Self Aware in Data and AI Literacy – Ep.564

Explicit Measures starts with two changes that ask teams to plan, then spends the hour on how self-aware an organization is about data literacy and AI literacy. Kurt Buhler is back with Tommy Puglia and Mike Carlo to connect those literacies to culture, trust, and the habits that keep fast building reviewable.

News & Announcements

• Moving off ODBC — a self-serve scanner for your ADBC migration — Tommy walks through the PQ ADBC advisor, a scanner for Fabric and Power BI customers as Power Query moves from ODBC to ADBC (Apache Arrow Database Connectivity) in Power BI Desktop and Microsoft Fabric, including dataflows in the service. He reads two stages: ADBC becomes the default for new connections while existing ODBC connections keep working, then a cutover where ODBC is disabled and those connections have to use ADBC. The advisor scans semantic models, dataflows, and pipelines, finds connections still on the legacy driver, and sorts items by risk — what will fail, what needs review, and what is ready — so a team can scan, fix the failures, validate the workloads, and scan again. The connectors he names are Snowflake, Databricks, BigQuery, Redshift, Dremio, Spark, generic ODBC, and OLE DB. Mike adds Azure Databricks, Hive, and Impala from the migration notes he drops in chat, and he reads this as a cloud-source change tied in part to lakehouse tables, with SQL databases left on their current drivers and ODBC still available for other sources. Tommy’s early Power BI work depended on ODBC through a virtual machine; he sees less of that now and still wants organizations to plan. Both hosts are glad the scanner shows where those connections live. The repair stays with the team.

• Power BI Q&A retirement reminder: February 2027 timeline update — The legacy Power BI Q&A natural-language experience, previously set to retire in December 2026, now runs through February 2027. Tommy reads it as covering the user Q&A experiences and the Q&A configuration tools, with a table of recommended replacements that often lands on a different experience: Copilot in Power BI reports for Q&A on a report, Copilot in Power BI mobile for the virtual analyst, Copilot for embedded Q&A, and prep data for AI for the old Q&A setup. He also reads that a Fabric Copilot capacity can be created on F2 and higher, which puts Copilot on a smaller capacity once it is enabled. Mike found little value in the old Q&A and would move anyone who still wants to ask questions of their data to Copilot. The requests he hears now are about handing Claude the data, or letting an agent talk to the semantic model, which is a different job from a chat box dropped on a report page. Tommy’s sticking point is the license. The old “show me sales by country” interaction added no capacity charge and ran in Desktop and the service, on Pro and Premium Per User, and he remembers it working on a free license in the training he used to give. The replacement sits on Fabric capacity. He sees that as a real barrier for organizations that still have Fabric turned off, and as a push toward Fabric for everyone else. He also wants a clearer line on narrative visuals and smart narrative summaries. As Mike reads that part of the article, embedded Q&A should look at Copilot for SaaS scenarios and at those narrative visuals. Mike thinks a team can also buy a smart narrative visual and drop it on the report, which keeps that path off a Fabric capacity. The route he would push is Fabric IQ, or the Fabric MCP server, so Claude and the other agents he already sits in front of can talk to the model directly.

Main Discussion

Topic: How self-aware a team is about data literacy and AI literacy once Microsoft Fabric and agents make building easy

Kurt Buhler of Data Goblins is back, after a YouTube comment called last week’s episode the best one ever. Mike asks him what the two terms mean, and where someone sits on that maturity spectrum. The hour stays on the gap between a person who can read a chart, a person who can argue with a number, and a person who can explain a decision an agent just made.

• Data literacy is reading the number, knowing what it means here, and being able to argue with it. Kurt’s definition is how well you can read and interpret data in order to work with it and make decisions. He treats it as cultural and individual: a superset of math and statistical literacy, digital and technological literacy, visual and information literacy, and critical and analytical thinking, sitting next to data culture and change management. Mike hangs on the variation. It changes by department and by person, and walking into an organization rarely comes with a course in how this company talks about charts. Tommy adds the piece they have hit on before: question the data, communicate it, and argue with it so the decision gets better. His point to Mike is that a consultant’s data literacy at a new client starts at zero until the pain points, the thresholds, and the meaning of those ten metrics are learned, because some of the ten matter more than the rest. Kurt’s sign of low literacy is the decision made on tribal instinct, and the room that cannot engage when the data shows up. Mike has been in the awkward version, where real numbers make someone sheepish. Tommy’s analogy is a philosophy book: you can read every English word on the page and still miss the meaning. Kurt puts empathy under that. When technical terms and business lingo pass each other, the consultant’s job is to build the bridge in both directions. Mike, who spans the business and the technical side, calls that empathy underused. Two ears, one mouth.

• Fabric is a tool. Direct Lake still asks a business unit to understand more. Tommy asks whether Fabric raised data literacy or did the opposite, now that an agent or a click can stand up a lakehouse. Kurt’s answer is that the tool on its own changes nothing, and that scaling a Power BI practice into Fabric carries higher prerequisites. He does believe agents can lower that barrier and help literacy grow, when someone facilitates it. His example is Direct Lake. He advocates for it as one of the quickest ways to query a lakehouse, including an ephemeral model you build and tear down. Telling a decentralized analytics group to consider it also tells them to manage the lakehouse, use notebooks, and vacuum. Mike agrees Fabric is one tool among many, and he feels the difference from the Azure path, where a blob account, Databricks, or Synapse had to be turned on. In Fabric he clicks create and the item is there. That availability pulls people in more often, and it adds responsibility: who in the company actually knows the data lake, how it shows up in a day’s work, and who manages it.

• AI literacy is knowing what the tool is, when to check it, and which job it fits. Kurt describes it as awareness of the tool. You know the output can be non-deterministic, you remember it is a tool, and you can say what an appropriate use is. That awareness has to meet a task you could already do on your own. He takes the deskilling worry seriously. Critical thinking is required to use these tools well, and he sees the same tools training people to practice it less. What he also sees, in pockets of the community, is coding agents walking people into work they would never have touched. The point of that walk is understanding: why Rust, what it is good at, why you would choose it over Python. Tommy hears their data-literacy definitions as the consumer’s job, which is most of an organization, and the early AI-literacy talk as the builder’s job. Kurt brings the BI engineer back to the same end: represent the business process and the business logic accurately in the model, then communicate it. Building is the wider blast radius, and conversational BI still needs the same awareness, because a miss gets built into whatever you ship. Tommy’s line from his own writing is that agents raise the floor and can lower the ceiling. Kurt names the risk for people early in a career as the unknown unknowns: they can reach deeper into the fire, and experience is what makes you aware of the gaps you still have. Mike has watched generative output leave the building with no one reading it, and he points to a Microsoft paper he places around January 2025 on generative AI and critical thinking.

• Slop is missing effort and a missing person, and “the agent did it” is the workplace version. Asked to define AI slop at the dinner table, Tommy calls it output with no thought behind it: a general prompt, a general result, no purpose. Mike identifies it by the patterns he now recognizes — stock phrases, the reflexive “you’re right, I’m sorry,” em dashes, the same shape of paragraph — because the model is choosing the most likely next word from everyone it has read, while Tommy writes like Tommy. Kurt has spent a long time on this. He used to hold a hard line on AI in art and writing. He now counts lack of effort, which is Tommy’s test, and lack of individuality, which is Mike’s, and he has seen work that was almost entirely AI-composed and still carried an original idea, or carried an idea someone with dyslexia needed help getting onto the page. Human attention, in his mind, deserves human effort. Fully generated text can still clear that bar. Inside a company, the slop is the symptom: the email no one edited, and the semantic model whose author says the agent did it. Kurt’s test is whether you can explain the decision. If the model converts currency, you should know that, and you should know why this pattern was the one you chose. Tommy calls “update this semantic model, it’s broken” a slop prompt. The literacy he wants for a BI developer is the workflow: intentional instructions, and context you decided to provide. He uses AI to write because he talks faster than he types, with skills built from the podcast transcripts, and he credits Mike for the phrase he keeps — build the thing that builds the thing.

