All About Rayfin – Ep.536

Microsoft shipped Rayfin, and Mike’s opening take is not a small one: this is the biggest thing since Power BI Desktop. It lets you describe an application, and get a real one — running on Fabric, inheriting your existing governance and security. This episode is the first full pass at what that means for report builders, for BI teams, and for the awkward question of who maintains the thing after you build it.

News & Announcements

• Rayfin on Microsoft Fabric — Microsoft’s official landing page for Rayfin, an open-source SDK that gives you a fully managed, code-first backend on Fabric. Database, authentication, and hosting come preconfigured, so you write application logic instead of infrastructure. Apps built this way inherit the governance, security, and compliance already established across your data estate.

• awesome-rayfin — Microsoft’s curated gallery of community templates and resources. It ships nine starter templates you can scaffold with the Rayfin CLI, from a blank app to a field technician app to a Power BI fixer. It also documents the core concepts — data modeling, authentication, and typed data access — which is the fastest way to understand how a Rayfin project is put together.

Main Discussion

Topic: What Rayfin is, and what it changes for people who build in Power BI

Rayfin is not a new visual or a new pane. It is a way to build a real web application whose backend lives inside Fabric, published as a Fabric item alongside your other artifacts. Mike has been building apps on Fabric since January by hand; the difference is that Rayfin removes the parts nobody wanted to own — auth, storage, deployment — and hands them to you preconfigured. Both hosts spend most of the episode circling the same question: if describing an app is now enough to get one, what exactly is the job?

• The backend is the product. Rayfin’s value is not the front end you can see — it’s that security, semantic models, SQL databases, and the OneLake platform underneath are already wired together. You are not standing up authentication. That is the part Mike keeps returning to.

• Model quality decides your results. Mike found the lower-tier models could not hold a Rayfin build together, while the stronger reasoning models understood how to structure the app. If you try this with whatever model is cheapest, you will conclude Rayfin doesn’t work when the model was the problem.

• A style guide is now a prerequisite. Without a framework for visuals, formatting, and layout, the agent produces something that runs but looks unconsidered. Design the style guide first and the output improves immediately — that prep work moved from optional to load-bearing.

• It fixes the table and matrix complaints directly. Every limitation Tommy has hit with expanding columns, conditional visibility, and bookmark gymnastics becomes a sentence you type. Mike’s blunt version: he doesn’t want to build Power BI Desktop reports anymore.

• The semantic model becomes more important, not less. A lake house full of tables carries no business context, and neither agents nor people can infer how they relate. Mike’s argument is that Databricks and Snowflake are adding semantic layers for exactly this reason — the model is what makes any of this answerable.

• The handoff is the unsolved problem. Mike can ask the agent to write documentation for the next agent, but if the client isn’t on GitHub and isn’t using agents, they cannot pick the project up. Tommy names the risk plainly: if the one person who built it is gone, who owns it?

• This is a squeeze on Power Apps. Different licensing, more cost, and now a faster path to the same outcome sitting right next to it in Fabric. Neither host was reaching for Power Apps before, and they are reaching for it less now.

• The skills that matter shift toward communication. Less hands-on-keyboard DAX, more articulating the problem, understanding what the business actually needs, and turning that into something buildable. Tommy’s framing is that soft skills are what make the hard skills work here.

Who should touch this first

Both hosts land in the same place on rollout: give Rayfin to your strongest Power BI developer — the one who understands the data and the people — and give them some AI training if they haven’t done this before. Handing it to everyone at once produces ambiguity and slop. The harder problem is organizational: dedicating someone to research and prototype for half a week is a real cost, and most businesses struggle to justify it against a maybe.

What it means for consultants

Mike’s take on pricing is that speed is the product. If you can compress months into a week, you price for the outcome and the urgency, not the hours — and you stop competing on being the cheapest shop. Tommy raises the token-cost question that follows: on a Rayfin build, whose budget covers the agent? For now both treat it as a client cost, but it is a line item that did not exist on a Power BI engagement a year ago.

Looking Forward

Start with the templates — pull one from awesome-rayfin, build something small against data you already understand, and pay attention to where the semantic model does the heavy lifting.

Episode Transcript

0:22 Hello everyone and welcome back to the Explicit Measures podcast with Tommy and Mike. Hello Tommy. Oh man, we have so much to cover today. How you doing my friend? I am doing well. Things are moving right along. along. Can’t complain today. We have a really exciting topic called Ray Fin. This is a brand new product that’s just been recently released from Microsoft. It’s a fabric item that you can go build full apps inside fabric. Yeah, go ahead Tommy. Tommy. No, I was going to say it’s not really it doesn’t exist in the fabric UI.

0:52 it doesn’t exist in the fabric UI. It is. It is. Is it a Is it? Is it? It’s an item in It’s in fabric UI. So you you publish your Ray Fin item up into fabric. So it actually acts like an item just like This is the neat part of this is you’re actually getting full-blown apps right inside Microsoft Fabric. And people are already building crazy things. I just saw Injae Park post something about a F1 race video thing that’s already published out there on on this example for a dashboard that’s around F1 race cars. Pretty cool. So I

1:22 around F1 race cars. Pretty cool. So I think people are going to go nuts with this. This is absolutely amazing. I firmly believe this is the biggest project since Power BI Desktop came out. Like already 100% 100% Already Tommy, it is so fun to build with it. It is lightning fast to be able to create something that’s decent and good. It’s pretty impressive. I really enjoy building with Ray Fin already. And I think today too, we’re going to talk about like from a developer point of view, but also how is this going to

1:52 of view, but also how is this going to impact BI teams? How is this going to impact organizations building what they want? want? Yep. Already it’s making It’s It’s making waves already, I think. Right, which means it’s making waves in governance. It’s making waves in the management. So, that’s a huge part of today. So, I think there’s a lot of impact here, there’s a lot to unpack, but Mike, where do we begin? Do you want Do you have any news or any items that came out that recently? I think we just dive right into it. I think think really have much either for news items. It deserves it. The only thing I think I can maybe

2:22 The only thing I think I can maybe mention here, Tommy, is I believe the June release of Power BI Desktop has been released. We’ll get to that. So, June has been released and with that release, you can now use the skills for Fabric to directly talk to Power BI Desktop and there’s now this thing called the Desktop Bridge, which you can now use to go talk from your agent to Desktop and actually start manipulating not just So, we could manipulate the data model using the XML

2:52 manipulate the data model using the XML endpoint and using the MC P server, but now there’s a bridge that you can directly talk from your agent directly into the report side now. So, now reports can be modified. Tommy, things are Agents are coming over everywhere. I’m happy to see these agents and experiences getting involved. I felt like Microsoft for the last couple months have just been behind. It was quiet. I didn’t know what was going on. There was like no announcements, there was no push around all these stuff. Now it’s all here. I I’m I’m extremely excited where this is going. Sorry,

3:22 excited where this is going. Sorry, Tommy, go ahead. What were you going to say? say? I was going to say that’s probably going to be its own episode, too, because again, again, dude, Microsoft is on a roll. It It has If anyone watched the basketball game last night, the greatest comeback in NBA Finals history, that’s what Microsoft’s doing right now. They just got the hot hand., any anytime you can attach yourself to something agentic or building with agents or, that just seems to be the right way to go. For the longest time, Tommy, everything around agents and Copilot felt like it

