Tommy was beginning his business. I said, “Tommy, I write all of my statement of works in Markdown.” Oh, that’s a great idea for a skill. Sorry. Go ahead. Well, [laughter] I’m I’m building I’m building on your idea here. So, Tommy and I both run consulting businesses. We do a lot of work with code and and talking to customers and writing everything down. And so, one of the things I told Tommy very early on, I said, “Tommy, look, I’m not the best at this.” And this is, I worked

22:48 For another consulting firm before I got my own business. I started my own business in 2019. Before that, I was even saying, “Look, I the the company that was I was working for gave me like a word document and said, “Here’s your template. This is the thing you must use in order to produce like a statement of work.” And I thought it was so timeconuming because I was manipulating borders and trying to add some standard text here. And I could and the thing was it was so annoying to me because the information in that document at the time was stuck in that Word document. I could put it on a SharePoint page, but I couldn’t

23:21 See other people’s estimates. I didn’t have a library of estimates. I didn’t have all the things I was building to produce estimates for customers. And when you do estimating for stuff, you typically find that people ask for the same stuff over and over again. And you just stylize it for their business, their company. And maybe there’s some common aspects, but there’s maybe a couple tweaks of things here or there. Well, to the end of my story here, I’ve been telling Tommy since the beginning of the business, write all your statements of work in Markdown. This is knowledge that you’re

23:54 Using. Some of those will turn into sales. Some of them won’t. But you’re you’re spending time thinking about what does the customer need and how do you write a accurate project for them? Basically, a project plan on how to get that done. And so, here we are now. I’m, seven years into this experiment of running your own business. And I’ve got a whole bunch of markdown documents of previous jobs that I’ve done, I’ve worked on, I’ve estimated. All that knowledge is now training material for a claude code. Like

24:27 I’m so angry about this. So the to your point Tommy about the skill, right? The skill should be, hey, you’re an architect trying to sell a customer on XYZ thing, right? The skill should be use these other documents. all the other estimates you’ve done is in these folders. An estimate follows this form and it has this, this, this, and this. These are what the sections mean. Mhm. When I tell you, I’m going to ask you to do another estimate. I’m going to give you the name of the company. This is what you need to fill out. I’m going to give you a a a general story. And this

25:00 Is something else too, Tom, that again, I’m not there yet, but I record all these meetings with customers about what they want. I have a call, a pre-sales call. That whole transcript can be fed in as input to my little AI clawed skill code. Okay, let’s have a pre-sales call. What do you need? And I can take all my history of stuff I’ve already built, the transcript, the video that’s already happened, stitch all that together, and then out pops another markdown document of what the customer wants. What are we trying to sell them? You then you review it and hand it back over to the

25:32 Customer. I’m so angry I didn’t think about this because the skill is much more than a markdown. You Mike, you could have this folder called architect skil and in it you could have markdown files for previous engagements. You could have one called transcripts say hey use the transcripts folder. Look for a relevant one around data governance. Okay, here’s all the markdown ones. So all you have to say is we’re looking around this many hours. It should fill in the price. You give it the rates. Boom. Done. Minimal effort. I am so mad I didn’t think about this because that’s I I still do all my proposals in markdown too. This is dude dude

26:07 This easy peasy. All right, there we go. There’s our there’s our homework. There’s our to get a collaborative homework. There we go. But again, it’s being able So again, let’s talk about like what’s going on with like cloud code work here and stuff like this, right? So the idea of this is like you already are doing things that are manual. You’re touching lots of computer programs. You’re doing lots of things to give visibility to these agents for them to be able to do actionable things. And let me go back to my very early comment here which was a long time ago, Tommy. We’re talking about agents and they were just providing information to us. We’re tipping the corner or tipping the edge

26:39 Of agents aren’t just being able to see just code. They’re able to see images. They’re able to look at your computer. They’re able to see screenshots of things. They’re able to look at. So the agent is being given almost like eyes at some level that’s helping me build more stuff around this. So be if if the agent was writing code or building tests or backend services it’s really good at that because it writes code it [snorts] runs code returns text of did it run or not run or there’s any errors

27:11 It comes out or doesn’t come out right so everything that the agent need was already in text form as soon as we start going more towards like building websites or front-end development or working with a document or things that aren’t like working with video or transcripts those things aren’t as easy to give give over to the agent because it has to transcribe or do certain things for it first in order for the agent to understand. Now that these tools are getting better, we’re able to give it more visual information and from that it’s able to produce better results. And now it’s it’s and now it’s