• Literacy and culture move together, and the hero gets more dangerous when one person can orchestrate the agents. Mike ties any rise in data literacy to a rise in data culture. The two pull on each other, and he extends that to AI. An outlier who is far ahead of the organization creates friction. Either the company raises everyone, or that person starts looking for a place at the same level, because the mismatch wears people out. Kurt reaches for the data hero, which he believes comes from the DataOps cookbook: one capable person shoulders the data processes, and the reports stop when that person is out, because nobody else knows how they work. AI turns that hero into the story of one contributor who can do the whole project by orchestrating agents. Tommy ties it to Tuesday’s conversation. A prompt and VS Code are enough for the three of them to build three Fabric apps, with three schemas, three interfaces, and three ideas of what data gets in, unless they share a process. BI already had tribal report styles wherever a team lacked a common language. He defines culture as traditions, process, and language, and he thinks that set matters more now that the building is easy. Mike borrows a point from Rob Collie’s book. If every line of an application were already written, typing it in could be a matter of days. A data project takes weeks, months, or years because of the coordination. AI makes it easier for more people to ship a full solution alone, which is the Power BI proliferation problem again: some models with strong DAX, some with trash DAX, some shaped well, some pulled straight from the operational tables. A new author does not yet know that a semantic model wants dimensions and facts, and the DAX gets heavy because the shape is wrong. That unknown unknown now covers apps and email, too.

• People are shipping artifacts they never opened, including measures that live in a Fabric app query. Kurt calls vibe coding self-service BI on steroids. He is meeting people who build and share dashboards as artifacts in Claude and prefer them to the Power BI or Tableau reports. He has barely opened PowerPoint in a year; workshops go out as HTML from a tool he has been building. Mike has watched an HTML file render inside Teams, and he has opened links to HTML that Claude Code generated and saved. The question they land on is what happens when Word, PowerPoint, and Excel are things you generate for one audience. Kurt’s caution is the other half. Rebuilding a mature tool teaches you how much professional work is already inside it. The super individual contributor still produces more output than anyone can review, and he thinks almost nobody making a Fabric app is reading its code. He has seen metrics defined in the DAX query of the app, a SUM or CALCULATE written inline, where a measure in the semantic model should be, and he has watched someone insist a dashboard was Vega when a look at the code showed it was something else. The literate response is to know which decision was made, and to put a check in the workflow that catches a metric born inside the app query. That is the thing that builds the thing, aimed at a failure they have already seen.

• A data and AI czar starts with a standing conversation, a position from leadership, and a gate on the powerful tools. Tommy’s scenario is a large organization where one person controls the budget: rapid development, a stronger data culture, more trust, room for individuals to create, and a way to validate what they create. Kurt starts with a cultural problem the way he would start an adoption problem. A data café or office hours, optional, aimed at one community, ideally in person, where the behavioral cues are in the room, and remote when the company is too big. No question is a wrong question. Walk people through the Power BI app. When a suggestion ships, name the person who brought it, and keep the rhythm until it is part of the culture. Mike hears the Fabric adoption roadmap in that answer, and he wants a position on AI from the top before anyone tells the whole company to start using it. A blanket invitation produces side quests and one-off builds. He would extend a data center of excellence into AI, and he would keep a small tiger team — his first name for it is a black-ops group — free to experiment and required to come back with findings. The lesson he uses is a Fabric app that was asked for Plotly, came back on D3, broke, and burned tokens. Someone then has to write the check, including an agent that evals the app against what the company will accept. Kurt wants the executive in that story to lead in public. An executive who posts unedited AI text, then complains about other people’s AI text, teaches the wrong lesson. The version that has motivated people this past year is a leader who shows a thing they made, walks through the process, names what went badly, and steps down. Tommy’s phases sit on top of that. First, proofs of concept that are allowed to stay out of production, including one the leader builds. Second, an agentic team with a shared directory of skills and agents, doing for AI assets what BI discovery did for reports — a team that makes flyers all day may need a skill, and the team writes it. Third, levels. Demonstrate that you can build something trustworthy, and then the agents or the MCP server open up. Kurt is frank that the tools are dangerous enough to need that segmentation, that locks are a weak substitute for the right person in the chair, and that people need a visible way up the ladder. The shadow version is already in the room: a company that believes it only uses Copilot, then discovers personal Claude subscriptions a couple of weeks later and panics about compliance. Kurt compares a total lockout to the old Power BI lockdown that pushed the real work into Excel. Withholding the tools people can see everyone else using clamps the team. Opening Claude Code with no limits gets expensive and unsafe in a hurry, the same way an open Power BI tenant becomes thousands of workspaces and a capacity bill. The balance is the job.

Looking Forward

Treat AI literacy the way the organization already tried to treat data literacy, and keep asking what the output is and where it goes. Tommy puts as much weight on the culture around that trust, and on the context of the team, as he puts on the skill. Kurt’s close is empathy in both directions: for the people adopting the tools, and for the context an agent needs, with the agent kept as a tool. Critical thinking has to be part of the culture, and he is clear that culture arrives through change management, in small steps you can repeat, once the company can say what it is walking toward. Mike brings back the three C’s Kurt named last time — critical, creative, and curious — and he is honest that curiosity is the one he leans on. Power BI’s blast radius was reporting and analytics. AI widens it to apps, processes, and workflows. The message has to come from the top: leadership names the direction, the teams that are experimenting report back, and the organization learns this as a skill it holds together.

Episode Transcript

0:01 Dance to the day to laugh in the mix. Fabric and A. I get your fix. Explicit measures. Drop the beat now. Pumpkins feel the crowd. Explicit measures. Hello and good morning. Welcome back to the explicit measures podcast. Here we are again jumping in. Tommy, hello. Good morning. How you doing? Good morning. Hello, Mike. I’m doing excellent.

0:31 excellent. Great. Good to hear it., we’re going to continue on with talking more things around AI and this particular episode is going to be talking around data data literacy and AI literacy. So, just kind literacy and AI literacy. So, just talking about how literate are you or of talking about how literate are you or how comfortable are you in the data and AI era and how do how literate you are in these spaces. So, we’re going to talk a little bit that’s our main topic for today. But before we get into that, Tommy, you’ve got a couple news items you’d like to run by us. All right, let’s run through the first

1:02 All right, let’s run through the first one. Mike, this is this means more to me than this probably does mean to you, but this is on the fabric blog. Moving off ODBC, ODBC, a selfserve scanner for your ADBC migration. And what this actually is in introducing is a tool called the PQ ADBC advisor. so many the buzzwords you’re saying right now, this is like when I talk to my family about anything that’s data related, they’re like, “What the heck are you talking about? What are you

1:32 heck are you talking about? What are you What are you describing?” The old lama pi. Yeah. Yeah. Exactly. Words that have no context to other people. We say these hugging face lama pq. Yeah. So, jeez. it’s a tool designed to help fabric and PowerBI customers migrate from OBBC to the Apache Aero database connectivity drivers and Power Query is moving away from OBBC to ADBC and PowerBI desktop and Microsoft fabric. So

2:02 PowerBI desktop and Microsoft fabric. So this is not just a power query thing in the web for data flows. There’s two stages ADBC becoming the default for new connections while OBBC connections continue working and then a cut over ODBC will be disabled and all connections must use ADBC. and the what the advisor does is help scan semantic models and data flows and pipelines. Identify connections still using legacy and classify items by risk. What will fail? What will need review? What is ready? And they have that. So

2:34 What is ready? And they have that. So you want to scan, fix items that will fail, validate those workloads, and then rescan. The support of connectors are Snowflake, Data Bricks, Big Query, Red Shift, Dreo, Spark, and Generic OBC,, O LLDB , O LLDB connections. So Mike, I when I first started with PowerBI, most of our things were done via ODBC and it was a pain in the gun because we had a virtual machine that ran that allowed us to do so. But our lives dependent on ODBC.