3:54 around agents and Copilot felt like it it was an afterthought. Mhm. Mhm. Hey, we’re going to build this Copilot thing. We’re going to throw it on the side window. we’re going to add Copilot into your reports, and you’re going to be able to ask questions to it and get answers from it. They all felt like a little bit aggressive, like as far as what I wanted the agent to reason about and get answers back for me. Now, I feel like the story has shifted for me. for me. The story has shifted from I use agents to get answers as now I use agents to build products that bake the answers. And I think for me, this is much more

4:25 And I think for me, this is much more effective when I think about how do you use agents and AI? I want AI to build something that runs on cheap, commodity, easy-to-run hardware, and just as reliable. I don’t need the agent to step in every single time I want to ask it a simple question. Right. And And for me, on the on a different angle is really like a It’s been like a partner in my business, rather than just a chatbot or this contributor. Yeah. Yeah. We’re really it’s helping on again, even

4:55 We’re really it’s helping on again, even if it’s in different harnesses, because of skills that can go across different harnesses. Mhm. Mhm. Because of the MCP, I have this continual context, right? Yeah. Yeah. I’m going from Claude to Notion to the MCP, and I have to reintroduce everything. Mhm. Mhm. There is this common flow, which like, all right, take a look at the central location of context, let’s begin the discussion. And to me, that that’s part of the revelation. But yeah, Mike, you mentioned something to hear, and

5:25 you mentioned something to hear, and let’s make sure that we give credit where credit’s due, if it’s true. You mentioned, and this is You’re not the only one who said this, that Ray Fen is the greatest base impact to the Microsoft data platform since Power BI Desktop. Yep. Yep. Let Let’s not gloss over this. And first off, off, I would want to discover why that is. How That’s a big Those are big words. Those are fighting words for some people. And I think first let’s discover okay, what really is Raven

5:56 okay, what really is Raven for a developer? How do you get started and what does it do? Like what can it accomplish? And I think it’s a great part to be Let’s try to sell it. Sell someone listening why Raven is the greatest thing since sliced bread or Power BI Desktop. Tommy, Tommy, when you look at Power BI Desktop and you have that table, Mhm. Mhm. right? The table you’ve been trying to build. build. Yeah. Yeah. And then it doesn’t quite do what you want, right? Automatic column widths or I want to be able to select something in

6:26 I want to be able to select something in a column have it filter something. Or you have you have that matrix, Tommy? Mhm. Mhm. And and in that matrix you can’t expand the columns with a little expand or contract icon cuz in the in the table in the matrix you can expand the row level, right? Yeah. Yeah. All those problems you had, Tommy, working with tables and the features you wish it had, boom. boom. Raven has all of that. Now, Now, you have to tell Raven how you want your table to work. You have to tell Raven to go use a a framework or something about

6:56 go use a a framework or something about a table inside a visual, but Raven has no context to any limitations that Desktop has. Done. Done. You Tommy, you have a a visual a visual that you built in Power Power BI Desktop and you want to change that you want to have rounded bars Mhm. Mhm. in the bar chart. Or you want to use something else special on there. Can’t do it. Doesn’t let you. With Raven, you just describe what you want. It figures out what d3. js to build

7:26 want. It figures out what d3. js to build and builds you the thing. Done. Like that Let me so for me the sales pitch is a lot of the things that we were trying to do in Power BI are now accessible to developers. If you can describe it and you use a good enough agent and I will say this, say this, you do not want to use lower end models to build these projects. You need a better model to reason about what you’re going to build and build good stuff. So, I have found some of the Sonnets don’t quite get it.

7:57 some of the Sonnets don’t quite get it. Sonnet 4 5, Sonnet 4 6, not quite good enough. enough. Chat GPT 5. 4, 5. 5, Opus 4. 6 seem to get it. Seem to understand how to build that app. Seem to understand how to put things together. So, you definitely when you are building with Rayfin, you’re going to want to be able to make sure you have access to higher end, better reasoning, better code development models because it does a better job of building the output. I’ve gotten gotten a couple times where I had to like I get stuck and like ah, it’s not quite working what I wanted and then I had to

8:27 working what I wanted and then I had to go back in and like modify some things and and change it. So, in general, that’s that’s my general pitch. Right. Right. I’m going to do one more feature then I’m going to kick it back to you, Tommy. Yeah, please. The other thing I’m pointing out here is I’ve just described to you things about the visual side in Power BI that we can’t do, that we can’t do in Rayfin. Yep, yep. Yep, yep. Rayfin is all that and a lot more, right? Rayfin is an app, an application, a web application that physically lives inside Fabric.

8:59 You I think right now the mental model is it’s a static web app. So, if you build websites or you build GitHub pages, you could think of it a lot of as a GitHub page right now. The difference between a GitHub page is if you’ve ever built one of these is a GitHub page is basically read only display information. It has no back end. There’s no API inside the back end to submit a form or go get data or So, the other nice thing about this is Rayfin has a back end attached to it and there’s a handshake between semantic models and SQL databases and the front

9:31 models and SQL databases and the front end, which is the app that you build, it handles that security for you. So, you can fully build like So, to be very clear you don’t need rayfin to build apps on top of fabric. You can do that by yourself right now. I’ve been building apps on fabric since January of this year because that’s I started experimenting with things. It’s fully possible and I really like the sequel database in fabric because that gives me full sequel stuff and I think we’re getting post grass here pretty soon and once that’s there I’m going to try that one and I I built Yeah and also build. So that’s another

10:02 Yeah and also build. So that’s another one I want to get my hands on. So all these different things I’m looking at these going wow, this is really powerful. Wow, I can really build so much stuff. Tommy, you can build any experience you want. In power right back connecting other APIs. You want to bring your foundry chat agent into the item, you can do that. Like there’s going to be the capability for us to do a lot of extra things now directly through this system. So anyways, just wanted to point that out.

10:32 anyways, just wanted to point that out. This is really interesting. I really like what’s going on here. I’m very excited about this project. it’s already gaining lots of traction. People

10:40 already gaining lots of traction. People are generally excited about this. Yeah. And so that I want to continue to like teach and educate people on how to get going. It’s actually really easy to get started. There’s a lot of there’s like four templates that they give you out of the box. There’s another four templates on another website called awesome rayfins you can get started. yesterday I’ve been doing a video every day this week because I started a playlist that I’ll put it here in the chat window of all the rayfin projects I’m building right now and doing a public demo of those projects and getting templates going and showing people how they work. So that way people can understand

11:11 can understand how this rayfin system is coming together. together. Mike, I want want to touch on the latter part that you were mentioning about why rayfin is so much of what we’re going to call game changer. It’s going to go on the t-shirt of what’s a game changer. Mhm. Mhm. Because I think it even if you were just say just the fact that you can build any visual, you can use the nav, it can use D3, it can build that report for you with just a prompt, it’s not using Power

11:41 with just a prompt, it’s not using Power BI desktop, is incredible. However, the greatest challenge for most individuals and teams when they want to build something custom under their data, and we’re not talking embedded here, is all the back end authentication, how the data flows. Those are That’s a different language, and that’s usually the hardest part. And the one thing that Ray Fin really focuses on, or really accomplishes, is the ability to build the back end directly on the Microsoft One Lake platform.