27:44 Just making stuff happen, right? Like to your point, Tommy, I don’t have to make a PDF. I don’t have to make a document anymore. I can just tell it this is the output of what I want. Build this thing. Amazing. Ah, dude, good articles today. Good news. Yeah, I think there’s a lot. Again, I I think we’re going to have a lot of these new conversations, and this is I think really going to lend very well to our main topic today, which is, okay, we talked about co-work, we’ve talked about Claude, we’ve talked about other agents, we’ve talked about open source projects and having agents and AI agents touching those open source projects. So,

28:18 Now let’s go into our main topic. Now, I think Tommy, dude, you another doozy. How does all this fit with AI inside Microsoft Fabric? Like what AIs can we use? Which ones are you finding success with? So, let’s just pause there. Any opening thoughts, Tommy, here around like AI and fabric. , first I just got to say the tagline of the Simon Taylor wrote this article and I love his headline here. A AI is rewiring the economy. It’s cheaper to cover a hole in the wall with

28:51 A flat screen TV than fill it. Stuff is cheap, services are expensive. Expensive AI is about to fix that. And what we’re dealing with here, Mike, is the core argument here is we have a metaphor that AI is going to do to services. What we have are the services that companies and CL or firms provide to what containerization did to goods is simply cost reduction. right before AI and when you had to do a service you needed someone specialized in that area

29:24 Or repository of information with the experience especially if it was even from marketing material to if you’re nice designs to even what we do well AI is changing that where the services why don’t we just talk about the first 30 minutes of this conversation services work presentations nice designs are even down to our proposals are things I know AI is going to handle and to emphasize more of a disruption and for me when I

29:57 Look at this with be business intelligence and especially fabric it’s going to help us in the day-to-day this goes back to what are the services and you think core services that usually we provide or just anyone in the powerbi space it’s writing tax measures it’s semantic models it’s the dashboards now with fabric obviously setting up the lakehouse creating the the ingestion and the the engineering side of it. Well, we’re getting more and more to the point where AI is just simply helping us write the code to a point where it’s

30:31 Going to really be a major part in doing it for us or doing it instead of us. And I think that’s where we’re beginning to see more and more of our space as well, not just AI in general, but really how it’s creeping into the business intelligence space. And I let me let me just give you some things that I think is happening, Tommy. So one one trend, these are just trends in general as I’m talking about AI and where this fits, right? One of the trends I’ve seen is Microsoft trying to push AI

31:06 To provide the insight to you. Talk to the AI, have the AI produce the insight or give you information about your data or the semantic model, whatever that is. Okay, I find that to be mildly successful, not super successful. It doesn’t wow me in any way. Like it’s it’s basically producing tables, a little text summary, but I don’t find that to be too actionable when I’m done with it, right? that doesn’t seem to really resonate with me. Other trends I’m seeing right now, I’m seeing people building SVG visuals. I’m seeing people vibe code a custom visual

31:40 For their stuff. To me, that feels like a better placement for where AI should be resolved. And and let me give you why. It’s difficult to learn SVG and attach that to what data you’re building inside your report. But I can describe what I want it to look like. I want these bars. I want this color. I want the edges to be rounded. I want this bar to be this color. I want bar that bar to be that color. I want a line that has a target value. Here’s the value of the target. Right? I can be very descriptive of what I want the visual to look like and do that. I can figure out where the gap for

32:16 Many developers is they can’t go from the vision of what they want something to look like or maybe you’ve seen an image. Maybe you’ve seen something on the internet. That’s a really cool table. That’s a really cool bar chart. I’d like to take that style and mimic it on top of something in PowerBI. That does not exist today. So, I think AI is much more impactful and and wows me a lot more by having the AI help me build the visual. Create it. Make it from scratch. Like what that to me that’s lowering the

32:49 Barrier for people creating visuals. Now, let me just caveat this with something. Microsoft has done a very poor job of being able to like Tommy, let’s say you figured out how to make a really good-looking visual, whatever that may be. There’s maybe some DAXs related to it. There’s conditional formatting. You’ve got the the chart looking really nice. You got the style, the properties all set correctly. How do you share that? Mhm. What do you do to give that to somebody else? And today, the way the solution lives right now, there’s zero sharing of styles on individual tables and bar charts. You can’t do it. there’s not an