3:07 But our lives dependent on ODBC. I know it’s not as prevalent at is at least I have not seen that but I can guarantee you a lot of organizations need to look at this and need to start planning now. All right, this is so agreed. I think this is definitely the right the right approach here to start talking about this one. There is a really good also article. So in this article I will put the link to the OBC and ADBC these changes that are occurring. I believe one of the main reasons why this is occurring is because of the lakehouse

3:37 of the lakehouse and the the use of tables and tables inside the lakehouse. I think that’s part of this., so I’m going to put the migration documentation in the chat window and then also I’m going to put in the other there’s a secondary article here talks about the transition as the learn documentation from Microsoft. There’s a little bit more detail around the technical pieces. It talks about the different drivers. It talks about connectors and the changing of drivers. So the idea is if you’re connecting to data bricks, if you’re doing Azure

4:07 to data bricks, if you’re doing Azure data bricks, if you’re connecting to Google BigQuery, Hive, Impala, Snowflake, Spark, these are the recommended switches that you’re you’re able to apply. So it’s not like I’m going back to SQL databases and I’m having to switch the driver on that. This is this is more of a big cloud technology piece. ADBC is connecting to specific data sources. So Tommy, you you said this to I’m like, where am I? I literally thought to myself, where the heck am I using OBBC on connections? Because I I know I’ve used them. I’ve selected them in various locations, but I think this is really

4:39 locations, but I think this is really dependent on there is sources of data that are using this new ADBC connector. And so if you’re connecting to data bricks, bricks, Dermo, Google, BigQuery, Hive, Impala, Snowflake, and Spark, this is where you need to pay attention with this. This is this is where by default it’s going to use this new connector which will probably be faster I’m assuming and it’s just the way of the world I think ODBC is its legacy right but however again the whole technology thing I guarantee you there organizations who rely on this and

5:09 organizations who rely on this and I’m glad there’s a scanner to help you understand where it actually lives because that would have been but it’s not going to it’s not a migration tool and I think that’s the difference it’s a scanner tool it doesn’t migrate it for you That’s on you to fix. Yes. And I think just very clear, it’s not like they’re deprecating or getting rid of OBC. That may still be used for other data sources, but for these data sources, that’s when you want to use it. Okay. Okay. Exactly. So, that’s the first one. The whole there for a minute.

5:40 The whole there for a minute. Speaking of things that don’t work, another update,, the co-pilot or PowerBI Q&A retirement reminder, the February 2027 timeline. all five of you who are trying who are still using PowerBI PowerBI Q mean so in case you did not hear for the five of you who need it PowerBI where Microsoft previously announced the legacy PowerBI Q&A natural language querying experience would retire in December 2026

6:12 retire in December 2026 it’s actually now been extended to February 2027 there you go because maybe there’s now seven they realized there weren’t five people there were seven there was seven people that need it And again, this affects two people or two types of people and user Q&A experiences and Q&A configuration setup tools and but they did such work on getting that Q&A experience the best they could before AI. Here’s where I’m curious of your thoughts, Mike, because they have a table

6:42 table that shows what is being deprecated, what is or what experience is being retired, and they have rather than just saying the new tool, the recommended alternative, and I’m not sure how I feel about this. So, for example, Q&A to report, use copilot, PowerBI reports, Q&A on dashboards, the standalone experience. So, that’s not a onetoone. the virtual analyst in mobile apps copile in PowerBI mobile Q&A embedded copile for SSAS Q&A setup prep data for a for AI so we’re really not we’re

7:14 a for AI so we’re really not we’re seeing seeing the importance of having the fabric copilot capacity can now that can be created on F2 and higher capacities thank goodness but you need to obviously enable that and it makes a co-pilot available in smaller capacity so Mike let me get your take on this. this., I didn’t find a ton of value in the Q&A before, so don’t really, there’s there are people that like to ask questions of their data specifically.

7:44 specifically., I think Copilot is a better solution than than the Q&A previously. So, if I if you want if you really need that feature, I think it makes sense to to migrate up to the new thing. Do I use a lot of Q&A on documents? I don’t think I do really. I don’t do a lot of that. And I think anymore people are not are not looking for a drop. Not not a lot. There are people that still want this. There’s people that request this interaction with their data. Not disagreeing with that, but I think anymore I feel like the sentiment of the

8:16 anymore I feel like the sentiment of the customers and the people that I work with is less around I need to drop in a QA experience. Rather, they’re saying how can I give Claude the data and the reports that we’re looking at. I I feel like it’s it’s we’re still asking questions of the data, but it’s more like I need to connect to the semantic model and talk to the semantic model directly rather than let me give you a report page and we’ll just happen to drop in something that’s on that page. So, I just feel like the the sentiment of the market has

8:48 like the the sentiment of the market has slightly shifted for me in my surface area of people I talk to and it feels like they’re trying we’re trying to do a lot more work around how do I give you access directly to the model with agents as opposed to I need a a text box that I can chat with inside the large the the report page. Right. So, I think there’s a difference there that I’m looking at, but I could be wrong on that. Yeah, I’m curious who I’m interacting with. Yeah, I’m curious about the cost difference here because when I would always do PowerB dashboard and data

9:18 always do PowerB dashboard and data training, the biggest one of the cells so to speak in the beginning for people who are new to PowerBI was you can get started with these AI features and then I had to change it to AI features pre- chat GPT or generative AI and that they just work and again it wasn’t anything too special but to build that visual show me sales by country or by region by region and building that visual did not cost you additional CUS to talk to it. Exactly. It cost you nothing. I could do on a desktop service. Didn’t matter about your license. You could have done

9:48 about your license. You could have done it on free. But again, now anytime you’re going to communicate, it’s going to cost you tokens. It’s going to cost you something. And this new experience is locked in to an FQ. I believe Q&A would work on pro and premium per user. Yes. Oh, they’re retiring it. So you need if you want to do any sense natural language you’re going to have to use the F2 with okay the FCC really that is I think maybe the the biggest so

10:18 that is I think maybe the the biggest so everything else I said earlier which like moot points interesting maybe okay fine Michael you’re weird and how you build things okay this point I think is the sticking point for me right the sticking point is we’re going from pro and premium per user to must have fabric capacity. Yeah, that’s I think the bigger barrier to entry for people., and maybe it’s a sell feature to try and push people into fabric more.,, I think a

10:49 into fabric more.,, I think a fabric has been adopting well, but I also think there’s a lot of organizations out there that still haven’t embraced fabric or putting or turning it on or applying it to different places in the reports. Here’s what the last bit I’ll say about this and then we’ll we’ll get to the fun. they mention narrative visuals or the nar narrative summaries and my question here is it they’re saying this for the embedded experience. So what they’re actually saying here is if you are using the embedded PASS Q&A scenarios

11:19 scenarios use narrative visuals and smart narrative summaries are those going those are I believe going away too. It’s really for most people not in a embedded PAS scenario. Those are going away as well. It’s not just the Q&A. It’s also like narrative summary and visuals, all those AI features. And I did like those a lot, especially if it was in a sense easy to do. So I think they need to do a little clarity on can I use narrative visuals or

11:50 on can I use narrative visuals or narrative summaries if I’m on a normal cloud or is that only in that scenario they’re talking about? Yeah, I think I think what they’re saying this part of the article as I read this one, it’s saying if you’re if you’re using embedded Q&A for things, you should consider co-pilot for SAS exper scenarios scenarios and you should consider the narrative visual or smart narrative summaries, which I believe those might be custom visuals that you can bring in. There’s some people or companies that you can buy a smart narrative visual from,

12:20 buy a smart narrative visual from, right? right? And those can be applied to your report. So that’s another alternative to not have to go down the fabric capacity route. route. And then the other one they also note here, this is what I think is happening, fabric IQ or fabric MCP server, that’s an actual MCP you talk to with your agents. So when you show up with Claude, when you show up with all the different AI systems that are out there, you can directly talk to the model that way. So, I would argue that’s probably the way I’m going to push

