Where Writeback Belongs – Ep.569

Writeback has been near the top of the Explicit Measures mailbag for years, and an anonymous note asks the question again now that Fabric apps are generally available. Tommy Puglia and Mike Carlo argue the report should stay a report, and that the action belongs in an application that can write forward.

News & Announcements

• All About the Power BI / Fabric MCP — The next Chicago Fabric and Power BI user group is October 22 at the Microsoft office downtown. Tommy’s plan for the night is installing the local Power BI semantic-modeling MCP and the Fabric MCP, then showing how the Fabric MCP reads across OneLake and runs on your behalf. The last meeting had Gil Raviv in to show a win, and they want the same from anyone who has something in progress. RSVP with a first and last name. A Meetup handle is not enough for the ID check at the door.

• What’s new in Microsoft OneLake — Tommy’s five highlights from the Barcelona OneLake post are IQ sharing of governed business context, connectors for Salesforce, Dynamics 365, and Business Central, consumption-based OneLake compute that does not require a dedicated capacity, stronger instructions for AI and agents, and Fabric apps extended to warehouses, lakehouses, and semantic models. IQ sharing, as he reads it, lets an organization reuse operational knowledge, including markdown agent instructions and RDF ontologies, across Copilot, agents, and applications. Mike compares that to a Notion-style system whose files land in a lakehouse next to the SQL. Tommy has already added a Notion sync to his skill-vault app, and he notes the Fabric feature does not yet understand a skill folder with scripts. The other half is Microsoft Foundry discovering OneLake metadata, plus the OneLake table-read API in preview: serverless access to Delta and Iceberg with row and column security, without starting Spark. Mike calls that the API a backend needs. Tommy’s name for the pattern, which he credits Mike with pushing first, is Fabric as a backend.

• Microsoft and Snowflake’s commitment to Apache OSI — Christian Wade’s post is Microsoft stepping into the Open Semantic Interchange work that Snowflake and Databricks were already in. The near-term scenarios Tommy reads are converting between a Snowflake semantic view and a Power BI semantic model, making DAX a recognized query language, and extending the spec for ontologies. The post cites the scale of semantic models, including use at 94% of the Fortune 500. Mike has been in the public meetings and had already opened a pull request asking for DAX alongside MDX. He says Microsoft’s converter landed and folded that change in, so DAX is already in the spec even though the post still talks about it as upcoming. His test for the spec is whether a measure and its filter context can be expressed so SQL and DAX return the same result. The project is still incubating and still accepts breaking changes. Mike also flags a related OneLake note: read and write from Azure Databricks, and a preview item for native OneLake storage aimed at the friction between Databricks and Power BI.

Main Discussion

Topic: Where writeback belongs once Fabric apps exist

The mailbag has no name on it. The writer points at Mike Zach, former CEO of ATRX, a Power BI-native planning and writeback platform now part of Vena, and asks whether the next phase of BI is acting on a signal without leaving the report: Power BI as an analytical layer, or Power BI as an operational application, once AI makes the insight faster than the workflow. Tommy puts writeback in the top ten mailbag themes of the last five years, close to the podium.

• Writeback is a narrow name for a wider job. Mike’s first cut is that planning and budgeting are a real case, and that they are too small. The useful question is which action follows the number: a comment, a captured reason, or a call to an API that is not the report. Writing into a semantic model is the hard version. Columnar storage and row storage are built for different work, and a dimension update is not an operational-database update. Tommy’s history of the pattern is a semantic model that sat on SQL, so “writeback” meant updating that database. Power Apps, which Chuck Sterling demoed for years, often took values from a model and acted on them. They were not editing the model.

• Tommy wants the word to change from writeback to write forward. Most of the requests he sees are not a correction to the gold table. They are an alert, a notice, or an update in a different system. Power BI, in his line, is for consumption. The operational change belongs in a Fabric app or, less often, a translytical task flow. He would not have someone change widgets inside a report now that those other layers exist. Direct Lake makes a custom visual that edits gold a non-starter for him.

• The price of a third-party writeback seat is the decision. Translytical task flows are in the product: a button submits to a function, and you pay for the items you create. Planner, which Mike traces to Lumel and a Microsoft partnership, is a separate item with its own capacity, and finance teams react to it the way people reacted to a first smartphone. Fabric apps are included, and an agent can shape the same planning experience. Mike is not willing to pay a third party for writeback anymore. Tommy’s challenge to anyone who has Fabric in a supported region is to pause the episode and build a proof of concept. He moved Fabric’s Quest onto a region that has apps.

• A table visual inside the report is the case Microsoft still does not cover. Mike likes translytical task flows for a button, and he keeps hearing the follow-up: edit this whole table. He wants a first-party visual bound to SQL, in a DirectQuery report, so the third-party tools stop collecting around that gap. Tommy’s objection is permissions. The real case is one role, with one responsibility, triggering one action. Putting that business logic in every custom visual does not scale, and a Fabric app can show or hide elements from Entra the way Power Apps did. His alternative, if a report must hold the action, is to embed a Fabric app in the report.

• Flip it: embed the report in the app. Mike’s revision on the spot is to start from the Fabric app, which can call functions, read and write a lakehouse, upload files, and mix DAX against a semantic model with SQL against a lakehouse or warehouse. The report page, or a single visual, is embedded where the app needs it. Direct Lake tables pick up a write immediately and can sit beside imported tables. A small refresh can target a table or a partition through the API instead of the whole model. He did this with a client the same week: a feedback form, Fabric SQL from a template, edits going back to SQL, in about 30 to 45 minutes. His translytical demos take longer than that. Tommy’s old Power Apps pattern matches. The event-planning app had a view-report button. Nobody updated a number because the report said 15.

• Translytical task flows survive as the exception. Tommy’s test is whether a translytical task flow does anything a Fabric app cannot do, other than living inside a Power BI report. His answer is no, on speed and on the same capabilities. Mike will not throw out the report framework. Desktop generates deterministic DAX for a visual, and a Fabric app that writes its own DAX has to be checked. Kurt Buhler has been making that point while building. A button in a report that sends data to a function still does not fall out of the report natively. Tommy’s split is about 95 percent Fabric apps, with translytical task flows for something like a streaming event. Jack in the comments asks if Mike is only redefining an enterprise app. Mike’s answer is that Fabric apps democratize application development the way Power BI democratized reports.

Looking Forward

Learn the agent well enough to get a process out of your head and into a Fabric app. Tommy is done waiting for writeback inside the report. When a first pass fails, Mike’s move is to split the task or have the agent plan the smaller steps before it builds.