12:11 One Lake platform. And it does the enterprise grade security, the governance, and any analytics that you need, and any application that you want. And this is the crazy thing here. So, if I’m a team and we need something custom, I don’t have to hire someone or learn all this language. That back end part is built into the skill that Ray Fin is, or has. Rather than, “Hey, you need to figure out authentication.” It understands, that’s its purpose. And this is really, I think, why people are seeing the buzz. This is also why

12:41 are seeing the buzz. This is also why we’re seeing so many examples of what you can do. Because, again, again, we saw fundamentally with Power BI desktop all these examples in the gallery because all you had to do is put something on a visual. You didn’t have to worry about authentication, you didn’t have to worry about your tenant. It was You had Power BI desktop, easy to do. do. Yes. Yes. We have not seen examples of Fabric as much because a big point of that is it’s got to run on some back end, it’s got to run on some authentication, or your own tenant.

13:12 some authentication, or your own tenant. And now we’ve eliminated that. That wall has gone completely down. Yep. Yep. And I think if you are listening, this is a fundamental shift that we have yet to discover how this is going to work in on team. team. I love all the I love what you’ve been doing already. But, Mike, let’s also be clear here. I don’t think we know yet how this is going to make an impact on teams and organizations. For example, I’m thinking about this in a few ways. One, how do I

13:43 about this in a few ways. One, how do I put this on a resume? Two, how do you deploy this on a team? How much do you actually deploy on a team using Raven? And then three, I think really just what the capabilities are. Those are the three things that I think we’re still very much undiscovered. Yeah, I think people are going to start trying to figure out how to use it, but I think the amount of community items that we’re going to build is going to be very unique. Let me give you another example here, Tommy. Tommy. We can now with Raven, we can now really stand back

14:13 with Raven, we can now really stand back and say, “What’s the problem we’re trying to solve? Mhm. Mhm. What’s the issue that my BI analysts have?” Right? I’m working on a number of clients right now, Tommy, that have pretty sizable semantic models. They’re pretty big. Tommy, there’s no capability for you to build anything in the reports by selecting a perspective. It just doesn’t exist. Like Like if you have So, let me just let me just paint a picture here for you. Let me just paint a problem here for you. And this is one thing I’d like to vibe code here or or agentically code with my agent here at some point.

14:43 with my agent here at some point. Imagine this, Tommy. I’ve got a model that has 30 tables in it. In one of those table In one of the So, we have a couple of facts, maybe two or three fact tables in there. And in one of those fact tables, I have a measure. How do I know when I select that measure which dimensions go with that measure? How do we know? Measure sets? We Yeah, would have been, right? We don’t know. But there’s no there’s no way to describe, unless you’re doing a lot of additional descriptions, me personally how I solve this problem is I

15:13 personally how I solve this problem is I go into the model view and look at the model view and say, “Okay, this measure is in this table. This table then relates these other tables.” I have a lineage lineage of what tables are available for me to be used with that measure, period. I be used with that measure, period., that’s something I’m missing. mean, that’s something I’m missing. That’s great, but it doesn’t help me when I give that model to a user to go build in the service or in Power BI Desktop. Desktop. They’re overwhelmed. They don’t They don’t have like if the the experience today doesn’t give you the lineage of

15:43 today doesn’t give you the lineage of where measures come from. So, So, when we roll this out to our BI analyst and people are just trying to drag and drop fields onto a a report, difficult. one other challenge I’ve been recent recently aware of, when you build a table or matrix, how you export the data from export to Excel, Excel, people don’t like the formatting of it. There’s no way to stylize the formatting. There’s no There’s no option to. to. Very difficult to work with. So, there’s

16:15 Very difficult to work with. So, there’s some there’s some really rich like data exportation experiences, there’s some really rich like user experiences that we’re missing, and right now there’s a lot of friction on we if we build big models, people get confused, we have a lot of education to get them up to speed. Why not build a Rayfin project that simplifies all that? Pick this measure, things filter down, you only see what you can use with that measure. Why not build things that, all that all that information already lives and exists inside the

16:45 already lives and exists inside the semantic model. It’s just not being highlighted in the UI. Why can’t I see the columns and tables and a lineage diagram all in the same spot and then build my table from there? Like that would be much more relevant to a to a business user. So, where I’m going after this, Tommy, is we can now really pick apart the problems, the adoption barriers that we have, and now Rayfin can be like, “Okay, great. I want to build Rayfin in a way a a Rayfin project that has models, dimensions, and measures. That’s it.” on

17:16 dimensions, and measures. That’s it.” on a page. a page. And And when you pick a certain measure, it hides or doesn’t show the measures or columns and the dimensions that don’t matter or that can’t be used with that. Right. Right. And then you can grab them and add [snorts] them to a table. Right? So, we can now use perspectives. Perspectives are something that has existed for years, but we’re never leveraging them in the front-end UI building experiences. Now we can use those. So, to me, Tommy, like like another one here, Tommy. Tabular Editor has this perspectives

17:46 Tabular Editor has this perspectives editor. editor. There’s no way There’s no place where you can edit perspectives other than Tim Doll. Doll. Right. Right. There’s no UI for this. Build a Rayfin perspectives editor. You can build lots of perspectives and use that with your There’s so many cool new things, Tommy, that we can go build that is we’re now we stop thinking about where we have friction in the system and we we step back and say, “What is the problem? problem? How can we solve it? How can we be creative to solve this one?” And then we just tell our agent and it builds out a

18:16 just tell our agent and it builds out a solution for us. Mike, I don’t know if we want to get to this yet because you’re already introducing things in my head that’s causing a slight concern. And I’m like Wait. What? I’m I’m thinking you’re going to be way concerned after all the things I say today. Well, well, and here’s the thing though, too. So, be before we get to that because that’s more of a confrontation type of thing. We’re only 20 minutes in and there’s still so much I think to talk about. But they still They I I do want to focus

18:47 But they still They I I do want to focus a little more on the value here of Rayfin, too, because it’s how to get started, it’s also a command-line tool in a terminal. It’s an NPM package. So, one thing I was testing out, Mike, that actually worked really well with Cloud Code, I can have my instructions and say, “Hey, build a Rayfin app.” And a prompt Rayfin with it. I don’t have to necessarily create the Rayfin myself, which is also really cool. Because it’s an MPM package, so I can say, “Hey Claude, take a look I literally

19:17 literally look at this Notion project. What’s a Raven idea?” There’s a Who created I don’t remember who created the skill, but it’s called brainstorming or it’s like superpowers for Claude. Okay. Okay. It’s a great set of skills. And it’s or it’s a plugin for Claude code. It’s like, “So let’s brainstorm.” It basically will think of different ideas. So it’s an okay, you can build an app with Raven. Told Claude, “Okay, let’s build a Raven app.” There’s an MPM package. It already knows what it is.