12:50 probably the way I’m going to push people more towards is is those systems because I’m finding so much value from having like a claude or a cursor or an XAI, something sitting in front of what I what I talk to basically, right? right? Okay, Okay, awesome. Good notes here., I just want to call her out. All the articles and topics of those announcements are in the chat window and they’re also in the title in the description of this video in case you want to check those out and go read up on these things yourself, just so you’re aware. All

13:20 yourself, just so you’re aware. All right, with that being said, let’s get into our main topic today. And in typical fashion here, I’d love to bring in our guest here. Kurt from Data Globins is back with us again., we have

13:33 again., we have Hello. We got feedback from our last episode on Tuesday. I saw a note on YouTube. It said best episode ever in exclamation points from last from last week. So Kurt, thank you so much for bringing up the quality of these episodes substantially to the point of these are best episodes ever. So for those of you who are joining in real time, you are going to hear the best conversation ever. Ever. This hopefully this is number two of this of the second one of these because Kurt’s here. He’s a great thinker. Kurt is

14:03 here. He’s a great thinker. Kurt is amazing in the PowerBI community and now data and AI community. I think you’re expanding just beyond PowerBI at this point, Kurt. and we we’re very happy to have you here and talk to us about these topics around data AI and unpacking what this means for you and your organization. Welcome. Thanks for having me. Looking forward to it. Lots to talk about as always. Lots to talk about. Whenever I hear someone say best something ever, I I I think it should be said in the way of the comic book guy from the Simpsons. Best episode ever. I

14:33 from the Simpsons. Best episode ever. I don’t know. So, we’re really excited. Yeah. So, let’s keep it up. We’ll keep up the energy. energy. We’ll keep going. Yeah. Let’s keep going here. So, let’s let maybe we should So, the topic for today is data and AI literacy like and how do you how self-aware are you of where you fit in this spectrum of your maturity level? so, maybe we should just touch on a couple initial topics. Maybe what does when we say the words the buzzwords we were Tommy was just this morning giving

15:03 were Tommy was just this morning giving us a whole bunch of buzzwords. Yeah. Yeah. What does data literacy mean to you and what does AI literacy mean to you? And we’ll start with our guest Kurt. Take it away. What what do these terms mean to you? you? Huge topic to unpack but also something that is like been deeply important to me for a very very long time even like before I worked in BI. But data literacy to me is all about how well you can read and interpret data in order to work with and make decisions on that data.

15:34 and make decisions on that data. So data literacy is something that’s very complex and it’s also very nuanced. it has a lot of different individualistic facets to it. because it’s it’s somewhat inherently cultural. And when I think about data literacy, like there’s a lot of different things that mix in with it. Like it’s a superset of other things like math and statistical literacy, digital and technological literacy, visual and general information literacy, as well as

16:05 general information literacy, as well as critical and analytical thinking skills. So a lot of these things are like bundled up into what data literacy is., and of course it’s very closely associated with other things that relate to how you as an organization culturally work with data. So your data culture, but also things like change management and AI literacy is not too far away from that. But I’ll park the bus there. But what do you guys think? I I like how you phrase that. I’ll I’ll jump in here quickly and talk I’ll kick

16:36 jump in here quickly and talk I’ll kick it to you here in a second., I like that phrasing. how well you can read and interpret data. I want to hang on the point of this kind I want to hang on the point of this varies by department and it varies by of varies by department and it varies by individuals within the department. I think different people step into this experience with what have you learned about data literacy. And also I would argue when you walk into an organization there’s not necessarily a hey welcome to Carlo consulting here’s our data lit data literacy course here’s how we

17:08 data literacy course here’s how we here’s how we communicate about data here’s the things you’re going to see and how you interpret those things here’s the language we use in and so I think there is when I think about data and data visualizations I’m communicating to you in graphical form form and so there’s there’s do I when I use the language of visuals and charts and graphs, are you able to understand that language and read it the same way I’m reading it? And so those are the things that I’m picking up. Tommy, what do you

17:38 that I’m picking up. Tommy, what do you think about data literacy? Yeah, I’m not going to try to disagree with you right off the bat, but let me add on to Kurt’s point first because I like what he said on the whole department thing is that’s where I’m like, but let me focus on what I agree with. The first thing is I agree is be able to read interpret data. But I would add on to that, this is actually from some of our previous episodes too where it’s also be able to question communicate the data and also one thing we talked about Mike in the past true data literacy is the ability to argue

18:08 data literacy is the ability to argue with data in the sense that having those conversations to make better decisions. It’s not. And I think data literacy as a consultant, Mike, your data literacy whenever you go to a new client is zero. And you have to build that up because you have to know in their data, not just how to read a bar chart, but what is their pain points and thresholds and what those numbers actually in a sense mean to them. So rather than going, I know you have 10 metrics. Well, some of

18:38 know you have 10 metrics. Well, some of those metrics are more important than others. So that’s where I would really say it’s not just reading and interpreting the data, but the communication and the ability to argue with that data to make better decisions. Yeah. No, definitely I agree with that as well. So the decision-m part is like obviously very key because you need to be able to use data in order to motivate decision-m like a sign of in my mind low data literacy is when decisions are m made based on tribal instinct

19:09 made based on tribal instinct or or for implicit behavioral reasons or patterns that don’t aren’t necessarily backed up by data and when confronted by data to Tommy’s point that people are not able to engage with a datadriven argument and that that’s not going to be a motivating factor in choosing something or or on the contrary that it’s just not certain on what metrics we should use in order to make a decision and why. So so yeah I definitely agree with that.

19:40 agree with that. I like where you took this one. Instead of maybe identifying high data literacy areas, also the the opposite is important, being able to identify low data literacy moments. So, so I think potentially there’s something there on that other end of the spectrum, which is and I like I thought I thought you had a really good point there. Low data is like those gut feeling decisions. We’re not leveraging data to make those decisions. And I and I think you’re right. I’ve been in those situations

20:10 right. I’ve been in those situations where I’ve confronted people with what the data actually shows and people feel a little bit sheepish sometime about being confronted by real data. yeah, yeah, that’s a bit of an awkward conversation for you sometimes when you’re actually using real information to back up your answers and someone else is showing up. And maybe I’m reading a little bit between the lines here. Kurt, you come from the PhD data. What would be is it biomedical sciences biomedical I was going to say genetics

20:41 biomedical I was going to say genetics but I think biomedical is better in that space I think there’s a like that there’s a lot of study of individuals in that arena that uses data numbers statistical analysis like you say the mean is something like this or the standard deviation is that everyone understands what you what like they understand those terms immediately but if Tommy and I step or other people in other business units step in started saying, “All right, well, the,, here’s our number and our standard deviation is X.” Tommy knows this because he does a lot of baseball stats

21:11 because he does a lot of baseball stats and he’s loves stats. That’s like Tommy. You do biomedics, though. You’re I’m going to go H2O. Yeah. So, but but regardless though, there’s a there’s a there’s a language of terms that we say. It’s the same buzzwords like when we’re talking about AI, Olama,, large language model, harness, there’s things that we know. And when I say that word, I need to make sure you understand the same. You can visualize the same thing we’re talking about when I say harness. Let’s do an analogy here because I think

21:41 Let’s do an analogy here because I think that think of the word just literate, right? There is where there’s a universal lit literacy when it comes to the English language, right? Reading a book. However, Mike, if I gave you one of my theology or philosophy books and you had no philosophy background, you would be illiterate to it. even if you understand the words that are on the page, you’re,, it’s the actual meaning of those words. And like it it circles back to what you said, Tommy, talking about communication because I think for me the bottom line on a lot of these like cultural

22:12 on a lot of these like cultural topics that relate to data culture and and how we behave like patterns of behavior that we have around data and BI. to me, the bottom line is is there empathy? Like is empathy a common value that you have? Because you you need to have it because when you go and engage with people and you go and talk to them and you immediately start using technical terms that you’re very comfortable with or vice versa, they start,, slinging the business lingo at you in ways that you can’t follow, you’re immediately not