Episode Transcript

0:01 mix. Fabric and A. I get your fix. Explicit measures. Drop the beat now. Pumpkins feel the crowd. Explicit Hello everyone and welcome back to the explicit measures podcast with Tommy and Mike. Hello everyone. Tommy, how you doing? I’m doing excellent Mike. How you doing? Doing well. Well, I feel like I’m on the wrong side of the screen here. I feel

0:31 like I should on the other side. We’ll leave it for now. The AI is going to get all confused up here when I when I try and edit videos later on, but we’ll see how good it is. Mike had really good points today. That’s what it’s going to say. Exactly. I’ll be clipped with Tommy in the front. Awesome., jumping in today, our main topic before we get into our news items is,, the signal in the report, right? Where does rightback belong inside your reporting? Now, now that we have Microsoft Rafen, it’s fab

1:02 Rafen apps, it’s a much better experience to write data back into fabric., and also there’s there’s this really compelling story right now is fabric is becoming a really robust backend for your applications and things that you’re building now. So, let’s talk about this. Now, there’s a translitical task flow that you can do some right back there and then now there’s also Rayfin apps. How do you choose which ones are better? So, let’s go unpack that today. That’ll be our main topic today. Before we do that, Tommy’s got a number of news announcement things

1:32 that we found across the internet we’d like to go over next. Tommy. All right. The first one is, hey, the next Chicago Fabric PowerBI user group is happening October 22nd. Same time, same place, the Microsoft building or office downtown Chicago, right off of talking, Mike, all about MCPs. We’re going to be diving into installing them both the local PowerBI semantic modeling MCP, how to install it, looking at the tools available, doing a nice cool demo, and also installing and getting started

2:02 with the fabric MCP, getting it installed and seeing what it can do. We’re going to go through basically how it can read across your one lake, execute on behalf of you., it’s pretty nifty. And like we did on our last user group too, Gil Reviv came to our last one which was awesome and he showcased. We like to we want people to show wins. So the beginning if you’re working on something and you want to showcase what you’ve been working on, what a platform. Yeah. Bring it. so we’re really excited. That’s going to

2:33 again happening October 22nd in downtown Chicago. So we’re super excited about that one. Time is flying by, Tommy. Like it’s already another month. We have another user group., we were just the other day was talking and we’re like, “Oh my gosh, two months have like just blown by here recently.” So, things are just really moving quickly. So, happy to have another user group. If you want to come out and join Tommy and be around the Chicago user group, it’s picking up momentum. They’re doing it every month. Got all booked up and join them at the Microsoft office downtown Chicago.

3:04 They have everything scheduled for the year. Hopefully, I have the next few for the 2027 scheduled as well. Again, every month, baby. So that’s our first one, but make sure to sign up and RSVP. One thing that we’ve been finding is if you do not provide your full name, if your username on Meetup H is like Ray, Ray, Ray 777, you can’t get in. So you need first and last name because you need to show your ID. So please remember to do that. So y Okay. All right. So that’s our meetup. Two great articles, Mike, on the fabric

3:34 blog. And let’s dive into this. The first one is about the growing ecosystem of Microsoft one lake. There’s a few major things here Mike in terms of this is obviously coming from Fabcon SQL con Barcelona but let’s dive into it. So the major focus is really seeing one lake has the unified foundation for analytics AI and copilot experiences. Let me I’ll run through the major ones and then I want to break one down with you Mike. First is the new IQ

4:04 sharing for govern business context. something they’re calling as what they’re calling IQ sharing expanding connectivity across the data state integration with Salesforce data dynamics 365 business central. The third is new consumptionbased onelay compute ondemand billing for onelay compute instead of requiring dedicated fabric capacities. Number four is stronger AI and agent instructions. You can imagine which one Mike here I want to talk about. Fifth is the fabric apps

4:34 extending fabric apps to support warehouses, lakehouses and semantic models, major governance and catalog enhancements. And then a few others for the growing fabric partner ecosystem URI for geospatial and a bunch of others. But Mike, the big one here, there’s two really big ones here. Okay. And I want to get your take on the AI first the IQ sharing and this AI and agent integration here because this to

5:04 me is pretty powerful here. So the the IQ sharing what this is actually called is basically this idea where you can actually allow organizations to share governant data, metadata and business across context across teams, partners and ecosystem. Basically the operational knowledge can be reusable across co-pilot agents and applications. This includes the ability to ensure

5:34 markdown and agent instructions and markdown or RDF file ontologies with plans for fabric IQ intologies and PowerBI semantic models. So this allows you to take those MD files for agent instructions and actually share them across your organization. Pretty cool. So I think this is like there’s been this challenge around your agents don’t really get to have a single place to learn things or have their similar

6:05 skills or it feels like this is a lift and shift of what you already built Tommy months ago of being able to get like you know hey I need to share content or specific markdown files. I do feel like it’s really nice that they’re embracing some more like file types here. it’s going to read other context files such as a markdown file. I believe there’s also this thing called the semantic view that they’re coming out with that has also been like another folder context or

6:35 files folder context inside the lakehouse as well. So as we look at like what’s makes sense here more and more of our world is using flat files markdown to to leverage this knowledge and put them together., when I look at like Obsidian, I look at I have another one, an open source version of that that’s log seek., I’m imagining notion is somewhere on on this as well. These are a lot of like markdown systems files

7:05 that an agent can rip over and do things with, right? So, Tommy, imagine with me for a moment, right? Let’s say your notion that you’re so much in love with it all goes to say that like it’s a bad thing. It’s not a bad thing. I’m just saying like it’s great. It’s a good tool. I’m the data all that goes into notion’s world notion’s land right I believe you can synchronize it with git but imagine the back end of notion now becomes fabric lakehouse how much more like that I think is

7:36 fairly impactful because now you have potentially like other data SQL databases and other things right next to where you’re doing like your day-to-day work things so I think this is a play for that I think this is a play for like a notion style system that lives on top of the lakehouse and when it creates files for you in the UI all those files are just getting landed back in the lakehouse and now you’re just talking to it. There is a definite potential there. So what you’re speaking of I actually

8:07 updated my skill vault app that’s available on GitHub that because before it would sync with your skills and you can push them into your GitHub repo push them back. Well, now I have a connect to notion. So, you can actually sync any skills that you have in your skill vault into notion and vice versa because of the skills API. But this, to your point, Mike, this right now wouldn’t do skills because it doesn’t support the skill structure with a skill MD and then folders and scripts. But the fact though that you can in a sense

8:37 share or have agents stored in fabric or at least the instructions for an agent, it’s I think the vision that you’re talking about is absolutely where they’re trying to go because I think just as much as our data needs to be governed, how do you manage the agents that you have and again outside of just data agents like reading and really you can have an agent for each of your department’s data, right? like what the weight is, what the importance are the metrics. This goes back to anthropic

9:08 self-service analytics article that I think has been very profound for me this year on how do you share that knowledge and also make sure that the data is in line with what an agent needs to know. So that’s the f first big part here with the share the IQ sharing. Obviously just like IQ in general I think we’re still seeing the big vision here of what it’s going to be in terms of like the really the application of it but it feels like they’re focusing on the infrastructure not so much what the output is right like like what the actual end

9:38 product is. Yes. All right. I’m going to play with this one. This one looks interesting. I’m interested to see where this goes. I feel like that’s a feature for other companies to develop on top of that Microsoft is opening up is where I think they’re going with that, but we’ll see. Yes. So, the next one, my friend, is stronger and AI integration. Speaking of which, so connecting to data is only the first step. What they’re allowing now is discover and use govern data in Microsoft Foundry and Excel. So, this