19:47 knows what it is. And it can start building with me and prompting it. And all I’m really doing is I’m still interacting with my normal agent. So if you’re like, “Well, maybe I’m not comfortable with MPM.” Okay, fine. Fine. You can use Claude code. You can use Claude the Claude Windows application to get this done or VS Code. And it can use Raven’s package. And that’s another thing I love about this, too. How they developed this, too. You can use coding agents to build Raven. You don’t have to build Raven yourself, either. So there are so many places here

20:20 either. So there are so many places here and and I think there is the development point of view that is so impactful. And then to your point, what you’re really highlighting here is the value, the output here, which the world is your oyster at this point. It truly like, it’s reporting, it’s applications. It can be to your point, whatever you can think of. Now, the first shift I’m seeing here and I want to hear your opinion. This really makes us developers. Even

20:50 This really makes us developers. Even though we’ve we’ve joked around about this, where you’re like, “Well, I’m a Power BI author. I’m a Power BI developer. Well, I’m a developer because I’m using DAX.” More and more Microsoft is realizing, I think, too. If you want to use agents and you want to use them effectively, you’re going to have to use a terminal. You’re going to have to use an IDE like VS Code or Cursor to be efficient. So, let me pose this for you. If this is so impactful for us and impactful for developers,

21:19 and impactful for developers, are we now true developers? I always like to go back to this. Well, what point does this go part of the resume resume? And at what point are you expecting to see this on the resume? No. No. This is This is so big. Tommy, if you’re not playing with Again, I’m going to This is such strong words. This is

21:49 This is Again, like I said earlier, this is the biggest feature since Power BI Desktop was released, period. End of story. story. I can do so many more things. This handles right back. I can talk to an agent. agent. At the Microsoft Build conference, Replit was announced as a partner with with Rayfin. You were talking earlier about like, I don’t want to use an NPM package. I don’t want to go use VS Code. I don’t want to write code stuff. Okay, that’s fine. But hey, agents,

22:19 hey, agents, Replit is this more business user-friendly experience. Talk to an agent. Get some experience. Get some UI stuff. They’ve done a lot of work on their end to make it easier for you to build out this business-facing front-end looking thing. Now, it’s just a limitation of what you can communicate to the agent. So, Tommy, I fully believe like we’re This is an a great evolution into what we want to be building, and before I was when we talked about Power BI, BI, we were essentially building data apps,

22:49 we were essentially building data apps, right? Things that are difficult to manage. It’s hard to get bookmarks to work. It’s hard to get visuals to show and not show. Now, it’s not. Now, I just tell the agent, “Hey, I want you to have a a chart on the table on the view here, and then I want a button in the upper right-hand corner that switches it to a table and then switches it back to the bar chart. Done. Complete like like I just describe what I want, and it happens. Tommy, I don’t want to build any more Power BI Desktop reports anymore. I Even like like I want to just talk to an agent and say,

23:19 I want to just talk to an agent and say, “Hey, agent, go do these things.” Right. so you’re expecting us then to see on resumes 5 years of experience in Raefen and Java applications? So 100% like right now? Like just like he said with Power BI? And people will ask you for like hey, I want to know who Raefen is If I’m a company and I’m going to hire new people, I’m not hiring anyone who doesn’t have some like familiarity with what Raefen is and how to build stuff like this. You missed the point. I don’t want you limited by what you only know in Power BI Desktop anymore. This This can do everything Power BI Desktop does and a whole lot more.

23:51 Desktop does and a whole lot more. Man, you are so excited. You missed my joke. joke. I asked you today, should you expect to see 5 years of experience in Raefen? No, I’m going to say yes. Today you don’t. Yeah, yeah, you just said you don’t. a real a real 20 years of experience in Power BI. That’s cool, you’re also lying. So yes, but like if you this is now opening up People that are experienced in this is now, right? You may not have experience with Power BI, but do I need that experience anymore? Do I Do I Woah, this is a strong point here.

24:22 Woah, this is a strong point here. Well, you could be an app developer, right? So app developers have been building static web apps for years now. But you’re not building Raefen specifically. And who cares if the NPM package switches slightly differently? Now you’re talking about app developers. So I think the challenge right now, Tommy, is this is this is going to be very detrimental to Power Apps because Power Apps is is more money, it’s a different licensing model. Like Dude, I’m never going to touch Power I wasn’t going to touch Power Apps before. I don’t want to touch it at all right now.

24:52 don’t want to touch it at all right now. Right? Right? honestly, listen, how much I I Power Apps, but you were saying this probably a year ago, and I’m like maybe we’re we’re not there yet. I have a project actually I’m probably going to I I’ve been thinking of contacting them and going to go, “Look, we can do this today. Like we can basically with what Ray Fin can do, I don’t have to go through entire PowerApp development.”, I think you’re right. There are Why would I build that? Why would I Why would I spend so much time like if I if I can go directly access or pull the

25:22 I can go directly access or pull the data down from Dataverse and put it into Power BI? Power BI? Okay. Okay. If I can bring Dataverse data into Fabric, right? Why do I want to go over to power Why do I want to pay more licensing for another thing when I could just build the same thing inside Fabric and just have it all work without an additional licensing fee or cost for that? that? Before we go to org apps, because I think that’s a great conversation about do teams start doing that today? Because again, again, take your hat take your self out of the hat of an individual contributor to do

25:52 hat of an individual contributor to do you actually doing this as a product for the company, but you mentioned something before we that I think is really permanent. You said, “Who says I need to have Power BI desktop experience at this point?” Mhm. Mhm. I need I need a quick pause here. A quick check here. Because here’s the thing. It still runs on Fabric data, semantic models, lake house, that infrastructure. Mhm. Mhm. You are forcing my hand here to ask the question that needs to be asked.

26:23 question that needs to be asked. Mhm. Mhm. Because I I wanted to wait, but you’re saying things and it’s bugging the heck out of me. I can feel like a jab. You’re jabbing me. I got to punch back, okay? So, at one point, how much of my development should be Ray Fin and how much should be Power BI desktop? Should Now I switch everything that I’m doing from Power BI development and I need Truly here, think about this because there are pros and cons to Ray Fin. Because guess what, Mike? There are pros and cons to any technology, to any product. If I do Rayson, I do not know

26:54 product. If I do Rayson, I do not know the language. I am at the whim of my prompt, right? So, we know there are some cautionary things. things. So, So, you can’t just say, all right, I’m not saying you’re saying this, but one person cannot say, “Forget all Power BI report development. Everything will be done in Rayson.” I don’t think you’re saying that. However, However, there is going to have to be some decision tree on when do I do Power BI desktop and when am I developing in Rayson? Rayson? Mhm. Mhm. So, there’s going to be there’s going to

27:24 So, there’s going to be there’s going to be decision tree. There’s already I be decision tree. There’s already, Tommy, to be clear, there’s mean, Tommy, to be clear, there’s already millions of reports that are already existing that are Power BI reports already. People have built them. They’ve been maintained. Get it. those items that are already out there, you can still modify them. You can still use them with these now fabric skills, skills on fabric, right? So, skills on fabric will let you modify your desktop with AI. Yeah, you can So, you can keep those things around. You can still modify them. But again, if I sense any bit of resistance

27:56 if I sense any bit of resistance to what the business requirements are for something report, whatever that is, if I sense any resistance to that, I’m going to move away from the Power BI like if there’s a feature that I can’t do easily in Power BI desktop, that to me right away right away, that’s the the first moment of moving over to Rayson. Now, let me give you some other context here as well. Rayson doesn’t have everything that you need. Like some projects have a lot more, some projects have a lot less. So, what I was experimenting with templates this week, I’m I’m running the Rayson

28:27 this week, I’m I’m running the Rayson project project and there is four separate templates that you get with Rayson. The first one has only one skill attached to that project. The second one they have has a ton of skills attached to it. So, as you think about Rayson and like a standard language, a component library, what does the visuals look like? So, there’s actually a good amount of work to get a good-looking report out of something. If exactly what you’re building, you can you can get to a really good result. However,

28:57 result. However, if you don’t know what you’re doing and you don’t have a framework of visuals and formatting and,, how everything should look on the page, you have to design a style guide to to help the agent get a better output for what you’re looking to build. So, that’s information that you have to go figure out how to go do and build some of that framework around the style guide. Once you have that, I think you’re then off to the races. You can build anything you want, right? the I one of my earlier projects was doing I I had some mock data I was going