22:43 can’t follow, you’re immediately not speaking the same language. And so, especially if you’re a consultant, your job is to be the bridge builder and to own that and to have empathy for that person and bring that empathy and to be able to facilitate the bridge building in the dialogue so that you can have a birectional conversation with each other about these topics that ultimately boil down to data. Ooh, I like the empathy piece of this. I think as an as that person who spans the business and the IT technical science, empathy is an understated value of

23:16 empathy is an understated value of being able to to to relate to them to hear like listen. We got two ears, we got one mouth. I run my mouth a lot and so I should probably be listening more. more. So I I have a question for you both here because I think when you think of value sometimes you also think of scarcity, right? there’s only so much of it and I think in the age of fabric right now and Curt you were mentioning something about this we can build data as quickly as possible now both with

23:46 as quickly as possible now both with from an angentic point of view but also just what fabric can do to have lakeous and anyone can do it right anyone can actually build a lakehouse right now that barrier is gone has fabric increased the an organization or a group of people’s data literacy or the ability for their to gain data literacy or is it doing the opposite here? here? Yeah. So, it’s a complicated question because fabric is just a tool

24:16 question because fabric is just a tool like it’s it it on its own is having no effect on that. But but in order for a team to be able to succeed to scale what they want to do out of PowerBI and into fabric obviously there are higher prerequisites in order to be able to understand certain things and we can already make the leap here into like talking about AI and agents but I do believe that successful use of AI and agents can lower that barrier to make it more accessible and to facilitate the

24:46 more accessible and to facilitate the growing of data literacy. I think that’s possible., but not in a vacuum. It needs to be done in a way that’s facilitated somehow. But,, but in general, that tends to be my experience. And the go-to example that I tend to give is direct lake. So, direct lake on paper like is really great, especially now with agents. Direct Lake, I advocate for it quite strongly because it’s one of the quickest ways to be able to query data in a lakehouse. You whip up the direct lake, you can query it, you’re good. and

25:16 lake, you can query it, you’re good. and it can be ephemeral., so you you build it up, you tear it down, it’s fine. fine. But,, but the problem is if you’re telling someone like in a decentralized business unit who are doing analytics and you’re telling them like, hey, you and you’re telling them like, hey,, let’s consider direct lake know, let’s consider direct lake you’re also implicitly telling them, okay, you’re going to have to manage this stuff in the lakehouse and you’re going to have to use notebooks and you’re gonna have to vacuum and stuff like this. And suddenly like there’s a lot more technical things to consider even when you start talking about like the various technical aspects

25:48 the various technical aspects of direct lake that are still scrap scrapping the surface scraping the surface. So and and that’s that’s an step up in terms of what you’re asking that person to understand on a technical level. So so I do believe that personally. What do you guys think? I so I had a comment I thought initial thought on fabric is a tool and I agree with you. I think fabric is one of many tools. It doesn’t have to be fabric. There can be many different tools that are applied here. here. not just fabric to do what we’re

26:20 not just fabric to do what we’re doing. doing. However, I think the ease of getting access to lakeouses and building items together, together, I feel like that is a function of this tool that was easier than what I was trying to build before in Azure. Right. So I So I Oh yeah, for sure. For sure. To me, there’s like this aspect of accessibility of the tool and how easy it is for us to create things. So, you it is for us to create things. So,, previously I would build a blob know, previously I would build a blob storage account. I would maybe need to go turn on some data bricks something or

26:51 go turn on some data bricks something or another or synapse, but these are all Azure based artifacts that lived inside Azure somewhere. With fabric, I don’t have to turn that stuff on. I just click a button, an item shows up, I hit create, and boom, it it runs. It’s there. So the availability or the

27:06 there. So the availability or the accessibility of the tool I think is is encouraging us to engage with the tool more frequently. And to your point, it just makes it more difficult or challenging for us as we talk about data literacy. What do people actually know in our company about to your point the data lake? how does that actually materialize into people’s day-to-day workflow? who’s managing it. there’s a with the great tool, we get a lot more

27:36 a with the great tool, we get a lot more responsibility. I would do agree there. So that that’s my maybe my my comment on that one. that one. I think I’d like to I think we’ve defined data literacy well enough. I’d like to maybe start unpacking well where does AI literacy fall? And then I think your comment that I really resonated with Kurt was you said AI can help improve or it’s an assistance to your data literacy and I wanted to that’s a really novel or ne neat concept you’re giving

28:07 novel or ne neat concept you’re giving there. So maybe we should just transition to maybe let’s talk about data or AI literacy. What does that look like? Is it the same thing? Are we talking about something different here? It’s to me it feels very different. So it’s like I’m going to segue into because you made me think of something that I have a super strong opinion about. okay but yeah so so for me AI literacy is indeed it’s similar in that the fact that you like you need to be aware of the thing that you’re using like for example that it can be non-deterministic that it’s also not

28:39 non-deterministic that it’s also not a person you’re not anthropomorphizing it you’re it you’re like aware of when you need to check things and when you shouldn’t and this thing what is an appropriate use and what is not. It’s not about stuff like your ability to prompt or something like this. It’s it’s more just about like your general awareness of what the technology is and what use cases it might fit and how you can reconcile that with your way of being able to conduct information but also to fulfill

29:11 conduct information but also to fulfill a task that you would normally do without AI. So like leaprogging a bit onto what you talked about how I I do think AI can help and agents can help facilitate learning and literacy about data but other topics like technological literacy in general because there is this prevailing I wouldn’t yet call it a dogma but like belief that AI deskills people and I think that that is true I think that that can happen and that is a risk that we need to mitigate like there is this

29:41 we need to mitigate like there is this belief that AI can lower critical thinking and that’s a bit of a catch 22 or whatever because we know that critical thinking is necessary to use AI and write and do critical thinking but at the same time like we’re finding that when people use these tools they’re also actively being trained to do those things less and not as well. However, I I speaking from my personal experience, but also what I’m observing in pockets of communities is

30:12 observing in pockets of communities is that there are a lot of people who are being taken by the hand by their use of coding agents to discover wide worlds far beyond where they used to live and work and things that they’re able to do. And I think that AI can lead them on this journey to your point by making it more accessible so that suddenly they can discover new things that they never would have dreamed of touching before. and it can really help them be able to understand those areas. And

30:43 to understand those areas. And that’s not to say that you’re going to learn Rust because your agent something in Rust. No thank you. Yeah. Yeah. But I might use it. No, but indeed but you can however focus on understanding why Rust, what is Rust good at? What is Rust not good at? Why would you use Rust as opposed to Python? What are the various things you need to be aware of? And like And like being led into that journey is something that is is really great and and so yeah, I’ll pause there. No, I think there’s a few branches here

31:13 No, I think there’s a few branches here because let’s let’s be clear that when we are talking about the data literacy, we were very much coming from the consumer point of view, right? The definitions that I heard from the three of you was all about consuming, communicating,, making decisions off of data, which doesn’t really sound like the data engineer, the BI professional. That’s most of the organization. And for the most part, I’m not saying that carb blanch. blanch. Well, I Well, okay. How would that definition then be applied specifically

31:44 definition then be applied specifically to a BI engineer? Well, what what is the bottom line of what you’re trying to do with data, right? Like yeah, you’re transforming that data, but the bottom line is it’s being communicated to the business in my mind. Like that is that is the beall end all is how can we take the the the business process and the business logic and represent it in data structures in an accurate way via data modeling and and whatever and then communicate that to the business in an effective way. And I think that is the engineering comes into play there. But that is ultimately

32:15 into play there. But that is ultimately to me still the bottom line and the goal. but maybe you disagree but no I have no qualms with that. I have no qualms with that. I think the only thing I would say is when I hear the words data literacy I very much primarily think about the importance of that for the organization or the consumer. Sure. Sure. Some somewhat we’re not saying that they’re not important and our but our definitions for AI literacy that we’ve said so far at least early on has been about the building side of this. And I think there’s two branches here. I made an argument in a podcast and I