10:08 deeper integration Microsoft foundry is making it easy to discover and build knowledge directly from one lake data. Foundry users can explore their data estate including the rich metadata that we just talked about. So that’s really where the play is. The descriptions label connect to one lake items with a single selection to add knowledge to the to Microsoft. And that’s the first part. The second part of of the agentic side of this is the one lake table read API open parenthesis

10:39 preview close parenthesis which allows agents to secure serverless access to delta lake and Apache iceberg tables. You don’t need SQL for that. It automatically enforces table row and column level access controls. You can query the data and external data engines can also work with one like through the APIs. This makes sense to me. Like this like okay great you’ve got tables showing up but I don’t want to have to run a spark engine to go get data from it. I don’t want to have to run like it

11:09 should just be an API call. Go get this data with filter. It’s almost like GraphQL is being wrapped on round on top of the lakehouse items. So it’s just an API call. Hey, I want to go get this data. This is the table I want. Here’s the I want to filter these columns by this information. This is huge. thing. I think this is going to open up a lot of opportunity to like again if fabric’s a back end,

11:30 you got to have this. You got to have file storage. Goes right back to it. Yeah. So Mike Mike keeps saying backend because I gave him a lot of love. I just wrote an article inspired by Mike and credit where credit’s due on this. No, no, no. It’s an idea we’ve been pushing on together for a while now. But you were you were the first one I’ve ever heard fab. Yeah. pushing this concept, right? Yeah. F aab the fob. But I think this is a relevant term and like this is just adding more I think fuel to the fire here around this one

12:01 and I I Microsoft has not come out at least to my knowledge Mike I have not I don’t think Microsoft has said or coined the term yet fabric as a backend but that’s that’s what they’re doing without labeling it right it’s like that’s the game they’re playing right now with all the work that this is doing the fact that I can use foundry or use my one lake data including looking at the metadata, not just the tables and that API allows claude or my whatever’s

12:31 in Foundry or whatever my tooling is to connect to this securely, which is the most powerful thing. It’s not like I need to use this one lane of co-pilot. So, that’s Yeah. So, just want to give that update. And going off of that, the second article, Mike, is something you’re going to be very happy with. We touched on this on Thursday, but is Microsoft’s official stance Microsoft and Snowflake’s commitment to Apache Aussie. , and this is from our very good

13:02 friend Christian Wade. And really what this is going is Micros Microsoft is officially backing, how do you say, Microsoft Apache Aussie? It it’s an Aussie like the the kangaroo Aussie like Australian. Yeah. I guess it stands for incubating, but that’s what it says in here, but okay, cool. no, they’re incubating the project. That’s what they’re talking about. I got so Aussie is the the the Aussie project is open semantic interchanges was the original name. They changed it to something Aussie, so it was more unique. And then Christian Wade is

13:34 now jumping on this one, but he’s very vocal about this Apache group. And I’m actually on these meetings. These are open to the public. Anyone can go show up and and go do them. I’ve been attending them now for a couple weeks and trying to stay involved. I’ve submitted my first change to the AUI specification to make it better for us to like use the tool. So, we’ll see if my pull request gets through if they need any changes., so this is actually a pretty big deal. Everyone’s been ragging on Microsoft. Well, you’ve been you make semantic models, but no one can play with them. And data bricks

14:05 and snowflake are doing this, but Microsoft isn’t. Well, this is Microsoft stepping into the ring and playing with everyone else. They’re designing things around this OSI specification which is going to be I think very useful and my clients right now are very excited. I’m very excited about it. But I think this is very helpful for us pulling real information out of wherever you do the work. , if you work in data bricks, if you work in Snowflake, there’s logic that’s buried in there and

14:36 information that should stay there. But how do you share that back to your business? Where does that go? where do you store these things? I think this is going to be a very helpful tool to help explain a lot more of this and just add Aussie to our bucket list of weird AI words that we are going to interchangeably use. so one of the things Microsoft says here is it’s not just you saying it Mike C Microsoft says customers increasingly want platform agnostic agent experiences but let’s be honest Christian just listened to the podcast and said oh wow

15:07 we should really do this and that’s really the only reason right he doesn’t too busy for us I know I know but basically some of the things that Microsoft highlighted here semantic models support more than 38 million active users monthly active who knows how many agents that is monthly active agents. I’m going to keep you that’s another one that’s excellent that you put out. 94% of Fortune 500 companies use semantic models and again Microsoft is he investing to contribute to this experience the the Aussie project there. Snowflake and Microsoft

15:38 are collaborating closely with semantic views and the existing integrations continue to expand. So really one of the major near-term scenarios that they want to do is generate conversions between a snowflake semantic view and a PowerBI semantic model. What they’re also going to do when they want to help is make DAX an Aussie recognized query lang language and also expend support in Aussie for ontologies. So crazy and this is exactly

16:10 what you want to see because this way people can create agents for the Aussie language for the semantic language that would understand DAX. This is awesome for us Mike. Yeah, this is going to be I think pretty useful for us., and this is going to really push us further ahead, in order to make sure that like this is we’re going to have to think about how do we share common terms across different tools. This this is regardless

16:42 going to have to happen., what we’re incorporating here I believe this is like Kimble there’s a lot of good Kimble and good old data modeling warehousing techniques that are being used here. the the lakehouse pattern is now fully adopting great data warehousing. But this is one of these areas where doesn’t matter what tool you’re in, you still need to be able to talk across tools, right? And my biggest complaint right now is if I’m a data bricks person, I want to run and build everything as much as I can

17:12 inside data bricks. I want to build what makes sense there. If you also think about lakehouse now just just the Microsoft fabric ecosystem if we’re even in just lakeouses I have a table that table can be consumed and used in a semantic model that semantic model captures relationships and measures that live on top of those columns some of them are basic measures some of them are more complicated measures but for most measures you produce you could produce an equivalent version of that in SQL

17:43 well then how do you go from that same table through to a SQL data warehouse and then run the same code. So there’s this there’s this really this interesting concept the the compute engine doesn’t matter but the output or the results should be the same. So how do you capture two different languages getting to the same result and how do you test that? So I think this this OSI specification is going to allow us to be more flexible with how we describe what calculations

18:13 and what relationships exist. Relationships are simple, right? That’s easy. They exist both in SQL and in PowerBI, but when you add like filter context, when you have measures that are like really complicated, how do you convey that same information across both? because I should be able to write or build a DAX table inside a PowerBI report as well as I should be able to get to the same result inside SQL. And how do how do I know how this works with agents? what this is probably going to h affect and actually impact us is if

18:45 the plan is to integrate and make DAX an acceptable language for Aussie. Well, Mike, there’s probably going to come out best practices for DAX. Wait, what do you mean? What’ you say? So, okay. So, if if when when DAX is an Aussie recognized query language, it is it is because I already made the change in the PR previously and so it’s already there. Okay. So, my initial I have a PR

19:15 I had a PR about a month ago where I was like, hey, we need to put DAX in here. This is a language that’s not MDX. I need to have it. And then Microsoft came in and built their converter and then they added a bunch of other things that added DAX into the converter as well. So my initial idea got rolled into the system. So yes, it’s already there. DAX is an official language to be used. it does say in this was six days ago, but I’ll trust you. So since DAX is a query recognized language for AIE, it is okay. we’re gonna have to update

19:46 we’re gonna have to tell Christian Wade that he needs to update the article then. So,, where does he say that as if it is? He says a future plan. I think it’s in the video. , no, no, it says, “Looking ahead, we’re committed to staying in advance the open standards.”, include helping establish DAX as an AI Aussie recognized query language. Yeah, it’s already in there. It’s in the spec already. The the the spec has been updated. Okay. I read looking ahead as anyways.