29:27 doing I I had some mock data I was going to bring in the report. I didn’t really want to have Raefen I wasn’t testing like the data and semantic model aspect. I figured we’ll wire that up. It should be no problem. So, I said make a bunch of Vega and Vega light visuals, put them on a page, and then for each visual I was working on stylizing and fixing and creating and manipulating the one visual. And tell me, VS Code has this new feature now, which is amazing. So, in VS Code now, when you run a Raefen project, it opens the local host

29:58 Raefen project, it opens the local host and it puts it in the browser window of of the of the They took that feature, by the way. So, yeah. yeah. Yeah. Yeah. And then now there’s a button that you click and you can select the items on the page, which gives context back to the prompt. I’m like, “Dude, this is the way to edit web pages and designs.” Cuz the agent doesn’t have any stinking context about like what you selected. So, that’s a good option. The other one I really like to use is a skill called agentation. Agentation’s like agent and annotation as one word,

30:30 like agent and annotation as one word, agentation. com. say cloud design, but yeah, continue. So, So, Agentation does this exact same thing. It has a little icon on the bottom right-hand corner. You can pause animations. You can select items on the page. And then there’s actually a skill that goes with it that your agent has. So, you can actually use the skill and it’ll find the details. It’ll cache the results for you on the website and the agent will be able to go get the details, find the information, and add it stuff for you automatically. So, these these it to be honest, Tommy, I I have fun building Power BI reports

31:01 I I have fun building Power BI reports and it’s nice to build good-looking things. things. The speed for me to get from idea to report or design or problem-solving for me, for me, Rayfin is now even further shortening the ability for me to make good stuff now because my I can go from idea to tangible prototype or MVP or product really fast and to be honest, it’s extremely fun. I’m having fun building

31:31 extremely fun. I’m having fun building stuff like this. It’s incredible. Now, for you for you it really is it’s been mind-boggling. It’s been mind-breaking. But, But, I know it’s not pushback’s the wrong word because everything you said I agree with, I support, I completely celebrate. There’s no other way to say it. However, all the things you’re mentioning here, my mind is going to handoff. My mind is going to the scaling or the deployment

32:00 going to the scaling or the deployment in an organization because you mentioned the agentation skill or using VS code or maybe cloud design. I also have that NPM, NPM,, the package. I also have to know how Rayfin works, right? Mhm. Mhm. This is all great and again, that’s what we do. This is what we like. Yep. Yep. But, whether I’m internal or as me as a consultant, I have built this app through Rayfin and I have used all these skills. skills. Yep. Yep. I used the question is not why is that wrong because that’s not the case, but what’s the handoff look like at some

32:30 what’s the handoff look like at some point because this is I think where we need to find I think what the product itself is and the knowledge transfer is because because to your point you can go there’s none of it there’s none of it there’s none of it so so what are you talking about Tommy we’re building an app here your documentation should be part of the app you just you tell the agent add information bubbles add an explanation page add a how to use this page like the

33:00 page add a how to use this page like the application itself that’s not what I’m talking about I’m talking about the ownership of the application the ownership of Raven give it to the company go go say hey here’s the here’s the app it’s a GitHub repo I’m going to check it in for you boom ready to go everything is self-contained and you tell the agent to document itself write documentation write the read me put all the information together that we need update the read me I don’t write documentation for any of my projects anymore agents are are yes yes agents are extremely good at that handoff and I ask the agent write the

33:31 handoff and I ask the agent write the documentation so another agent can go pick it up now will I give this project to a company that’s not using GitHub that’s going to be very hard for them if they’re not using sorry if they’re not using a GitHub if they’re not using agents to build things right right they’re going to be stuck because there’s a lot to go through there and I I am not going to think that anyone’s going to be able to pick up that project and code it’s going to take them a long time to understand the structure and all the pieces this is what I’m talking about that to me that’s a barrier that is

34:01 that to me that’s a barrier that is again the whole point of Raven is develop with an agent right right it’s not write code manually one line at a time so if the whole point is to give agentic experiences to building applications any company that’s going to hand this off I’m consulting from my first question to you is do you use agents do you have access to GitHub co-pilot or some cloud code co-pilot some something’s there that you can put at repos and work with code if the answer is no then I’m going to say this is not a solution for you

34:31 to say this is not a solution for you you need to first go work with your security group, unblock yourselves, get the IT to accept some form, whatever that form is, of agent building experience. Only after that is done, then you can use this. Because I don’t think you’re going to want to maintain this one without having that in place first. I I’m very happy with No, it is what you’re going after. It is It is It is. I just want to make sure I’m like I Listen, I love you, but I’ve heard things on this podcast that has shocked me. But no, we’re on the same page here

35:01 me. But no, we’re on the same page here because I think that unfortunately right now is a barrier. And I’m thinking too, if I’m like let’s say I go back to some of my previous roles and I remember building some things that was something I learned myself. It was more or less when Power Apps came out. And it was like, this is great, but you have to train the team. Like we have to get everyone on board with this before we’re going to use this in the organization. Mhm. Mhm. Because otherwise, again, if I had a heart attack or I got hit by a a car when I was riding my bike, which just happened, right? Who’s going to own the

35:32 happened, right? Who’s going to own the app? And to your point, if I built something in Raythen for my organization that was solid and I we’re going to see so many things that are going to help businesses and help people’s process and they’re going to be dynamic, but if I build that and the company goes, “Well, Tommy, we don’t use AI. Like we haven’t blocked these things.” Then that app becomes a non-starter, right? Then Raythen becomes a non-starter and I can’t now deploy this, right? Or I’m not going to start getting people to use it because

36:02 getting people to use it because again, it’s using something the organization doesn’t like. And listen, I’m with you on this where it comes to if your organization is not realizing like everyone says AI AI AI AI and they’re thinking about this all from the co-pilot and the chatbot point of view, but I think the first step is enabling these types of tools, too. Tell This is a little little random, but I was talking to someone the other day and they posed a question to me. They said, “A lot of consultants are saying

36:32 said, “A lot of consultants are saying you need a year to shape your data before you can use AI. What’s your take?” I said, “That’s completely off base because you’re trying to pull Who’s making that junk up? That’s ridiculous. giant consultant company., the people Yeah, like so right, but Exactly. That’s what the heck is that coming? There’s literally a prep your data for AI button inside semantic models that are like are like Right. Right. do like do like 50, 60, 70% of what you need to get done. done. who don’t understand what AI is really. I think they just This is the buzzword

37:02 I think they just This is the buzzword because there’s AI for the consumer, there’s AI for the developer. Obviously, you want to shape your data. You need to clean your data a little, but slime balls. [sighs] [sighs] The It goes back to that solution looking for a problem. If we just say we want AI, you’re never going to get there because you’re going to be trying to fix things because you haven’t identified the issue. Anyways, that’s a little off a tangent. So, there is this really critical point here with Rayfin where you and I can use this all the day long.

37:32 you and I can use this all the day long. I can stay up late,, and build things that are life-changing in terms of my own workflow or what I can provide value to a client. Sure. Sure. But, that only gets me so far. That only gets me so far because if I go to my client go, “Yeah, you everything you asked for, it was going to take 3 months. I did it in 2 days. How about that?” And it does everything you asked to do it no problems. They go, “Well, that’s great. Okay, well, how do I manage it?” “Well, do you have Opus 4. 7 and do you have Copilot?” They go, “No,, that’s we don’t want to spend

38:02 , that’s we don’t want to spend that money.” All of a sudden that app becomes no value for them. So, Yep. Yep. Let me ask you there. Let me Let me pause. Okay, yeah. No. No. That just means you just secured your next 6 months of work with them because they’re going to ask you for changes for everything. So, so it’s a really good, Tommy, actually. So, how do you all to this? Yes, you keep hiring me. Look, I built this thing super fast. It took me a week where it would have taken us months to figure this thing out. So, we’ve we’ve compressed the time sign to build something. We’ve got this amazing experience. Awesome. Love it.