32:47 I made an argument in a podcast and I wrote an article to m not I didn’t write the article to Mike but the conversation was simply that agents raise the floor and will lower the ceiling. ceiling. They raise your floor in terms of what you can do but there’s a there’s concern that they’re going to lower your actual ceiling, ceiling, right? right? Because what you said. Yeah. Yeah. So definitely for sure. And so I I mentioned this when I wrote about co-pilot at some point like I think it was last was last that was a great article. Yeah., but I I talk about like unknown unknowns and

33:17 I I talk about like unknown unknowns and how for especially people who are inexperienced or who are early in their careers and therefore inexperienced like that is one of the biggest risks is that they venture into these new areas and they’re able to do a lot of things. They’re able to put their hand deeper into the fire., but they’re not they have a bigger pool of those unknown unknowns. And that is where your experience and maturity in your career helps you out to be able to use these tools is you’re more aware of coming to self-awareness. You’re more aware of what your unknown unknowns are and how

33:49 what your unknown unknowns are and how you deal with that. And you can kind you deal with that. And you can close that gap. And to me when you of close that gap. And to me when you talk about like raising the ceiling, that’s what I think about a little bit. Are the unknown unknowns in that concern the same for the two types of people? I say the people building or the people consuming because again if I Yeah. Yeah. So it’s it’s a very good point and to me it’s like so building is a bit a superset of consuming because you need to consume in order to build. So but

34:19 to consume in order to build. So but indeed like building is a wider blast radius but and but consuming is still you still need that right like you still need to be aware of these things when for example you’re doing conversational BI and you’re having a you’re interfacing in a chat window with the agents to be able to answer data questions or to be able to do research or think through a process hopefully alongside other traditional tools like you still need to have the AI literacy in the same way to be aware of what what it can and can’t do and again

34:51 what what it can and can’t do and again not anthropomorphizing it and making sure that you understand that it it it can be semi-random or non-deterministic these things. so but just the consequences of not knowing that when you’re building something can be propagated out to the deterministic beast that you’re going to release upon the world. Yep. or or semi not deterministic. I Yep. or or semi not deterministic. so a lot of what we build with mean so a lot of what we build with right if you’re using an agent in

35:22 right if you’re using an agent in something that you’re building like the agent is part of the chat experience or something there is a there’s an aspect of does the agent even respond with the answers you would want to be in that product that you build. Let’s say you’re building a website and you want to add a chatbot or you want to ask have the chatbot ask questions. How are you grounding that agent in that information? So,, I’m going to I’m I’m going to reinforce your point here slightly, Kurt, because you talked about the impact of generative AI and critical thinking. Microsoft wrote a paper on

35:52 thinking. Microsoft wrote a paper on this in I think it was January of 2025, so early earlier than where we’re at now because I think for me the end of 2025, this,, November, December, things really started ramping up a notch and we started getting a lot more of like you can do it. agents became much more capable. But regardless, this paper talks about what this generative AI piece is doing. And I can’t tell you the number of times I’ve been in conversations now where I’m having other

36:22 conversations now where I’m having other individuals really comment on I’m letting people use AI in our company and no one’s even reading the output of the AI for humanizing the words and then sending emails. And so you’re getting you’re getting content from people that are that is solely AI generated and never human reviewed. Oh boy. Oh boy. And I think that that feels sloppy to me to

36:54 sending BBX an email. Yeah. So so this is an interesting thing and like I’m in the middle of almost I’m almost finished writing an article that touches on this but when we think about No, it’s it’s no it’s all good. have it out by next Tuesday so we can talk. So, so, so when you guys when we think about AI slop, right? Like AI slop is like,, whatever it’s like one of the words of the year or whatever to you, what does AI slop mean when when someone says AI slop? How would you define AI slop like to your family at

37:24 define AI slop like to your family at the kitchen table? Great question, Tommy. You want to take that one first? I’ll I’ll come back to it next. the most direct way is anything without thought but it’s gen it’s generally any output that may be seen that is clearly written by AI with no purpose or meaning. I that would at least be I know that’s not the right definition but if you were to say how I would describe it to my family it is very much just a general prompt and it comes from a general prompt with a general output.

37:55 prompt with a general output. Yeah., so I’m gonna I’m gonna maybe vary your your answer a little bit, Tommy. I’m going to go. How do you identify AI slop? Now, how would you define it? Yeah, depends on the medium of where it comes from. So, I’m at the dinner table. I’m trying to tell my my family about AI slop. I’ll say AI has been trained on lots of patterns of how it communicates things. Different AIs communicate in similar patterns. there are if you’re in the AI space and you’re reading and working with it frequently, you start identifying common

38:26 frequently, you start identifying common phrases. a lot of times the AI will say certain words or phrasings that it uses over and over again. however like they like however it’s comprehensible that you may have and then it gives this really long run-on sentence thing or there’s these like common phrases. So I think when you start seeing some of that over time, if you’re working with AI and you’re reading its output regularly, you start seeing patterns in these regular

38:56 seeing patterns in these regular phrases. And one of the ones I I feel like I see a lot of the times right now is when I correct the AI, there’s the oh you’re right. I’m sorry. That is what was going on. Like so there’s always this like need for it to appease you. So there’s like common. So I would say what you’re seeing is when you see AI slop as you become more familiar with it, you’re going to see these things called EM dashes. You’re going to see the same pattern of thought and structure in information that’s being passed on. Yeah. Yeah. And that’s not necessarily how I talk.

39:26 And that’s not necessarily how I talk. Everyone has their own spin. I think there’s an uniqueness here. Everyone has learned how to speak and type and write and and pervade and convey thought differently. M so when I talk to Tommy Tommy you’re going to write things Tommy the way Tommy writes things because that’s how you’ve experienced them uniquely as Tommy AI is not doing that AI is the exact opposite it’s taking the aggregation of every thought and trying to find the most common pattern in whatever words are being said ne so it’s

39:57 whatever words are being said ne so it’s literally doing I say this word what’s the likelihood of the next word being this what’s the likelihood of the next word being this so it’s it’s doing the exact opposite. It’s summarizing every piece of word as opposed to what we’re doing, which is our unique experiences of words. So, it it’s a a feeling, but you see regular repeating patterns of this. What do you think about that definition, Kurt? Am I am I off base here? Is that what you’re seeing too? So, so yeah, I think like with writing definitely there’s there’s

40:27 with writing definitely there’s there’s that like you you have these shibiliths or like like repeated words or phrases or even symbols that come to occur. Like a common one that’s a lot more subtle is even like the little dot

40:38 more subtle is even like the little dot instead of a dash to separate something like when you see that or like uni code characters that are usually very difficult to type and you see them in everybody’s presentation now. And how did you get there? You use the alt alt key. Wow, I didn’t know you have a fancy keyboard. Yeah, indeed. So, but to me like slop ultimately because I’ve spent a lot of time thinking about this because I’ve I used to have really really strong opinions and I still do to some extent. I used to have extremely strong opinions about what can and should be AI generated in particular when it comes

41:09 generated in particular when it comes to art and writing and creative expression. But I’ve encountered a lot of people who for various reasons pertaining to accessibility or pertaining to other things are using AI to facilitate the output. So like when I think about AI slop to me like to Tommy’s point it is a representation of a lack of effort but to your point Mike it’s a lack of individuality. So when you take something that is ultimately meant to be directly or indirectly some act of creative or

41:39 act of creative or intellectual expression and it can be part of work but it can also not be part of work of work then and it doesn’t at all represent either effort nor individuality then to me that’s where it can be slop. And there can be things like these tells that just feel loud now and when you read it or you see it like you’re like oof. Like or also those like design like notches on the left side of things. You see that and you’re just like h no one’s ever done that ever. Yeah. So