20:17 Okay. tomorrow. This is a recent change regardless. I mean, this is like last couple weeks something is regardless what this is going to do is since it is an acceptable language, there’s going to be best practice around that to your point to translate that to SQL, right? Because most semantic models outside of PowerBI would use SQL. This is going to one enhance our skills and also two I think it’s going to change best practices around DAX because the language itself has to have acceptable

20:48 patterns right for calculator filtering and showcasing certain aggregations or time intelligence that needs to be in both that I would imagine so there are going to be best practices that come out of this so they can translate well to back to SQL or with the other languages right it can’t just be a standalone language so I think this is going to really Well, also the agentic side of this and also we’re going to learn a thing or two. John picks up on us. He goes literally John’s laughing about us about laughing. We are arguing about semantics

21:20 of things which is funny because we’re talking about semantic models which is also semantic. This is why so many companies have problems. This is good. I at the end of the day this is really good. I’m I’m very happy Microsoft is participating. I think this is going to be a big win for everyone. anyone can participate. So if you have an idea, if you want to improve the OSI specification, you can definitely go check it out. it is on GitHub. It is open source. There’s a whole area of discussions. So if you are interested in using it or incorporating into your your projects, you’re more

21:51 than you can. It is flexible right now. So they are accepting breaking changes currently. so you if you start adopting it, just be aware that it is inside the incubation stage. It’s still flexible. It may change and it may break some things, but you can start looking at it right now. The link is here in the chat window as well in case you want to check out the OSI specification. There’s also a Google calendar that they’re using to go to that. So, if you want that and want to

22:21 participate and show up to these events, there is a public forum. Anyone can show up and meet in these meetings and talk about what they’re interested in related to the OSI specification. Okay. All right. I think it’s telling Mike I Yep. It is topic. Main topic, dude. So, are you ready? I am ready. I’m trying to see if there’s anything else I needed to talk about here. I don’t think so. the only one one thing I did want to comment, Tommy, back on your one last

22:53 thing that I think is relevant to call out here that I thought was interesting. you had made the comment Tommy around

22:59 that article around for all the updates. Mhm. One of the items on here that I found was very interesting. A lot of people have been complaining about how data bricks doesn’t really play well with PowerBI and how there’s a bit of a bit of friction there. Also around this one lake area there is a read and write one lake data directly from Azure data bricks. And now there is a new item inside fabric which I do think this is worthy to call out. There is native storage in one lake is now supported and

23:30 in production. So there’s a an item that’s in preview currently that you can use directly inside fabric that will let your data bricks talk directly to the one lake. I’m not sure exactly how this works. but this is really interesting to me because I see a lot of friction between people in data bricks and people in PowerBI. Anyways, just FYI about that. All right, you ready to go my friend? All right, let’s go to the main topic. So this is a mailbag. So, thanks again always to submitting a mailbag and we

24:00 do not have a name on here. So, again, it please submit mailbags. We love we love our mailbags. We love when people submit them. We love also when you put your name so we can give you ample credit. But again, go to powerba. tipsodcast. There’s a link with a funny picture of Mike and I in a envelope. Click on that. Submit the form. We love your questions. So, here we go. Mike and Tommy, I’ve been following some of your conversations around power, how

24:30 PowerBI and fabric are evolving beyond traditional reporting, and I had a topic that I thought could spark a good explicit measures debate. PowerBI has gotten incredibly good at telling users what happened. But should the next phase be letting them actually act on that information without leaving the report? Mike Zach has spent years building this problem around as a former CEO of ATRX, a PowerBI native planning rightback platform now part of Vena. He’s been a pretty strong he’s had a pretty

25:02 strong point of view that the next evolution of BI isn’t another dashboard. It’s closing the gap between seeing a signal and doing something about it. There are a few directions you can argue through where or when rightback actually belongs in PowerBI. whether PowerBI should be remain an analytical layer or become an operational application layer and what happens as AI makes insights faster but business workflows remain disconnected. So again no name here but we want to thank you for your question

25:33 and let me probably yeah I’ve done we’ve all done that., but but let’s start with kind of the big picture here because Mike, rightback is absolutely one of our top five submitted things in the mailbag. If I were to look at all the different mailbags that we’ve received over the past five years, my right back’s up there. If not top five, definitely the top 10 in terms of categories, but I know it’s close to podium and there’s a

26:04 reason for that in terms of the need. We were talking about in custom visuals before fabric when there were custom visuals that came out that allowed for it and then as trans analytical and we’re in a new age of this now too and yet this is a recent mailback submission and yet we’re still seeing that there’s a great need for the ability not just to look at my data and email someone or act on it but really act on the data within the report. And

26:35 there’s always been complications with rightback because is it touching the semantic model or is it touching the SQL database? How is it actually working? As of today in October 6, 2026, we now have something incredibly unique that is generally available to us. And when we think of right back, where does it belong? Let me start with you with does the PowerBI journey change at all based

27:05 actually let me back up. Let me get your impressions of what Mike Zach has really talked about here and what his directions are. What’s your take initially from the question? Yeah. So right back is interesting. And so really I think we need to understand like what does rightback mean and how is it actually helping us or or hurting us in any way with what we’re trying to do. I think Microsoft has been very clear from day one especially when they started adding like power apps inside fabric when they did translitical task

27:36 flows of trying to write back to a SQL database something else that’s there. There’s always been this story around your data is needs action right insight to action. I may have heard that phrase a couple times from Microsoft. This is where I think we’re where this thing belongs., when I look at rightback, rightback is very specific. I think I think it’s too narrow of a scope in my opinion. It’s about planning or budgeting. It seems like it’ be a pretty solid use case. But

28:06 I think if you open up the scope to say what actions do you need to take upon seeing the information, where do you put a dashboard that has either adding comments or explaining why or capturing data? I think that is a good place. But actually, if I think about the actions part of this, I may want to go do something somewhere else. I may want to go call an API outside of the scope of PowerBI. I may want to go take other actions that are not just pure rightback as well. So

28:37 I I think rightback is a very narrow scope. I think another another initial thought there on this Tommy is rightback is hard on semantic models. There’s a there’s a lot more infrastructure that needs to be set up for a semantic model to actually do. I enter data into something and then that data immediately shows inside a table, a report or something like that. Or you know let’s say I’m updating some information in a dimension table. depending on how big your data is, that could be a lot of work to update that record inside that dimension table. It’s not like an operational database.