38:35 amazing experience. Awesome. Love it. Right? Okay, by the way, Mr. Customer, Right. Right. do you have agents stuff to build this stuff? No? Okay, well, I’ll maintain it for you. Just pay me a retainer. I’ll maintain the app. You tell me what you want and we’ll build the features for it. So, honestly, it I’m going to use that as an opportunity to sell into that customer even more and say, “I have the tooling. I’m using agentic development experiences to build these things for you. You don’t need to go hire someone else or go open up to yourself. We will build what you need for your for your use cases.” I am so happy we’re on the same page

39:06 I am so happy we’re on the same page here with this because I was this was the word part of like, we are going to fight about this and we may not talk for a few days, but we’re on the same page. I think everything we’re talking about on handing to our organization. I have another question for you and you mentioned this earlier on about application. You want application developers? Who is best served building Rayfen, building applications in Rayfen? Well, So, unless you have it somewhere else, yeah. yeah. let’s talk Well, a couple of things I want to note here, right?

39:41 Replit is the business-facing version of building apps. Replit is the power apps of app building with agents. That’s how I see this. Yeah. Yeah. Lovable, Replit, other other tools out there. there. Tommy, on the podcast, we have heavily ragged against Microsoft about many times we’re talking about Lovable or Replit. Right. Right. Why isn’t this here? Why can’t we use these experiences? Well, it’s This is the answer to that problem. So, the griping that we’ve been doing, this is the solution. So, Microsoft, I know you’re listening someone over there.

40:11 you’re listening someone over there. Thank you for for listening to part of the podcast or allowing us to gripe for a bit while you built out the solution. I wish I could have known earlier this was coming cuz this is really really impactful. This is what we want. So, when I look at the spectrum of what users need to build things, I’m not going to assume a the Replit developer needs to be, you’re technical, but you’re not like writing every line of code. You can see the code, the files are there, but you’re doing more guiding and orchestrating of like how you want the the shape of the product or application

40:42 the shape of the product or application to run. to run. So, that one on one hand, that’s really good. good. And that’s getting to be very well known. if we’re going right now, when you look at purely only using Fabric alone, it’s it’s a more technical right now. Now, Tommy, I’m going to I’m going to go out on a limb here. Oh, boy. Oh, boy. Here we go. I I don’t know if this is true or not, [snorts] [snorts] but this could get very meta very quick.

41:12 but this could get very meta very quick. Meaning, Meaning, why not build a Raven app that builds Raven apps just like Replit inside the application? Right? So, I’m seeing people So, the new model the new model from Claude called Fable is now out. now out. Fable’s now out. I’m seeing people with minimal amounts of prompting getting like someone just said this could be like made up stuff, but someone said they prompted Fable and said build me a

41:43 Fable and said build me a Minecraft-like copy application. And it just straight-up built the whole thing. It almost looks identical to Minecraft. Mhm. Mhm. So, either it cheated and copied the code and just ran the existing code or it just like also impressive by itself. Which could be what it was doing, right? I don’t know. But it’s also still impressive regardless. But the But the we said last month or months ago, not even There was like last month. No, probably two months ago. Two months ago, we were saying software is going to zero.

42:13 software is going to zero. Software price is going to zero. Now, I’m going to say no, that’s not exactly true true because you still need to pay for the tokens, you still need to give it some information. So, and and in in lieu of the the June 1st price increases on cos all the agents six spaces, right? There’s going to be some level of cost to using these tokens. tokens. But, I’m going to argue the costs have

42:38 But, I’m going to argue the costs have just greatly dropped from what it was to build software to what it is now. You’re not paying a developer for months on end to build you software. You’re paying an agent to work with a developer to orchestrate out what you’re building. So, the costs are now weeks long and hundreds of dollars of tokens as opposed to thousands of dollars for people and developers. So, I think that’s the shift that’s happening, right? That’s That’s what we’re seeing here. There’s no reason why we can’t throw Fable at this and say, “Hey Fable, go build this really complex thing and copy

43:10 build this really complex thing and copy this really interesting experience.” And boom, it now gives you this really rich awesome experience for you. So, I might have mentioned this to you before. before. I think it was actually on a Teams message offline. But, with the introduction of Fable, with the introduction of Raven, the fact that it’s going to consume a lot of tokens to build things that previously either would take months and months to do or were almost impossible under the fabric environment. Mike, who carries the cost then?

43:41 Mike, who carries the cost then? Right? And I think this is Let’s talk from the consultant to client view. If I am If I can tell a client that I can build can build, your dream application that we, you tried searching heaven and earth for by Friday, well, who’s going to carry that token cost? Who’s going to carry the agentic cost there? Is it the client? Am I writing that in the statement of work? Like any token,, usage is on the client or is that on me? Because it’s different

44:11 or is that on me? Because it’s different obviously if we’re if you’re FTE because you’re hoping that you do have a pretty wide token amount, but is this even a conversation yet? Because I feel like to me this is a one of the costs that a client would pick up. The the agent? Yeah, if I’m using Raytheon, which we know is going to carry a lot of tokens, and I’m using Raytheon Fable and Raytheon, excuse me. Like I can build their dream application in 3 days. That would take yeah. Well, you’re you’re going to price accordingly, right? You’re going

44:41 price accordingly, right? You’re going to price your rate as Tommy appropriately [snorts] with getting I mean I’m going to go back to the I’m going to go back to a a summary here. So, Tommy, time is money. Mhm. Mhm. Time is money, period. If you can get things done faster, you spend more on it. Right. Right. The ultra-wealthy fly around in jets everywhere because their time is so valuable. It’s it’s faster for them to get to the place to talk to the person to make the sale, the deal, the whatever. Their time is literally physically money being made by if they

45:11 physically money being made by if they don’t get there where they need to be, right? That’s that’s their deal. I guarantee you that with the client, by the way. how Jeff Bezos flies on a [snorts] jet? That’s me. I’m flying on a jet. I’m flying in Raytheon jets. That’s what I’m doing. doing. I’m flying a jet with Raytheon. Raytheon. So, but I’m but I’m I’m I’m looking at this going,, time is money to these clients, right? You can either have it fast and now, but that’s going to cost you more money than slow and longer-term like slower and over more more time period. So, how

45:42 over more more time period. So, how quickly do you want to move the business? Do you have urgency around this problem? Is this really important to you, right? If it is, okay, Tommy’s rate is going to go up by $20 an hour, $30 an hour. You’re Tommy, your rate should reflect the ability for you to go faster than the next person. So,, the the message here is yeah, we’re not we’re not the cheapest consulting group in the box, but we’re going to be the fastest and the most thorough. thorough. And here’s here’s here’s how we’re incorporating AI into what we do to give

46:12 incorporating AI into what we do to give you deliver results faster. We We need to stop focusing on time and materials. That that conversation agents have really adjusted how that works. We need to focus more on features, software, deliverables. What what is the enabler that you need because like nothing before we’re able to get through and enable companies to do things at at a clip that we’ve never done before. So, moving away from time and materials is not a thing. It’s going to be more about retainers and deliveries and outputs and

46:42 retainers and deliveries and outputs and outcomes. That’s what we should be selling. selling. Cuz now that cuz what that does, that wraps up all your knowledge, Tommy, as you have Tommy, you’ve been working on agents since they came out. That’s months and years of work that you’ve incorporated in your mind. You’ve built tools. how to use different tools with the agents. You different tools with the agents. which tool to pick. You’ve become know which tool to pick. You’ve become an agent architect. It’s to some degree. They’re paying for an agent architect to show up and do work with them.