33:23 Easy way of doing that. And so to me, that’s one of the major misses that I see in like PowerBI is a lot of like it’s a very rich tool. It lets you build anything you want, but unless how to build the things, you can’t make visuals show up and they won’t look the way you envision them in your mind because there’s no way of communicating a shared knowledge of how to build that visual. Does that make sense what I’m describing there? Yeah. Because as you’re talking, I’m thinking, well, there’s a way now with cloud code, but I I digress. No, I I agree. I think

33:55 But to your point, yeah, this is where agents can make a difference, right? So, so this is my point exactly, right? Claude code or other coding agents, right? That’s the area where we can start stepping in and say, “Oh, we can use these agents to help you build the initial visual.” And then what you can do is you can then hopefully help implement it. Help it implement that and then like what about the prompt, right? So let’s let’s let’s go backwards a little bit here, right? You have a desired visual vision in your mind for this visual,

34:27 What it should look like. I want it to be this bar chart with these styled properties on it. Unless someone else gives you the answer for that or you build it into the JSON theme file, you’re not getting there on your own, right? I can’t I can’t give PowerBI an image and get back an equivalent visual of the same sort. So in lie of that you could use cloud code to help you build that and the prompt you use at the beginning can produce the visual. So you almost need like a secondary prompt after that. Okay, once you get the result go back to cloud code and say

35:02 Give me another prompt that would give me the same visual. Well, dude, well, you could even go a step further and you can have its own agent where, hey, give everyone in this workspace this new visual I created. It’s not and it’ll just add it as a new custom visual if especially if you’re using a git repo, right? You could just say now it’s available to everyone in that workspace because my the agent just implemented that or integrated that as a custom visual. Who’s to say they can’t do that? But that’s it’s a bit of a niche. I don’t want to say niche use case but I

35:35 Think what you’re talking about more Mike is these tiny services that I think before AI because I think the scenario you’re saying is absolutely something that the what we have today is capable of it’s a service that before was previously someone skilled in that area that would have to implement first they would have to create it develop it make sure it’s right and then there’s that whole other step of helping the other reports adopt it which was all complete manual effort. Well, we literally just walked through brainstormed a way to do

36:09 That which would work. So, and I think when you when you think about companies, why am I I always go back to a question of like why is someone going to pay me to do what I do especially in this space like yes, who’s paying a business intelligence developer or an architect? What’s the value they’re getting out of it? This goes perfectly with the article because we’re talking more about less about the services I provide about, hey, I’m going to build you 15 dashboards. It will have 10 DAX measures. You’ll have three reports, but you’re getting a professional to do it. Fine.

36:41 Okay, great. Well, if someone told me that now and I was a company, I’m like, okay, well, what DAX measures because I can get it pretty close here with if I’ve developed my own skill or if I have an agent set up to do just that and understanding the data sources, etc. And I think a lot of those services when we’re talking about even updating power query transformations, helping write the SQL for lakehouse pipelines, , vetting and cleaning a data model, a semantic model. Who’s to say we can’t have an

37:15 Agent to do that or we’re getting to that point. We might not be there now. I don’t think we’re there yet. No, we’re not there yet. But Microsoft is exposing a lot of things that are going to give you some I don’t think we’re there yet. But I I see the writing on the wall here is is the market is moving in front of Microsoft at this point. The market is moving towards generating AI based product pieces. Microsoft keeps focusing on the insights being AI generated. I think that’s a miss. I think I think that’s to maybe that’s something useful there. But rather I think it would be more useful

37:49 For us to say let’s step back. Let’s actually have the AI help us generate known visuals. Like what’s the value in this? Let let me let me spitball your idea here Tommy a little bit. Right. AI is really useful for creating lots of stuff very quickly. So if I’m if I’m able to use AI to commoditize the creation of visuals, why not let AI make 30 visuals and I pick the one I like? Like the time it would take me to build 30 different visuals of the same

38:21 Information. Difficult, right? I should be able just to throw that information at the wall and say, “Here’s the table of data I’m looking at, or here’s the model that I’m looking at. I need a visual to do this, let it build 30 versions of it, and then I pick the one I want. Boop, that looks good enough. I’ll slap it into my report. Make that a standard across all my reporting. No problem. It becomes a standard across all all the reports that you build.” That to me, that’s where I think things get more powerful. And I think this weird there’s one thing about PowerBI visuals that are difficult