42:09 Yeah. So So I I asked my bot what some of the So I couldn’t pull the phrases off the top of my head. I’m going to say some of these phrases and I want you guys to tell I I want you I’m going to say the phrase and I’m going to do are these what you hear? Like if I say these are these are these the triggers of AI slap to you. You’re not allowed to say more than three. three. Okay. I’ll I’ll do just three then for now. And maybe maybe I’ll sneak a couple in later. I’ll actually use the I’m going to use the phrases as I talk to you throughout the rest of this conversation now. Okay. So it is

42:40 conversation now. Okay. So it is important to note that. So that’s one of the phrases in the ever evolving landscape of and then another one there’s a lot of buzz intensifiers robust seamless cutting edge leverage unlock and then there’s a lot of like that’s a fantastic point or you’re absolutely right these are phrases I feel like I’ve seen that a lot of time it’s why yeah exactly it’s why yeah so

43:11 it’s why yeah exactly it’s why yeah so anyways Okay. So, but it’s it’s the the bottom line though is I think that human attention deserves human effort, right? Like in my mind. So, it I think I’ve still seen I’ve seen many cases of someone using AI to where what they’re what they made was largely almost completely AI composed. but it’s still something that they spent substantial time and effort to be able to construct and it does represent some idea or vision that they

43:42 some idea or vision that they had that that is itself original and you can see that or or to some extent like they for accessibility reasons like they they for example through dyslexia or other reasons like they are using AI to help facilitate the communication of an idea like through written pros. and so so I have seen cases where it’s like it it’s it was fully AI generated but it but it wasn’t sloped. So now connecting this to our point right AI literacy in the organization is

44:15 right AI literacy in the organization is when you start to see this slop right like then that that becomes almost like a symptom of a sickness right and it can it can spread because also when people get like like Tommy was saying you get these emails that are clearly AI generated or comments or whatever that are are AI generated like it’s tempting to then just like throw your agent back at it or whatever., or same thing like when someone creates something like with a semantic

44:45 creates something like with a semantic model and then you ask them like why did you do this and they’re just like I don’t know the agent did it. They don’t even bother to explain. Oh, I can’t stand that. Yeah. And that that that’s indicative of in my mind challenges with AI literacy and not knowing where that line is of okay you don’t need to know the exact specific place where the filter context the context transition happens or whatever but you do need to explain why you made this decision instead of that decision like you have currency conversion in your model you

45:16 currency conversion in your model you should know that and you chose this pattern and not that pattern. So why? Well, and it’s also not just the the building side of it too, but it’s also I think the workflow that you create for yourself, right? So yes, I Mike’s been saying this for a while and I don’t know if I should I gave you credit in the article, but building the thing that builds the thing is one of my favorite phrases. So, I just wrote an article about that and but again, yeah, but but to your

45:46 and but again, yeah, but but to your point, by the way, this whole conversation, I’m very I have to use a not have to, but AI is a big tool for me to help write because I talk. My wife jokes around that 90% of the day I’m on the phone, but I’m really just having conversations because with ADD, I talk a lot faster than I can write or think a lot faster than I can do things. add that with skills that I’ve created for myself with transcripts for the podcast. It I try to be intentional with that. All that to being said is I think especially for the

46:17 being said is I think especially for the BI developer it’s not just hey update this semantic model it’s broken to me that’s a prompt slop prompt right or because I think to re the part of the literacy is not just how you write the prompt as we said but it’s also understanding I think that workflow so I’m very intentional about how instructions are built and what context is provided did the other side of the coin and I wanted to just get your idea

46:48 coin and I wanted to just get your idea with this is I don’t think you can get higher data literacy without a higher data culture. I don’t know they I think they’re very intertwined, right? Yeah. Your literacy is going to dictate your culture and I think vice versa. This is in and maybe we can extend that conversation both to not just data Tommy but I think also to AI that’s exactly where we’re going with that. Yeah, they both go hand in hand and I think sometimes we maybe observe

47:19 sometimes we maybe observe and this could be just me. I believe there are observations where we push on we like to use these tools. We want to use BI. We want to use data. We want to use AI. It helps us. It can make certain things of our our jobs easier and faster to get accomplished. But the more you push into them, sometimes the organization resists and not all individuals have the same understanding of these literacy pieces. And so when you have an outlier,

47:51 you have an outlier, a Tommy show up that has or Kurt show up has a high level of literacy comparatively to the rest of the audience. audience. It creates friction between that employee and the organization. And I’ve even seen sometimes maybe this is a a hot take a little bit when we are using when we see someone who’s out the outlier in the curve. either the organization takes on the initiative that like we’re going to raise the bar for everyone to pull the literacy level of the entire organization up or that person might

48:23 organization up or that person might need to start thinking about I maybe need to go find another organization that has the same level of literacy that I do I do because I think people get frustrated when you are at a different literacy level than others for a long period of time time and it gets challenging. So sorry a lot of time pack there thoughts. Yeah. So, so this reminds me a lot of like this concept from like the data data ops cookbook I think so is the idea of the data hero and this is something that when you you end up

48:54 something that when you you end up having one person who has a high initiative and also a certain amount of capabilities and it could be related to data literacy but it also just could be in relation to just what needs to be done but they they end up becoming the hero that has to like shoulder the burden of a lot of datadriven processes. but it’s an unsustainable situation because if they’re out suddenly like the the reports and dashboards aren’t being delivered and everything breaks down and

49:24 delivered and everything breaks down and no one knows how it works and this kind no one knows how it works and this thing. So I think in those of thing. So I think in those circumstances those people can also become data heroes rather than the organization seeing like okay how can we first of all like deal with the capacity constraints but also to to to scatter the knowledge more amongst different people so that you amongst different people so that in our organization we have an know in our organization we have an elevated amount and I think with AI this becomes especially problematic because

49:55 becomes especially problematic because you have this idea now of individual contributors or super individual contributors. yes yes and so suddenly you the narrative of like one person can do everything because they’re orchestrating they’re orchestrating agents. Yes. Yes. Yes. Yes. Let me Yeah. Let me go ahead. Let me tie this back to our conversation on Tuesday. This was one of the things that I brought to YouTube was if I have all these people building fabric apps because I can all I have to do is a

50:25 because I can all I have to do is a prompt and an mpm and have VS code but we are all building our in very individualistic apps right that do not follow any continuality. You are going to have this ridiculous schema and and definitions and way people are from a user interface point of view to the process to what data is actually getting intaked. Unless there’s a process, right? It’s like if I just told the three of us to go to an organization and you’re going to work with these 10 people, you’re going to work with these 10 people, you’re going to work with these 10,

50:56 you’re going to work with these 10, build them apps because you can. It’s super easy. You don’t need any skill in that area. Well, if we’re not talking to each other, what’s going to happen is you’re going to in a sense, I don’t want to say create three different cultures, but there are going to be three different ways those apps work, right? because I’m going to use my own prompting style, my own skills, unless that’s shared. We already had this problem in the in the business intelligence world where if I was building a report without being part of a common language in my BI team or with

51:27 a common language in my BI team or with the the people who were building reports, everything became diverged and all these tribes tribal ways of creating reports were being done. I think to your point Kurt that’s just going to get exacerbated now not just with fabric apps but if the way I can build things and I think that’s part of culture because if you define culture it’s a set of traditions process language that goes into it into it and I think that’s never been more important

51:57 important because it’s so easy to do right yeah defin want to keep touching on your point here Kirk because I think Tommy great point on that one I agree with I think the individ so I want to I want to pick on a point from Rob Collie’s book his first his book a long time ago he talked a lot about if I gave you an application to build and I said I already know how to write every line of code inside that application and I just give you okay your go your job is to take from this

52:28 your go your job is to take from this written out piece of paper and type in every line of code to make the application work you can be done you could just type literally type the code in like days. It’s it’s it’s a lot of typing, but you could type all the code out and it would work, right? That kind out and it would work, right? That aspect. But when you work on a of aspect. But when you work on a project or even particularly a data project, why does a data project take weeks, months, years to get done? Yeah. Yeah. Because there’s all this connection and coordination happening between other individuals, right? It’s the you’re not

52:59 individuals, right? It’s the you’re not an individual contributor that can land the entire project by yourself. No. So No. So yeah, yeah, I think this is potentially part of the aspect you’re you’re touching on Kurt here is,, AI is almost encouraging silos to some degree because now everyone and I would even maybe argue yeah more people more people in the organization have the ability of creating more complex bigger software pieces full solutions without any

53:31 pieces full solutions without any involvement for anyone else. And I’m I had a thought after our conversation on Tuesday which was in the in the same way PowerBI created a proliferation problem for reports. Everyone can create a report now. It’s here’s a model drag and drop things in every you get sometimes you get models with great DAXs. Sometimes you get models with trash DAXs. You get sometimes good models. Sometimes you get models with very horrible designed models. models. trash. trash. Trash Daxs.