47:13 work with them. You’re flattering me, but honestly though, too, I this is wrong. wrong. No, I I I I I’m You can send you can send me the check in the mail later. Yeah, it’ll be a Raven app, so it’ll Raven will send it to you. but honestly though, too, this is the just a little change This is the fun part of what we get to do. This is the best part of being I feel like in our space. space. Mike, when I remember we first saw ChatGPT when it first came out, it was a moment just like when I saw

47:44 it was a moment just like when I saw Power BI Desktop where hey, this is going to be big and I remember like I’m going to invest the time. time. I’m going to put my eggs in this basket, my proverbial agent agentic eggs or at least time into it. And it’s very nice to see that, that’s paying off and I think for both of us too. You were the same like, hey, this is more than just kind like, hey, this is more than just cool. Like there was a lot of of cool. Like there was a lot of research. The terabyte and a half of local models that I have on my machine testing things out has finally paid off

48:15 testing things out has finally paid off and I’m doing command line tools and all the things because really Yes, can you simply prompt Raven to get a pretty cool cool tool? Absolutely, but, there are so many other skills here here and experiences that you need to know because all things we’re talking about today that I love that I don’t think other people are talking about. What if you come What if the company doesn’t use AI? What’s the deployment look like? How does it actually work from a cost into a value point of view? Yes,, I’m going to say yes and no, Tommy. There’s There’s not You’re

48:45 no, Tommy. There’s There’s not You’re making it sound I think in my mind the way you’re articulating your words. I know you’re not meaning this way, but I think that’s the way it’s coming off. The way it’s coming off to me is you’re saying there’s a lot of net new skills and things we got to figure. No, I don’t think so. I think all we’re doing is we’re moving away from hands-on keyboard and writing DAX and clicking buttons in a desktop application. We’re moving away from that and we’re moving more towards how well can you articulate the problem? How well can you communicate to the business about what their needs are? How well can you take their needs and materialize those needs into

49:16 materialize those needs into an agent command script thing that you can go use. Not I’m not disagreeing with you. There is some barriers for organizations to not use the tooling. Does this Raven thing another another Right. Right. side note to this, Tommy, is who do we give Raven to? Is it everyone in the entire organization that we give out the Raven items to? items to? Absolutely. right now, right now the answer is no. Right. Right. Do not, right? So, right now it is find your best Power BI person. The one who

49:48 your best Power BI person. The one who knows the most about building DAX and models and reports. That’s the individual that you identify. If that’s the the developer that you’re looking at, go after them. That developer understands how to handle semantic models. They already understand what visuals are there. They already know the limitations of what is going in and out of the reports. So,, yeah, William on the chat here is saying critical thinking is coming back to the center of of focus here, right? Finally. Finally. It but this is what we need. It’s it’s

50:18 It but this is what we need. It’s it’s about about emotional intelligence. It’s about critical thinking. It’s about identifying what is the problem and then you have to be creative enough to be able to to ideate what solves that problem. And what Raven Raven is just a tool, just like everything else. So, it’s So, I’m looking at this going find the best people that you have developers, give them some Raven, let them get really excited. Another gentleman here or person, I’m not sure if it’s I think it’s a guy, but I can’t tell from the name on the thing. But, someone else said

50:48 said my wife is now just found Raven. And she’s all fired up. She’s really excited about agentic AI and on unlocking a new level of excitement around Raven and building reports. I 100% agree. This is exactly how excited I am. And I said,, I I responded in the comments like, be prepared. Your wife is going to disappear for hours upon hours in the evenings and at night working on trying to figure out how to use Raven to get solutions done. It’s fun and it’s

51:18 solutions done. It’s fun and it’s exciting. exciting. I’m going to pause there, Tommy. What do you think? Well, first off, I just got to give a quick shout out to my grandmother because she always said that critical thinking and emotional intelligence were some of my best things. Like, all right. And she also said that she could She also said you were handsome, Tommy. She she did. And that obviously worked that did not work out nearly as I thought it would. So, what, Tommy? I’m going to I’m going to let you in on a little secret. Grandmas lie. I know. Especially Italian ones. I got to think so. She’s right, Tommy. She’s right. She’s

51:48 She’s right, Tommy. She’s right. She’s absolutely right. what my grandmother’s favorite game was? Guess who died. Guess who died? Oh, no. Did you hear about Aunt Cathy? What happened? happened? You’re like, oh, okay. So, yeah. So, yeah. No, but no, but I think really too, you’re you’re two huge things that you’re mentioning here. There are some very soft skills that make the hard skills work here. Unlike I need to know DAX and I can do DAX well. That’s not the case here where I can do prompting, I can do agents well. There is a way of framework. Mike, you don’t go into

52:18 framework. Mike, you don’t go into projects thinking in the same way you used to when it was pure development. There is a different way of thinking and it’s hard to articulate in terms of that critical thinking idea, the framework that you put in. I want to touch on another thing you said because, dumb minds think alike and you and I are clearly on the same page here. There is a concern if you were to roll this out to the organization or just start creating about that whole idea of AI slop. AI slop. Because we have to be careful just

52:49 Because we have to be careful just because we can doesn’t mean we should, right? Just because you can build every application doesn’t mean that you should start building everything and testing everything that way. You’re going to waste a lot of tokens. You’re going to build a lot of things that are probably not great,, that That’s going to happen beginning. It’s part of the learning curve. It’s going to happen. But if you but to your point, if I just say, “Hey everyone at the company, start building on everything. Here’s your fabric thing.” and everyone go crazy. Would that be successful? No. It’s not

53:19 Would that be successful? No. It’s not going to be the best whatever is best floats to the top. What’s going to happen is there’s going to be a lot of ambiguity. There’s going to be a lot of not sure what should we use. There’s going to be a lot of slop and at this point anytime even when it’s something so cool like this comes to the surface and comes available, there has to be a process of to your point, who do you give this to? My best Power BI developer who understands the data and the people, Right. Right. and I would ideally like to give them a little AI training if they’ve never done

53:49 little AI training if they’ve never done it., if they’ve never if that’s not part of their,, tool belt. And you also want to give them you want to give them set problems to do. It’s different for you and I because you and I can run wild on my own computer and with my own machine, my own tokens to test things out. That’s part of our job. Like part of our job is experiment. Yep. Yep. That’s not one’s,, everyone doesn’t have that luxury. So you So you Us as consultants, like as a consultant we’re expected to come in with like the

54:19 we’re expected to come in with like the newest, the greatest, the patterns, the things. So like experiment testing yeah, testing every angle, what works. So our job as company owners is to come in with a little bit of mix of like we have to do work to pay the bills, but we also have to do experimentation and development and like net new things to figure out what are these new patterns,, how does this new technology fit into our existing governance and knowledge around the fabric ecosystem. So 100% agree with this one. and so I think this is where we we we as a consultants need to do a bit