38:54 To work with, Tommy. And I would, maybe you can weigh in here as well. It’s the conditional formatting. It’s when I need DAX measures specifically designed to support that visual. That’s stuff that doesn’t easily transfer because it has to be part of the semantic model or it has to be part of a what’s the other visual calc, right? if it’s not there, you can’t use it. So, as we look forward here, and I think you’re speaking to a really great point here, that the mode that we’re going,

39:28 Especially I’m taking this as a consultant, but honestly, you can take this even if you’re FTE. Sure. When we look at the future, when we look at where everything’s going, what does it mean to be a professional then? And and I’ll phrase this a little a little different. It’s not because to me it’s more than just I am really good at DAX or I can develop DAX. I’m going to develop a semantic model or lakehouse and I can build that end to end. To me I would answer that question is rather than trying to compete against those

40:01 Minimal SC or not minimal task but trying to compete against AI and doing my normal workflow is I have to make AI help me scale my expertise. Yes. And is it is it best? So I’m let’s go back to your time comment, Tommy. Yeah. What does it mean to be a professional? I’m a professional. Does it make sense for me to spend a bunch of time working in a visual and changing the legends on or off? Does it make a bunch of time for me to spend a

40:34 Lot of time in there changing all the colors of the axis, digging around for the setting in the general properties? Does it make a lot of sense for me to stylize a button? Yes, that has to be ultimately done to get the output produced. I should just be able to send an image and say, I want a button that looks like this. It should just try a bunch of settings. Here’s a bunch of settings I’ve adjusted and here’s 10 versions of a button. Which one do you like the best? That one. Great. We’ll take that one. That would be more useful. Tommy, I’ve been the one building these power. So, Power

41:08 Designer is the tool that we provide. Yeah, no one likes the new [snorts] formatting for buttons and things inside desktop. No one likes it. there’s this whole on visual editing thing. The menu’s been changed. Yeah. Yeah. Yeah. Object. If we’re a if we’re a professional, let’s be real here. Like, if we’re a professional and I want to adjust settings on a visual, it’s very difficult to dig through all the menus that they have. And there’s nested menus upon menus upon menus. It’s not easy to get through. So, as a professional, does it make sense for me to spend my time

41:40 Doing that? No, it does not. It doesn’t make sense for me to dig through the menus. So, where does AI make sense? I’m going to actually now probably bring some real value to people actually 41 minutes into this episode is what tools can you use today that actually let you use AI on PowerBI things? Well, there is one called the PowerBI MCP server. So, I found incredible value from using this MCP server. I threw it I pointed it at a a PowerBI model. I was able to have it make folders,

42:13 , make measures, suggest measures, move measures into folders that I have made, , provide, , every column that has a number attached to it, make a measure that’s a sum or average depending on what makes sense for that data point. It did a good job. It was able to rearrange things. organize my data model. go through the model and write a description for every column and table. Give me a summary of what you’re going to write and then I can adjust it and then we’ll edit and push it back in the model. Like those are things that are really useful from a PowerBI MCP side of things. There’s also a fabric

42:47 MCP that that’s another again you could use whatever agent you want. You hook on the fabric MCP, it authenticates for you and then it hits a bunch of APIs in the fabric experience that help you do things like it literally builds the things that you need that that is where AI force multiplies things here. So, I’m just going to pause on that. I have another thought, Tommy, but I’m going to pause there. Your reaction to like the the MCP side of things in accelerating the professionals use of AI to make PowerBI and fabric. I

43:20 Think that goes align with what we’re talking about with the agent skills. , , I think we’ve talked I think we’re going to talk about it in the future as well, but we we me we’ve been mentioning the MCPS a few times and one of the conversations was is that going to be the future is the first iteration of an agent that’s dedicated to a particular expertise or domain. Yes. And I think we’re just going to see more of that where to your point that is now becoming [clears throat] busy work. them organiz organization of measures the building of measures the sum what do we

43:53 Used to say about that we use tabler editor and then you’re like oh no there’s timole and now you’re like I am just going to go to cloud and just feed it a chat and or whatever the the tool is and so we’re finding those busy work like tasks that are in power and fabric are becoming less and less we’re becoming less and less hands-on to it we’re Technically to your last point, your CFO or CTO could go into cloud and update all the measures and make the sums and organize the model

44:26 Just like you could, right? Because that did not require the only thing it required you to do is install the MCP. But everything else, it didn’t matter if you were your son who wrote that question or you as the seven eight nine time MVP didn’t matter. And this is a big point where I think organizations are going to wise up too. Microsoft is firing I think 1% of their workforce a month or something like that and we’ve already talked about this big disruption. So that’s in your article too. Yeah. Like