29:08 [snorts] This is a reporting system column or storage. We’re we’re shifting from one structure to another structure., you know, SQL databases are more row level based. PowerBI semantic models are more column They’re designed two different ways to be efficient in two different patterns. Right? So when we start talking about rightback I think a lot of people don’t understand the concept of columner storage and rowle storage and the difference between parquet

29:39 delta tables and SQL server tables. Yes, they still look like a table to the end user but the technology under the hood is substantially different and they’re designed for two different systems. So we just pause there. What do you think about that? Yeah, I’ll I think I have a concept here you’re going to like. I’m going to I’m going to introduce it because I think you’re 100% correct. Back in the day, the whole concept was writing back directly to the data that you saw, right? It was this column or this value needs to be from five to seven or a to

30:10 b, right? Based on what was already existing in the semantic model back in the day too, those semantic models were connected directly to a SQL database. Yes. So you would be in a sense pushing updating the SQL database. [cough] Excuse me. So to that end, I just want to hang on that point there. That’s a really good point, Tommy. The SQL database was really the backbone of the report, In in Rightback. And honestly, I’ll see that everywhere else I see that as being very much the same pattern, right? Right. Even now, if we’re if we’re talking

30:41 about Rayfin, right? So Ray Finnen and Rightback, which is another database. Yeah. It’s you’re putting that SQL database behind the scenes of that. And I think the difference here for me is and maybe this is just more conceptually for my mind when you build a semantic model you’re premputing all the aggregations or you’re you’re preloading you’re doing a lot of compute beforehand with the M language or the power query or the delta tables in the lakehouse right you’re premputing all

31:13 the data that way when you go to the report you already have the relationship is defined You already have the measure is defined. You already have like you’re writing a SQL statement essentially where the wear clause is already built. The filter clause is already built. The relationships, the inner joins or the joints are already built for you, right? And there’s also some hard rules around making a good performant model where you’re doing one to many, right? So you’re not doing a many to many join or

31:43 trying to avoid those inside your semantic model. So when you think of it that way, it feels to me a lot more like, hey, we’ve got this semantics layer that’s a bit inflexible, but it what you with you with you removing that inflexibility, you now get high performance and speed, right? And then with Rayfin and now with you know fabric translical task flows, all of those things are also still sitting on SQL. So now instead of precomputing

32:13 the data, you’re now using views to resolve your data at the time of need, right? And that allows you to like immediately update rows, add new records, change things, and then immediately see the results of those things right away. And the part two is even when the Power App integration came out, and bless Chuck Sterling for the amount of demos he’s done for the Power App right back feature, but it wasn’t necessarily right back. It was in a sense taking data from a model and be able to act on it. You

32:44 know, whether send an email with certain values. But Mike, I I I want to talk about right now because I think the the question does need to be framed differently. It’s not about where rightback lives. It’s there’s a new con concept here. We’re no longer focused especially with trans analytical and fabric apps about writing back. I want to introduce writing forward because what we can do now is not about affecting that last layer of the pipeline, the gold layer of data that I

33:14 want to update. Most things you’re trying to do is funnel it to another action. I’m not replying back to my data like in an email. I’m forwarding this information, right? because that’s what fabric apps obviously fabric apps kind of live in its own place but I think what a lot of people are trying to achieve when they’re looking at reporting right they’re looking at a sales number or they’re looking at you know all the orders or they’re looking at the operational equipment and they need to act on it the real goal is not

33:45 necessarily just to update that value right it’s to send an alert it’s to notify someone it’s to update a different system and the reason we always talked about right back back in the day was because it was the only really way that you could interact or in a sense write data at all. There was no really writing forward. But now trans analytical with the Python libraries available or with fabric apps where you can still showcase your data. Your goal is not just to update a value in the existing table that you’re in. your goal

34:16 or what really what is available to you really what people want is that ability to send out a send out a workflow or trigger something outside of that existing data and I think that’s a

34:29 different way we need to think about this is not a report is always going to be a report and I think we just need to finally I’m going to put my foot down there if you’re in a PowerBI report it is not meant to be operational it is not meant for me to change the different widgets and update existing data. There’s different applications for that when we can build that now pretty quickly. But when I’m in a PowerBI report, I don’t want that to be right back and I don’t have to now because I can build this in a fabric app. I can

35:00 use transit analytical but I don’t want to update the existing data and I think we just have to put our foot down. I’m going to put my foot down here that PowerBI is not the place that we’re going to do right back or is not really. We now have better layers and we now have better tooling to do right back where I no longer have to rely on a PowerBI report. I’m Yes, I’m going to agree with you there, Tommy. And I want to add to your comment as well. How much are you willing to pay for

35:30 rightback? Right. There’s a lot of tools that are not PowerBI that are trying to incorporate some level of rightback. So you could do analytical task flow. It’s built in. you’re paying for the different items that you produce, right? So you you have a function, you have a SQL database, you’re inside the report, and then you’re pushing data from the report back into the translitical task flow. Okay, interesting. It’s there, right? But when I look at a series, this other company that this Mike guy is a part of, there’s other companies out

36:01 there as well., there’s also this planner view that came from, it’s a new item that’s in there, but you have to pay for that separately. There’s more CUS that cost there as well. Microsoft official planner, right? Planner IQ. Yeah, the planner which came from that other company, it was Lumel was the original the creator of it and then Microsoft said, “Oh, we should probably have this in our product and said, “Let’s build let’s build a joint partnership.” And so I believe Lumel still builds it. I don’t ever use it. Right. It’s not part of my I don’t think it’s

36:31 needed, but finance teams, man, they are I I’ve showcased it to them and it is like showing them what an iPhone is for the first time. But build it or create like build it there or build it with a fabric app that’s included with everything like you could design whatever you want now. Like so so with a little bit of effort now and an agent you can get this basically the same experience customized to exactly what you need. So the question I have is what’s it worth to you? How easy do you want this to be?

37:03 And I’d argue per user, the payment per user should be very small. These companies charge a lot of money to have access to users and these kinds of rightback systems because it is difficult to set up and get going. But do you really need it? I don’t think so. I I don’t think this is I don’t really feel like this is rocket science anymore. And now with agents and the ability to have Rafe in I don’t think I’m willing to pay for any third party programs to help me build right back. I’m just going to figure it out.