54:49 we as a consultants need to do a bit more of this one. I think organizations should run in a similar way, Mhm. Mhm. but often the allocation of resources, it’s it’s difficult for an organization to to tell one employee like I don’t want you to work on things half the week. week. I want you to research this new Raven thing or build these new prototypes or experiment. That’s a large cost to the business business 100%. 100%. in hopes that a future something will pop out of it that’s actually really useful. Right? So I think a lot of

55:19 useful. Right? So I think a lot of organizations are unwilling and unable to unburden employees with the daily tasks of running the business enough that you can actually get value out of it. And this I think one of the reasons why you hire consultants. As consultants come in, they’ve had more business experience, they’re doing different things, they’re sharing their knowledge down to that business level. and I think that’s that’s an a very interesting way there as well. as well. All right, Mike. I love the angle we took today because I love the how much we talked about the capability of R and and

55:49 capability of R and and, let’s be honest, this is not going to be the last time we talk about this. This is I I have a feeling it’s going to be part of our Beat from the Streets for a while. Probably more episodes and I I really want to hear from users if they’re even allowed to do it right now. if we if I were to pull 100 of our listeners and how many of them would even have the capability to download and install the NPM package and to get started? So, maybe that’s another poll, too, because

56:19 maybe that’s another poll, too, because this is so if you speak to me here, finish my thought. thought. Yeah, get some water, Tommy. Mute yourself real quick if you don’t mind. so, I’m going to actually pick up a a question here in the comments that I think is actually extremely relevant when we’re talking about R. We have a comment here that says, we have complex DAX measures that are that to to determine what is an expense and and when the expense is occurring, right? How does this translate directly into R? Well, R is just an application that runs on top of Fabric. So,

56:49 that runs on top of Fabric. So, if that DAX measure lives inside a semantic model, right? My word of advice would be is first and foremost, start making sure your models are AI ready. Make sure that all the columns have definitions. Make sure all the measures have descriptions in them. Make sure that you have good relationships. Make sure it’s a star model. Like there’s a start following all the best practices in your semantic model because then you can then bring in those DAX measures directly into the semantic model and then use that right inside

57:20 model and then use that right inside inside the experience of R. And so, now R then has all the business context, the user availability. R can go write its own DAX on top of the existing DAX you have. It can build tables. It’s going to write build tables of data using the existing measures that you already have. So this just emphasizes a whole lot more of how important the semantic model becomes because now again Tommy we’ve said again on this on this podcast before there’s a whole bunch of data engineering items we have on one one one

57:50 engineering items we have on one one one side of things. Right? The the pipelines, the notebooks, the storage accounts, the lake houses, all all of that lives on the front side. We narrow all that information down to the semantic model. And and many and not even just us semantic model, there’s multiple semantic models. Semantic models carry all the business logic and definition. You can give someone a bunch of lake house tables, that has no bearing on how they’re going to how those tables relate. You’re not giving users end users, you don’t we

58:20 giving users end users, you don’t we don’t give them lake houses without some definition or metrics on top of it to help them build their their data. So first and foremost that’s missing, right? What you supply that with is you go put the semantic model in the middle here and the semantic model is how do those tables connect? What is the definition of those tables? Where did the data come from? What measures or calculations do I want to reuse? The output of that is the semantic model. And so I see on the everything narrows down to this really rich thing called

58:50 down to this really rich thing called semantics layer and this is important important because every single other big data tool is now building a semantics layer. Data bricks built metric sets way late way later than Power BI did. Power BI has had semantic models out for years, 15 plus years. Micro Data bricks is now just catching on and say hey we need metrics views. The exact same thing as metric sets. All right sorry the metric set that they’re building is the same thing as a semantic model. Snowflake is building their semantics layer, same thing. They’ve got

59:20 semantics layer, same thing. They’ve got a bunch of tables but nobody knows how they all relate together. And the reason these two companies are adding these capabilities to their tooling is because they need agents to talk to them. There’s no business context in these big freaking lake houses that no one can look at. So, you’re you’re killing me, smalls. Like, just do the semantic model thing. So, everything, no matter what platform you’re using, needs some semantics layer. So, the semantics now expands the other side of the the perspective. Now, I can build reports,

59:51 perspective. Now, I can build reports, paginated applications, Raviton projects. Like, the other side of things is now a rich layer of outputs. Again, they all hinge on a really good semantic model. model. This is fundamental. And it’s becoming even more clear to me as we build more rich experiences on the user side, the BI business user side, those experiences are are changing. Raviton’s just one more of another great library of tools that you need to have

60:22 library of tools that you need to have on top of all of those semantic layers. And you answered my question about who do we give this to. This is why it’s so fundamental, even with the introduction of Raviton, that you have the context in the experience of what how data flows in the data Microsoft data platform. Is so essential, it’s going to be more important. So, like, my closing thought here here is we’re at the another like, hill of what’s unseen. There is something that we know now exists that

60:53 something that we know now exists that we have yet to really discover. I cannot wait to see where this goes, but thank you for having this conversation with me. I cannot wait to see really, I think, how this really gets applied at scale. scale. I think it’s going to be very exciting. Stay tuned. You want to definitely follow the Power BI Tips channel. So, I will throw down a another playlist here for everyone who’s watching this playlist. We have started another playlist called learn Fabric apps. And it’s all Raviton. So, well, I

61:24 And it’s all Raviton. So, well, I want to be very clear. Like, I I I think this is really important for us to know because because the Rayfin side of things is going to be,, very impactful for your business. It’s going to greatly change what you work on, how you work on things. And so,, I just want to be very clear. Our playlist here about Rayfin stuff is going to be incredibly important for you to learn how to use and and gravitate towards,, leveraging these things. So,, I think that’s going to be incredibly important., here’s the link. I’m stalling a

61:54 , here’s the link. I’m stalling a little bit cuz I want to put the link in the chat here. So, here’s the link to the playlist on YouTube. Make sure you go check that out. We already have four videos deep., we’ve done a recap on episode 17, just a build recap and a very slight demo getting started with your first Rayfin project. Episode 19, 19, we start building a Rayfin app and we we talk more about just in general what a Rayfin app is, just more communication. Episode 20, we start building templates. We have two templates that we use in episode 20. In episode 21, we build from a template that’s actually a game, a

62:25 a template that’s actually a game, a video game. So, you can play a full game in inside Fabric now. So, now that exists, we’re going to continue to rip out more videos. We’re going to hopefully have one video,, for the rest of the week this week and maybe,, some next week as well. We’ll see what happens., I’m doing a little bit of traveling, so that may or may not happen., but stay tuned. There’s going to be a lot more Rayfin things coming. We’re going to continue to involve,, involve,, our skills and using that to help teach and educate the community about what Rayfin is. Anyways, thank you all so much. Tommy, where else can you find the podcast?

62:56 can you find the podcast? You’re on mute. You’re still on mute. Can’t hear you. Tommy, learn how to use your computer. I thought the keyboard shortcut worked. Anyways, that’s because Tommy’s not feeling great, so we’re muting ourselves a lot. You can find us on Apple, Spotify, wherever you get your podcast. Make sure you 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 bi dot tips slash podcast, leave your name and a great question. And finally, join us live every Tuesday and Thursday a. m. Central on all Power BI Tips

63:27 a. m. Central on all Power BI Tips social media channels. Thank you all so much, and we’ll see you next time. Explicit Measures pump it up high. Tommy and Mike lighting up the sky. Dance to the data labs in the mix. Fabric and AI get your fix. Explicit Measures, drop the beat now. Hop and skip, feel the crowd. Explicit Measures

63:57 Explicit Measures Drop it loud.

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.