44:57 I’m a little bit more skeptical on so I understand that point and I heard I’ve heard that I’ve heard similar things that companies are starting to downsize and they’re not I don’t think it’s I think there’s a little bit of extra hype and again I could be wrong on this one. This is just me speculating. I’m a small person. I’m I’m in my business figuring out what happens here. I don’t see it as as disrupting the jobs you have currently. I think it’s more of like the people that are adopting and augmenting are going to do and produce more. And to your point, Tommy, like I understand this idea of like we don’t

45:28 Need more people. That’s with the caveat. Yeah. Yeah. Right. At some point, you reduce so many people that there’s nothing doing anything. And I think Elon Musk is you love him or hate him. He made a company called Macrohard. So Microsoft and he made a company called Macrohard which he’s trying to tool as a business operation tools company that is only built using AI. AI runs the entire company. Like it’s literally an AI shop of a of a ton of agents.

46:00 Like it’s going to build the software. It’s going to have the product support. It’s going to build the next feature. It’s going to it’s going to have all these agenty things that are going to just build an entire company around things that Microsoft provides. It’s all code. If it’s code, agents and and AI things can provide that as a service as well. So, we’ll see where that goes. I don’t know if that’s going to be a good thing or a bad thing, but let’s let’s come back to the topic at hand here, right? More users, less users, who’s going to be involved with this? I don’t think we’re getting to a place where we’re absolving users from the system. I think what we’re doing is we’re using

46:33 We’re we’re removing a lot of busy work out of the system. And I read an interesting article from someone who turned on a lot of agents for their engineers, their team of 10, and they thought, okay, my team of 10 people can now produce like 50 people. We’re, we’re 5xing what they can do. Well, what wound up happening is you just moved the bottleneck from people producing code to people approving pull requests for people approving the content that’s being built. Right? So, let’s think about this in the world of

47:05 Of fabric and developing reports, right? I can I could produce a lot more, right? Let’s let’s Tommy, you’re a professional. We’re going to give you claw code. You’re going to 10x your ability to create stuff. Great. Well, you’re going to get those reports into QA and you’re going to hand it to the person who was asking for the report. You may be able to produce that and turn around in hours as opposed to days and you may have seven or eight things for them to review. At some point, the reviewer, the other side, the person who’s asking for the content, they’ve got to say, “Is

47:38 This good? Do I approve and move forward?” They’re now bottlenecked. And you can’t you could potentially put some agency things in the way that to help streamline that, but you really need people’s eyes on top of it. Does it look right? Does it answer the questions? Does it actually is it really meeting my needs to do my day-to-day job? And so I think what we’re going to find is what agents are doing is it’s just pushing the bottleneck somewhere else in the stream. And what this article was talking about was they had 10 developers. They were producing like 20 or 30 developers and they found the next

48:11 Bottleneck was the review process which is still gated by humans. Still needed people there. So they shifted people from development over to QA because they needed more people to get through that part of the bottleneck. They could produce more but where the works work level sits it had to be shifted. So, I think I think Tommy, we’re it’s it’s going to get better. You’re right. I think we’re going to be able to do more, but I think we’re also going to create a lot more and that’s going to create a lot more human intervention needs. I I think I want to focus on this point because I think this is for people listening especially where you’re going,

48:45 Hey, what what’s next for us? Well, this is where that domain expertise comes in and more importantly this idea of the intelligence developer because it’s not going to be so much my selling point to an organization is hey because AI rather than doing five reports I can do 15 is you still need that domain expertise both in the product itself but more importantly for the company and their their own context on what they’re looking for and I can scale but there to your point there still needs eyes on it,

49:18 Human eyes at this point. , and to validate it, to make sure also it’s what they need and I can do rapid iteration with AI, which is something when you think about all the problems we’ve had like, oh, if I only had more time, I could do X. Well, now rather than that sandbox phase of building something is, whatever your normal time frame is, we can rapidly develop and focus on that one product. And for me, I still need that expertise at this point. I’ll go back to the example I said what if your CTO was using cloud to organize the measures okay does do they know it’s

49:52 Correct is it work is it working out what do we need to change my role is shifting to rather than it matters that I know all the little ins and outs and that I am doing it myself is that I can build this in a scalable fashion validate it and do understand what iterations or changes that need to be I can do more things at once, but I still need that expertise. And more importantly, Mike, I’m not selling myself with the ser like necessarily the technical services I provide. I have to