37:33 And I I think there’s still a place for Planner IQ. I think the jury is still out. But to your point, we we talk about fabric apps a lot and we talk about I guess officially data apps. That’s what they’re really called at least when if you’re creating new artifact. So data apps, fabric apps, same thing. But fabric app sounds cooler. I I like it just comes off the tongue better. But well, there’s there’s already an app inside. Like there’s already like a section with an app button and like the app term is way

38:04 overloaded workace app. Yeah, too much. Yeah, this is true. content packs. How about that? So, has that been used yet? that hasn’t been around for a long time, but that was the original. Let’s reuse that. We’ll reuse that name. no. No, we’re not. That’s going to confuse everyone. I know. Content packs are not back. We’re not going to use that. No. Oh, those were fun., but data apps. I I know we keep talking about it, Mike, and you’ve been hammering it for months

38:34 now, and I finally come into the bandwagon. But honestly though, let’s If you haven’t built one, I’m going to challenge you listening today. If you have fabric and you’re in the right region, which now support us too because I migrated Fabric Quest over. So, thanks again for that., if you have not tried a proof of concept with a fabric app, you need to do that right now today. Pause the podcast, test something out, whatever you want to build, use cloud code. You don’t have to

39:04 install anything. It will do it for you. But if you have that ability with a, you know,, MCP and you have that ability or you have the repo, just get started writing this. And what you can do with this just tells me that the right the writing feature because Mike from here on out I am no longer going to say the words write back. It’s right forward or write data because honestly I think write back is I’m just thinking more and more about it. Write back is dead to me.

39:34 It’s dead. The writing feature of writing on my data really an application is available in a fabric app. If I want to show a report that people can really act on, I’m going to use a fabric app, right? And PowerBI is meant for the consumption, the governed layer of my information and data and organization where we’re currently at. A fabric app is used for the consumption of information, but also the ability to really act and write to my other systems

40:04 or trigger workflows. PowerBI doesn’t exist in that conversation. And I think fabric apps have such a capability because it’s not going to take me three months to do. Again, the person I’m talking to, yeah, you who hasn’t done a fabric app yet. In an hour, you’re going to have something that’s going to blow your mind. And again, just make it a proof of concept. The fact that I can connect that to any of my data already in one lake, there’s no point for me to use a custom visual. Sorry CEO,, the,,

40:34 the native planning rightback tool that’s part of Vena. Sorry for spending the time on custom visuals, but why? Yeah, I I don’t see it as being a big use case here. So,, I think there’s definitely a lot of places where people want to use rightback items, right?, this is this is where I really get like what is what are we trying to do? What what’s really trying to happen here?, right.

41:05 Yeah, it it’s how do I describe this scenario a bit better here? I’m trying to balance the value versus cost and time, right? So, and and so I’m going to lump time and cost in the same bucket here. I see XY chart in your head. There’s a there’s a there’s a chart running in my head here and I’m looking at okay where’s the sweet spot for where we’re going to be doing right back? Where does that look like on this scale of things? How how much how much money do I need to spend? How much time do I need to look at building it? Right?

41:36 While I do really like the translitical task flow and it does, transitical task flow lets you hit functions directly from a report. So if you want to take action on a thing from a report, you can submit. You can click a button and the report will then send information or data directly to a function. But inevitably whenever I’m doing these demos of translitical task flows, someone says, “Look, I have a whole table. How can I edit that entire table of data inside the report? And this to me looks like a mi this feels like a

42:06 slight bit of a miss from Microsoft. Instead of having third parties build the SQL table that is inside your report, there should be a SQL connected table visual that just appears that that that is part of like the semantic model. It is a direct query model, a report. Like I know this is hard, but the fact that there are so many companies clamoring to build something like this, it seems to me like Microsoft should just make this part of the product and have a a new visual that is the

42:39 rightback visual and it’s a table that shows up in PowerBI. That table that shows up and you can just edit things right there inside that table. I I feel like there’s so much clamoring around the edges here that Microsoft should own this and just make it a first class citizen and then just had this argument all go away. I am sure they’ve attempted this in the past. And I know that they weren’t in a partnership, but they were I forgot the company who was really on the cusp of building the custom visual table that allowed for right back. And they

43:10 invested a lot with Microsoft to do it. But Mike, here’s the problem though. what I can do with a fabric app right now from a rightback point of view when it comes to the entra access or the user access right to if I log in I have certain capabilities not just to see the data like role level security but also in a sense you can showcase and hide certain elements on an application based on the user very much like power apps but obviously in the fabric environment in powerbi if I have had a

43:41 visual you have to incorporate all of that role level security or all of those capabilities per user into that custom visual. That’s not a place for a custom visual to me like and I hear what you’re saying, but what’s it really trying to do? Right. Right. Back the theory of it or the generality is yes, one user updates a record does an action right to a particular record point of time. But that’s never the g that general case is

44:11 never what people are doing, right? It is a certain user with certain features or permissions needed or responsibilities to trigger certain actions, right? It’s not this general just everyone’s going to write back, you know, yes, no, approve, not approve, and anyone can do it. To build all that infrastructure and permissions and business logic into every,, individual custom visuals does not make any sense for me. And again, if I really need to do writeback or writing, again,

44:43 I I said I wouldn’t use rightback anymore. If I really need to interact and modify data, because I think that’s really what we’re saying here. I don’t think I’m no longer going to ask someone, especially Mike, if I’m dealing with Delta Lake or Direct Direct Lake, I’m not going to want anyone to modify that data in gold. Yes. Right. Okay. So, if we’re using direct lake too, now it’s a non-starter. It’s a deal breaker for any right back capabilities in a PowerBI report. Would you agree with that?

45:14 Yeah, I agree with that one. Okay. So, that custom visual goes away. No, it poofs out of out of it is gone from the ecosystem to me. You have to again, it goes back just like the fabric has a backend or fob as we said. I just like saying the word. There has to be and I started my article about the the whole iPad and I know you’re probably sick of hearing it. the whole iPad argument that Steve Jobs and I even linked to it in the art my blog article on prompting BI about the fact that Steve Jobs and I was just watching it

45:45 this morning. The third device between a phone and a computer has to do things better. PowerBI is not great or even good at rightback. It was just the only available tool. So, I’m going to push back on the custom visual idea or an

46:01 integrated visual, especially if we’re pushing people towards direct link to say, I don’t care. Embed a fabric app into your report. If you really want to do something, don’t use a custom visual table because every writing scenario is unique for the customer or process. Yes. I I think we’re getting to the era Tommy and this is where fabric apps I think really pushes us now is we are now able to

46:33 hyper customize what does our business do? How does an how do I say this? the AI is now reducing friction for us to build really custom processes around data rewriting back and forth that that process and now that we get fabric apps we get even more of this and where I was going with this one Tommy and and I think killing the idea that the that the report is not letting you write

47:03 re right back maybe we should flip the head where I was going with this is we should probably flip this okay we should probably flip the reaction We should be putting embed reports inside our fabric apps when we need visualizations of data. Right? So let’s I’m just literally unpacking this now as you as you said that last comment Tommy which I was kind of like looking at at something to get some more information here inside an app. The app is super

47:33 flexible. You can have functions. You can read and write back to lakehouse. You can upload files. you can process things like the application is very creative and what you can do there what you can dream up. Okay. So what if we take a portion of what the PowerBI report is doing and say look instead of trying to embed the readwriting back portion back into PowerBI flip it. The PowerBI report is still there but it is part of the actual application right? we can connect a semantic model to a fabric app. You can run DAX

48:05 queries against the semantic model., you could mix DAX queries and SQL queries together inside the fabric app. So you could use some information from the semantic model and some information from a SQL database or a lakehouse or a SQL data warehouse all the there’s more things coming to the fabric app. So, I think if you flip this mentality on its head and say, “Let’s start with the readr capability from the fabric app and then pull from there what you need on