50:28 More than ever sell the outcomes that will be result of hiring me or working with me. That’s what I have to do is like we’re going to not just build 15 reports for you, but we can scale and make sure that everything’s perfect. we can make sure that exactly what everyone needs is going to happen on this time frame. , and that from the actual me writing DAX, I shouldn’t have to say that anymore. Yeah, that I guess that’s going to be assumed, but we’re getting to a point where the agents are going to accomplish a lot of that. And I think

50:59 Yes. Yeah. I I want I like I like where you’re point you’re going here, Tommy. It’s going to be more about the thought, right? So, I think where people are going to need to apply themselves is what is the process, right? There still needs to be the liaison between the super technical IT shop people like here’s the data that’s coming in the door and here’s the visuals and the needs of the business someone you can do some of that with AI you can maybe ideulate some things but like at the end of the day someone’s got to really put together those two pieces of the world and interpret like what’s going on I may not be the one building the visuals but I’m still going to be the one

51:33 Documenting what table is the is the description right that table description needs to be made by someone, I need to explain what that means. That still requires people input. So, I think what we should be focusing on is where can we apply people in concert with agent e things that help us give better semantic models, more descriptions, , match the business requirements down to data that’s coming in the door. I still think there’s a lot of data mess

52:05 Honestly. I I think there’s a ton of money to be made in still helping organizations just identify their data sources and get them into the same place. Tommy, I talk about this all the time on the podcast and I’m I keep complaining about it, but I haven’t solved it yet. So, if anyone in the chat wants to solve this for me, I would really appreciate it. We’d love for you to just resolve this for me. But I go to all these different social media platforms. None of them talk together. None of them have aggregated an analytics together in any shape or form. And I spend probably too much time just digging through stuff. I don’t know

52:40 Which content pieces hit the best, the the are most impactful. I just can’t see it easily. And so for me to see analytics across all media platforms, it’s a it’s a it’s a representative sample of all organizations. They’re going to have a lot of data silos across their whole org. How do you get it together? That’s that’s where I think you the the the human using AI is going to be most essential. 100% 100%. And I think also just allows me to scale.

53:12 Am I doing the exact same thing? I I really have to start thinking more Mike about the services that I don’t know how to do or would just take a lot more development to do is that I need to just also provide that where I can I can start talking to companies and start asking better questions on what they want to do with their data. And I think that’s where my selling point is is not just we’re going to do your normal services but what are you really trying to do? Oh, you need that in application. I’m not I don’t do application development or push it to something. But guess what? I can now ask

53:45 That question because I I am I don’t have to spend the six hours of doing research on hey did that did that call request work? No, that didn’t work. And I’m doing that research. I can allow that to run because I know that’s something that they’re looking for and I can start asking better questions on what we want to do with the data. One thing I want to note as we get at the end, one thing you and I have not talked about or mentioned and I think this is important is providing the insights here because I think we’re still far away from co-pilot or anything doing the hey you’re up 5% because of product

54:17 X. I think that’s still very much an eye of the beholder thing on what an insight is. It’s up to every company to figure that out. , but but I think another part of our solution though is that insights is still going to come from the semantic model. And let’s let’s tie it all the way back here. At the end of the day, especially if you’re in the fabric ecosystem, if you want insights, right, and you want that agent work on not just the developer side, but your data, this is more important than ever that our focus with semantic models, our focus

54:51 With the data is that it can talk well or it has that right structure to work with AI. And I think this is going to be such more and more of a focus for us too because we can start now if we develop it correctly having semantic models that are geared towards helping the marketing team develop hey all of your email subjects that said buy now well a the agents can look at that start developing the marketing material for for a company coming from a semantic model because the semantic model would basically help you say here are all the

55:24 Emails that worked here’s that 20% And that is where I think our role is going to be more than ever. But that comes from a semantic model, not just raw data. The semantic model is also going to be more efficient for that helping operations say where the bottlenecks are because we can pull in all those records. We’ve filtered it out. We have the DAX measures. So I think we’re more and more to a point where our major focus and selling point as people working in business intelligence is developing the data that’s going to work

55:57 For a company using agent-based tools. I this is exactly where I think things are going Tommy and I 100% agree with this one. Let me give you a quick rundown of things. So we did a little bit talking about a couple tools that we use. We talked a lot about the concepts and like how this is. So, I’m going to quickly quantify. Let’s call this the the 10X fabric developer, right? I’m the I am the 10X fabric developer. What does my toolkit look like? What is Tommy and Mike picking from the library of things that we have access to that we’re using? , I’m going to quickly

56:29 Rattle off a couple things, Tommy, and feel free to add some things in here if you’d like as well. I’m going to say I’m definitely using VS Code, and I’m definitely using VS Code with various forms of MC MCP support. I’m using Fabric MCP. I’m using the PowerBI MCP. I’m using it to help me do simple manipulations. I really do feel like the PowerBI MCP does a lot of the things that tabular editor does. And people I love Tableau editor. It’s a great tool, but I don’t have to write C# for scripting. I just talk to it and it just does what I want. So already I’m already using a language that’s easier for me to

57:03 Like leverage and use. So that’s that’s one that I find very valuable. C-Pilot in Microsoft Edge. This is a free one. if you have co-pilot if it’s in edge today, you can use that. That’s a that’s an AI agent your company doesn’t have to pay for. You can just go ahead and use this. it it’s already available to you. I use that all the time. Copilot. I’ll have it write Python functions. I’ll give it a little sample of of information. I need a function to do this, this, and this. Write it in Python. I’ll have it translate a SQL statement and put it into Python. I will take M code and I will translate M code into SQL. I’ll

57:37 Take SQL and push it into M code. Like there’s all these translation things of like which tools we can and cannot do and I use co-pilot to do it. Like I just translate things. It’s really good at that stuff. So that’s another one that I’m using. I’m using the edge co-pilot to build color palettes. Here’s an image co-pilot. Can you please turn this image into a color palette of 16 colors and hex color codes? Boom. Done. And I’ll also argue my tool that we build power designer which is available today in fabric. You can go use it and try it for free today. If you’re making color palettes, you can

58:11 Use that for free. It’s built into the tool. Go to Copilot on your Edge browser or whatever browser you’ve got. Give it an image. Ask it to build you a color palette of however many colors you need. Copy and paste the hex color codes into my tool. Boom. Theme file is done. You’ve already got that built. So, it’s super easy to get done with. With a couple clicks, the tools are there to make it much easier. There’s no way to do that in desktop. You can’t copy and paste hex codes into the color theme palette in desktop. we just can’t do it. So that’s another advantage. I use Copilot in notebooks. Notebooks has the

58:43 Ability of running C-pilot. Now I don’t love it because it’s a bit expensive. It’s using my capacity. I like to use Copilot from other sources that don’t require any extra cost to my fabric capacity. So that’s one that I I use other places. And then the last place I use is I’m using large language models with VS Code to help me build visuals. Right? So, I’m exploring building SVG visuals, building custom visuals using large language models. Now, am I productionalizing those things yet? Probably not yet, but I’m experimenting with it. And I believe honestly there’s

59:17 A lot of people on the internet that are already using this stuff. They’re already doing examples of this. So, just check check out LinkedIn, follow people that are talking about fabric and agent-based things. People are building a lot of really cool tools that are being built there. Tommy, what other tools are you using around AIS in fabric? You took all you took all the good ones. So, I’m going to add on and from the advanced side of that because you have the sidebar. So, going off your idea of using copilot in edge, not only can you do that, but I would push you one step further. If you use chat GPT or

59:50 Cloud or you already have those tools, you can basically create a side on URL. So, I basically have the sidebar, not just copilot, but I have all the other tools there. And I have projects that I’ve created for each called the fabric guru that I have fed documentation on notebooks and from Microsoft on pi spark etc. and what fabric is and who I am and when I’m working on anything fabric I’ll just pull up that project and so it already has that context more than just the free co-pilot does and we start building very quickly on those things.

1:00:24 To your point with I use cursor but I also I’ve been doing a lot more with VS with cloud code and going back to that point with the cloud skills or an agent well there’s a theme one so I just say hey scour the internet for the branding for this company give it the URL there’s already a template file of JSON from a theme build a theme from this type of idea and not only is going to do the color palette but it’s also building out that theme file but implementing it in that report So, it’s really all the things you’re talking

1:00:55 About, but then it’s those I think spending a little more time on on building out those things that you’re doing every day. So, I I try with most things I’m doing fabric and AI that I have a particular project, context, skill, or agent that I’ve worked on so I don’t have to repeat that information. , and then I use from the GitHub review, especially for all my workspaces that are in git, is we have a whole review process with those workspaces to validate everything all the changes that were made on a new commit.