48:36 the report side. Here’s a report page. I’m just going to embed it here. Hey, I’m going to just snag this visual out of here and render just a visual somewhere in my app.” I feel like that makes a lot more sense. And then the only thing you’re trying to negotiate is this little part around how do you keep the data synchronized and what data are you trying to write back or change and what does that do to the semantic model right we now have direct lake and import mode inside semantic models. So

49:06 you can do direct lake tables which means as soon as you update the table in the direct lake you automatically can use that with imported data. so there’s there’s actually a large amount of options to figure out how to refresh the semantic model. We we’re always going to have to deal with this issue of the semantic model is precomputed and if you’re not premputing it, you’re always going to go back through like a direct connection to something behind the scenes. It’s never going to be like

49:37 fully like a SQL database. It’s just not going to be there. And so this is where I think fabric SQL really lit up for me. When they started announcing this, I started dabbling with it. Tommy, this is where things got really interesting. when when SE like you were always a big proponent of SQL inside Power Apps and I was like yeah but I don’t want to pay another license to go all the way over to Power Apps and do things right and then you could put Power Apps inside of Power Report but now you could do PowerBI reports inside inside a fabric app or a Power App

50:08 and I thought we’re just so close. It’s just the pieces don’t quite fit together correctly. And so when I think about fabric apps I feel like okay these pieces are fitting my mental model of what I want to build. So that’s where I was where I’m thinking this is going. I think you’re on to something. So, it’s funny you say this because my general concept with my when I was doing a lot really when I was doing a lot of power apps was most of the requests came well we do have a report that’s related to that application not

50:39 necessarily a onetoone but okay what was a good example we had one for event planning and it was to basically to make sure that everyone was registered it was all a power app but people wanted to view the report not so they were going to update based on that but just see the general members like the normal report that they viewed. So most of my Power Apps had a view report button that brought them just to an embedded page or a few of the embedded pages of the power the original PowerBI report. I think it’s important to know

51:09 at this point that it was never the intention to look at the report say oh it’s 15 update this number. There was a in a sense indirect relationship between the two but it was very important that that was existing right we’re saying what are our again the event totals how many people haven’t registered oh it’s about 25 people haven’t replied let’s go to the back to the app and I’ll go to that event and that workflow worked well and this is exactly what you’re talking about right this is what

51:40 makes more sense leave the report for what it’s good at right powerbi is good at and then leave the power or the power app or the the application interface. Correct. With fabric apps to do those again recordbased items. You are not doing bulk updates in you can. Yes. But you again you’re dealing with an account, a product, a category and let the app be best at that and then leave the in related report

52:11 somewhere that can be even embedded in the fabric app. There’s a there’s another part to this that I think really pushes us. There’s another area of this that really pushes me why I like fabric apps a lot more here. And let me try to paint another picture. This is very technical, so just bear with me. When you have a semantic model that’s being supported by a a report, right? Yeah. Put your glasses on. Tommy just got his I’ll feel smarter. Smarty smarty pants glasses on. so

52:41 with the fabric apps give us direct access now to fabric functions. Mhm. And a fabric function gives us direct access to APIs that are all across fabric. So one of the challenges I think that was an issue for us previously was when you were building a report there was very little interaction that I could do to call specific APIs inside fabric. And the reason why that’s important is because if you do this calling those APIs would allow you to refresh a model

53:11 and it’s not the whole model, right? If you’re updating data, the entire model does not need to be changed. We’re talking about a very small change. And a lot of what these programs do like when you’re talking about like aceries and all these roundrobin tools that are doing refreshes on reports what’s happening I think and this is my understanding of the system is you have a visual on the report page you’re typing in a number there and you’re hitting save the visual is retaining that updated

53:43 number even though it’s either wrong or you’ve updated the source system doesn’t actually have that difference in number. So what happens is at around that time you are hitting save or you’re updating a number on that chart that chart that visual is then roundrobing a message to something on the back end likely a SQL database. Hey this record on this visual is being updated right and then when it does that the direct query is then the visual is like waiting

54:13 for that data to be actually updated on the back end to actually show it correctly on the front end. So it’s like a maybe a second or two of of loop around time right update the visual push in the number visual says hey I’ve updated my number send a message to some SQL database SQL database says oh I know what to do I will update this record and then it the view the information that’s actually rendered in the visual the the visual can say okay now go grab the latest data from that view and the view updates and you get the new data. So there’s this period of time where the

54:44 number you’re punching in doesn’t actually match what’s in the visual, but the visual has to know like, oh, I have to understand this versioning concept, right? So that’s a lot of work to get things going here. Yeah. But with but with fabric apps, you now have access to an API that can do very similar things without having to pay for a program, a third party, another program. All of this now is built into the application. And so you can say, “Oh, I want to build an input screen or a form or I want to pop up a window that

55:16 lets me edit this data.” And all this is now possible. There’s now no barrier to how you want to enter that data. And then once you hit submit, it then goes right back to likely the SQL database. And then you can trigger from there not just a refresh of the semantic model via an API. you can say I want to refresh this table, this partition on this bit of data. And I think that’s where you can still leverage like import mode and these other things where you’re doing a

55:46 very small refresh on top of a potentially a very large model. I was just with a client recently and we were talking about a feedback form and he was he was like well maybe we should build a fabric app. I’m like yeah let’s do it. So in about 30 to 45 minutes, we had stumbled our way through the fabric app creation. We had spun up the datab the the fabric SQL the template. We picked a template. We said here’s the information I want to ingest. Here’s where I want to put it inside the SQL database. And so

56:17 we did the whole thing the whole thing. Stood up the the fabric SQL made a form, enter the information, pushed it back into the SQL database, and was able to update and edit records. That’s I’ve done demos on fabric transolitical task flows way longer than 30 to 45 minutes to get it going. Yeah. Yeah. Right. So, this is this is a huge shift in what we can do here., and now you need more of a technical eye. You have to know what you’re doing to get the performance out that you want. I

56:48 think in this situation, what do you think about? No, I I I completely agree. First off, as you’re talking about that demo you did, I really wish I started my company now because when I I first I created a timing app in Power Apps and it took me forever and it was good, but it was basically logging my hours and worked with QuickBooks and all this integration and my projects and estimated time. I could have done this in two hours with a fabric app and it would have been looked so much prettier. But I I think the the bigger point here is

57:19 the transanalytical I think while it’s a great concept for me I think it’s probably going to go to the way of metric sets right it

57:31 because again it’s is it better let me ask you point blank and I’ll answer it first but is trans analytical task flows better than anything that a fabric app can do and to me from the action point of you no because I can have better integration, faster development and the same capabilities in a fabric app. The only thing that the feature that it has that it lives inside a PowerBI report but in terms of what it’s capable of

58:01 to me I don’t think it serves a better purpose in terms of its writing feature than a fabric app. So would you agree or disagree that can a trans or would you actually state can a trans analytical task flow do anything better than a fabric app in terms of the features that it has not just that it lives in PowerBI there’s some built-in features that are there that are just part of the product that just work so

58:31 an app you have to so in an app world you have to throw away the report layer and start Right? You can’t you can’t use an existing report the way it is. So you know when you start saying let’s go fabric app I got to like think about okay I got to chunk the idea or the concept report. Now you can also fully rebuild any report that you have. I’m fully convinced that any report you build in PowerBI could be replicated inside the fabric app 100%. There’s something that the PowerBI

59:01 desktop does really well is it generates code for you very easily inside PowerBI desktop like the DAX that it needs to build that visual. It’s very deterministic. It’s not going to make up weird stuff. It’s not going to generate new measures. So if you’re trying to build a visual inside a fabric app, it writes its own DAX to generate that visual, right? It may be using your measure, it may not be using your measure. you got to check it. So, this has been a caution tale

59:31 from Kurt Buer and I as he’s been building things with this as well. Like you have to check the DAX that’s being written inside the the Rayfin app or the the fabric app. So, from that perspective, we have to be mind be careful about it. But Tommy, there’s nothing that I’m aware of that makes it easy for you to put a button in PowerBI and press that button and have that data sent to a function. You can’t you can’t do that. You can do that with an app. it. The equivalent of that exists, but you don’t get it out natively out of their port.

60:01 Right. Right. So, so to me, it’s like these are two separate worlds. You’re either picking one and going with it. Where I’m a little bit nervous, not nervous, where I’m a bit more caution here is I really do like a lot of the structured reporting paging. There’s a lot of really good stuff that the PowerBI report has done to make it easier for us to build reports. There is a framework. It works really well. I don’t want to throw that out. I like it.

60:32 I still want to use it. So that’s where I’m still trying to figure out how to balance this. where does righteback happen?, but I think if I’m if I’m being honest, more and more of my rightback scenarios are going to just start with a rafen app. I’m going to start there. Yeah. Yeah. And unless they I think it’s very much Sorry to cut you off, but I No, no, go ahead. Trans analytical to me in incredibly unique scenarios, right? It’s it’s 95% of the time what I’m hearing you say is is going to be in a fabric app in Rathon. There may be situations where

61:04 transnalytical can serve a better purpose, but I think they’re going to be in a sense like the streaming events like a very oneoff situations that I may want to use. And I think just right now the bulk gen,, from general to incredibly integrated work streams, fabric apps can achieve what I needed to achieve and or company can achieve what it needs to achieve. There may be situations that trans analytical just makes better sense but I think it’s going to be like one out of 10 or you

61:34 know one out of you that’s going to actually make sense if that I think with the addition of fabric apps more and more people are just going to abandon trans trans analytical to move away from it. I I don’t think we’re going to need it as much. Jack in the comments here I think makes a really good point. he says, “Mike, aren’t you just redefining what an enterprise app is?” Well, I’m not redefining what an enterprise app is. think but I think what Microsoft has done here with fabric apps is in the same way Microsoft democratized reports with PowerBI

62:05 they’re doing the same thing with application development apps are being democratized in the same way that PowerB reports were that’s what’s happening here you can bundle up your business process land it inside a fabric app and that becomes now your official business process moving forward and I think that’s where things get really interesting yeah 100% dude love it. Any final thoughts, Tommy, before we wrap here? Dude, I honestly I think this is a great conversation. Maybe we finally put to bed, we put to rest right back in

62:36 PowerBI. That’s what I’m hoping. That’s what I’m hearing today is let’s understand what we have available to us today and let’s utilize that and take advantage of that. To me, I’m putting to bed the fact that rightback belongs in fabric apps and in the off situations, I’m going to use translical task flows. I’m not going custom visuals. I’m not hoping and praying that another writeback feature gets into PowerBI is meant for the consumption. It’s meant for the interaction. It is

63:07 not meant for the writing. So rather than write back, Mike is going to be cheesy. I’m writing forward. You’re writing forward. I think I think Microsoft is democratizing app development the same way they de they democratize report development. We’re exciting point right now. The the apps are very exciting and I think I think using an agent to help you build this is going to make it even easier for more access to the

63:37 broader part of the business. People are going to build amazing things. there will there is one knowledge piece that everyone must learn. You must learn how to work with agents. Yes, fabric apps isn’t isn’t going to be for the light-hearted. You’re not going to write the code yourself. It’s just not worth it. you’re going to spin apps with an agent. So, one of the core new, I guess, lessons people need to learn is you need to go get familiar with how to use an agent in order to build things.

64:07 Get good at it., whatever you need to do, go buy a subscription, go spend some time, go figure this out. There is technique, there’s the way you prompt things, there’s what you describe to do and what not to do to the agent that gets it to build what you want. So, the main goal here is you’re going to have to figure out how to get what’s in your head into the mind of that agent thing. And odds are the models are so good right now, when my agent is unable to build what I want first time, it’s likely my prompt or I didn’t describe enough or I didn’t give enough information to it to

64:37 actually deliver what it needed to. And so, if it’s a really big task, I’ve got to think about how to break it down into smaller tasks or have it plan smaller tasks first and then have it go execute. So, I’m fully aware that this is a new world that we’re stepping into. There is learning that needs to happen here for this to work well. 100%. That being said, Tommy, where else can you find the podcast? You can find us on Apple, Spotify, wherever your podcast, make sure to subscribe and leave a rating. It helps us out a ton. Do you have a question or topic that you want us to talk about a future episode like today? Head over to

65:08 PowerBI tips PowerBI. tipsodcast. Leave your name and great question. And finally, join us live every Tuesday and Thursday, a. m. Central on all PowerBI social media channels. Thank you all so much. And we’ll leave you with this little treat. Oh boy.

65:50 Monday memo says accelerate. Burn the tokens. Don’t you hesitate. Dashboards [singing] watching every prompt you send. Use too little and you’re at the end. Secret agents. building systems in the basement. Secret agents [singing] under the radar. Just trying not to get the paper.

66:24 Tuesday memo flips the script again. Token budgets bleeding. Cut it then. Same old boss with a brand new fear. [singing] Yesterday’s hero is over here. Secret agents. Secret agents building systems in the basement. under the radar. Just try not to get the paper.

67:04 So we spin up tools they never bless. [singing] Shadow stack to survive the test. Buy a seat for Claude when the gate gets closed. Buy a seat for Gro [singing] when the path gets froze. More tokens, less tokens. [singing] Make up your mind. We’re solving work they leave behind. Unsanctioned ain’t the dream we chose. It’s the only way the deadline

67:34 goes. Business users in a quiet war. Sanction path don’t open anymore. [singing] If the company can’t decide the lane, secret agents keep us in the game. Secret agents. aent building systems in the basement. Secret agents under the radar.

68:06 Just try not to get the paper secret agent.

68:43 Explicit measures. Pump it up. Be it high. Tommy and Mike lighting up the sky. Dance to the day. The laughs in the mix. Fabric and A. I get your feels. Explicit measures. Drop the beat now. H feel the crowd. Explicit

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.