1:01:28 That way I’m not breaking anything. , especially when I’m using all the the P by the Timo files. So this becomes a validation system. it becomes me to implement everything and also when I am developing something in fabric I have particular projects with the expertise needed I yes these are great so expanding on that like base list there like that’s the stuff that we should be spending our time on like I think that’s that’s what turns us into the 10x fabric developer right we we can

1:02:02 Still do the things we need to get done we’re still talking to the business getting IT stuff figured out but we’re able to manipulate and go through that data faster with these tools. So, you’re calling us 10x developers now? Is that our label? I don’t know, Tommy. I still think there’s things to be desired. [laughter] Maybe like 4x. maybe four, maybe three and a halfx, right? that’s but regardless, I think I think this this is all about figuring out where these tools fit your workflow, right? And I think that’s something that’s difficult

1:02:34 Because I think everyone’s workflow might be slightly different and certain areas you struggle in or not struggle in. I definitely will say this. I have found a good amount of success with using any code generation anywhere in fabric you’re trying to code generate supplementing that with some AI really helpful. Like that that’s where I think that it really shines. building visuals, getting insights on data, the success is there, but it’s not as great as where I see it as it helping me create content. That’s that’s really where I see the

1:03:06 Differences. Create a skill, man, to come back to me. We’ll see. All right. Yeah. This is where we stand today. We should revisit this in the end of this year or next year sometime. Who knows? There may be a whole new world of things that are changing and making this all different for us in another year or so. Anyways, all right. We’ve rambled on long enough. This is a great topic. You’ve seen a lot of our our world, Tommy and I’s world has changed a lot. We still talk about fabric. We still do a lot of things around PowerBI, but you can see by the nature of our recent episodes, we are heavily seeing AI creep into this space and and change how we

1:03:40 Build things. And we hope that this episode has been some food for thought for you around how you can leverage AI to start just adjusting how you do your workflow. Just learn it. It’s gonna have to be something you’re gonna have to push time, figure out how to use it because if you don’t, someone else in your organization or a competitor is going to figure out how to do it and that that will advantage them over you. So, I think that’s why we’re trying to learn this stuff. I would agree with you, Donald, as well. Everything is changing super fast and so stay tuned, hang out with us at the podcast. We’ll try and stay up with the changes as much

1:04:12 As we can and you’ll hear it first hopefully on our channel. We’re we’re pretty much ahead of the curve here. I think a lot of these things we hear a lot of things here. Yeah. And I feel like after we say them, they become like a reality in other places. Doing a lot on LinkedIn. I I really just want to touch on new thing because I’m sure some people are frustrated how much we talk about AI. But Mike, what we’ve been doing since the beginning of this podcast is we talk about what we’re experiencing. It’s still the PowerBI water cooler. However, what we’re doing and impacting us, not just because we’re I don’t think we’re hyping about it. Granted, I do find it exciting. Is this is part of our daily routine now and

1:04:47 Affecting how we do our work. That’s that is no change since day one of this our podcast and what we’ve been talking about. So I would agree with that. All right, that being said, thank you all very much for listening. We really appreciate your ears. If you like this episode, if you found this valuable and it gave you some insights out of AI based things that you could be starting to use, please, please, please recommend this to somebody else. We do all this through sharing and word of mouth. We’d love it for you to subscribe. It’s free. If you’d like to subscribe, if you want all these episodes without any ads, you’re also more than welcome to become a member of our channel. Our

1:05:20 YouTube channel has a membership level. You can listen to the whole podcast episode without any advertisements. We’d love for you to become a member and help support us as well that way. That being said, thank you so much for listening, Tommy. Where else can you find the podcast? Ask us questions. Oh, yeah. You can find us on Apple, Spotify, or wherever you get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. Do you have a question, idea, or topic that you want us to talk about in a future episode? Head over to power.tipsodcast. Leave your name and a great question. And finally, join us live every Tuesday

1:05:54 And Thursday, 7:30 a.m. Central on all of the PowerBI tips social media channels. Thank you all so much for listening and participating with us today. We had a lot of fun. Great topic, good discussion in the chat. Thank you very much for chat people jumping in and giving us great ideas and suggestions. I really appreciate that. We do that because we enjoy our community. That being said, thank you all very much and we’ll see you next time. [music] [music]

Thank You

Happy 2026! If AI is changing how you work in Fabric, we want to hear about it. What tools are in your toolkit?

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.