54:01 Trash Daxs. That’s a t-shirt. Okay, that’s a that’s a shirt right there. I’m making trash DAXs. So, but there’s now because the

54:10 So, but there’s now because the ability to create is now much easier. You get a a wider variety of what people know. And going back to your point earlier, Kurt, which was novice or new users to the space, they don’t know the unknowns. I didn’t know that the way you shape these data model tables from dims and fax was wrong or you’re over you’re you’re literally pulling in all the tables as they came from your operational system but that doesn’t work in a semantic model. It’s not as efficient. Your DAX gets incredibly complex because your model your data

54:40 complex because your model your data hasn’t been shaped correctly. You learn that over time, but that’s one of those unknown unknowns new users stumble into. And now with AI, you get it not just for data, you get it for everything like app building, email writing, like you now have this individualistic contributor piece of a lot more people with a lot more capability and maybe they don’t know the unknowns. I Whoa. this is a huge concept that you’re proposing here, but it’s it def it’s definitely true and to some extent like vibe coding and

55:11 to some extent like vibe coding and stuff is just self-service BI on steroids like and and we’re not just seeing this happen inside of like self-service BI tools like I’m also encountering people who are for example like they’re sharing and creating and sharing dashboards in their cloud enterprise as artifacts and they’re preferring to use those over like the the PowerBI or Tableau dashboards or whatever. Yeah. Yeah. stuff like this PowerPoint, which I don’t know if you saw the latest anthropic update on Yeah.

55:42 saw the latest anthropic update on Yeah. Which incredible. Yeah. Yeah. Yeah. Yeah., I’ll be happy all day long. That the more you can kill PowerPoint, the happier I’ll be. So, I haven’t used PowerPoint in like a year or something. Like I I I haven’t used PowerPoint in a really long time. like only well I I do use it if I have to for specific reasons but when I have to give a presentation or a workshop or something like it’s all just basically HTML and like a custom tool that I’ve been slowly building over time but cool I like that I want to I want to even I want to double down on your point

56:12 even I want to double down on your point Kurt there is an article that says the unreasonable effectiveness of HTML have you seen this article on X I’m not sure if you have okay okay I I HTML is Crazy cool now because I don’t need to write it. I can produce it. And something I discovered just yesterday. If you’re in Teams and you add an HTML document into Teams like a like a markdown file or you just like put it there, you can click the file, it will render the HTML in Teams as you wrote

56:42 render the HTML in Teams as you wrote it. So if you have doc just just showed up one day for me. So I love writing documents in HTML. And Kurt, I’ve also had the same experience. People are now using cloud code and cloud code is generating HTML artifacts. It’s saving it on the cloud servers and then here’s a link. I’m sharing this document with you and it’s just an HTML document that I’m getting a link to Tommy or whoever and you open it and boom, it runs like a web page. You’re like, “Oh, this is awesome.” You can like read through it. It is very effective on communicating.

57:12 It is very effective on communicating., and I this is this is the new world that we live in. Do we even need Word? Do we even need PowerPoint? Should I be even using Excel anymore or should I be using something that’s more homegr Like do do you just build the things that you again this is to me this is like hyper customization on a lot of these things now like we’re opening a huge can of worms with this I think yeah yeah for sure it’s yeah there’s

57:42 yeah for sure it’s yeah there’s there’s there are a lot of things that that you discover though like to our point of la what we talked about on Tuesday also though when you try to like create your own version of these tools and stuff like this. Like you come to appreciate like the maturity and robustness that’s been built up by professionals over the year and you realize like you realize like the how naive you are to think that you could sling stones at the giant. But yeah agreed. But but however again like connecting it to the idea of like the super individual contributors

58:12 like the super individual contributors and data heroes and stuff that we now live in like there are many people in organizations that do think like well I can I can just do all this stuff and so then they start producing like massive amounts of output too much for any any single person to keep up on. And when people are generating any kind And when people are generating any BI artifact or HTML artifact like the of BI artifact or HTML artifact like the odds of them reading the code a lot of the times is is really quite low. Like and especially with something like fabric apps like I can tell you

58:43 fabric apps like I can tell you almost nobody I think who’s making a fabric app is actually looking at the code of that fabric app. I’m pretty sure. and same thing when people are yeah that that can be that can be but then when you have a whole bunch of you then when you have a whole bunch of like for example like metrics that know like for example like metrics that are being defined in the actual DAX queries or something or instead of in the semantic model the measures yeah I’ve seen that yes instead of doing a sum or it’s literally writing its own sum calculates inside it’s like it’s almost like a thin

59:14 inside it’s like it’s almost like a thin measure that lives inside the fabric app now that where do you track that It’s not a is it really real? Is it really coming from the right data set? And and I even had like I got in an argument with someone who was talking about fabric apps. We were talking about I was saying like you should really try D3 and use D3. And he’s like no I want to use Vega. I’m going to use Vega. And I’m like okay so you’re using Vega and he’s like yeah I’m using Vega. And I’m like we were looking at their dashboard and I was like is this Vega? I was like, “Do you do like have

59:44 was like, “Do you do like have you looked at it like and and it it was they didn’t even know like they just assumed it was Vega and it was it wasn’t Vega.” Vega.” So it was like a lot of the times people like don’t even know but in that case it doesn’t necessarily matter. It was a pedantic discussion like but but and and for some of those decisions it doesn’t but it just goes to show like again coming to AI literacy and slop like you do need to be aware about what decisions are being made and you do need to be aware of

60:14 made and you do need to be aware of something like if you’re creating that dashboard is there a metric that is defined in the DAX query of the fabric app app and if so should you find a way to get that in the semantic And that’s something that you could say like you have to read that but to to to your guys point like you’re talking about the thing that builds the thing. There’s ways for you to architect this in your workflow where you you you have the again with AI literacy you have the maturity to realize that this is a

60:45 maturity to realize that this is a challenge that can come up because like Mike you say I’ve seen it and then so then you’re like okay well let’s let’s learn from that and let’s build something into our AI workflow that will recognize if that happens and we can deal with that in a semi-automated way or or we can prevent that from happening as we’re developing this thing. All right, it’s it’s time to pull out the big guns. So, all this is we already talked about.

61:15 we already talked about. Have I introduced you to Raspberry Pie? No. so, yeah. Yeah. No, hold the horses here. Mike, this is one of my favorite little bits to do and I’m gonna I’m gonna ask the room here and I’ll start so you have some time to think. But all these things we’re talking about are well and great. Awesome. phenomenal in terms of I can do these things with fabric apps and we can build this way. However, again when we go back to AI literacy and we go back to data literacy and the culture

61:45 back to data literacy and the culture side of this one of those things that that all works is on trust. So I’m going to propose to you. I’m going to propose to the three of you that you’re at an organization, a large organization, and Mike, what I like to use? You are the data zar. I call it the data zar. You basically have all control of all the budget and whatever you want to do can be done. We’ll call this for today the data and aisar. You are the data and aisar at the company. Your goals for the year or for however long is to do rapid pace development,

62:18 is to do rapid pace development, improve the data culture and trust, but also promote individuals. Everything that we’ve talking about that individualistic idea of being able to create, but you want to make sure things are validated. And again, from a high level, tell me the things that you put in place. Where would you start to get to that point? Because again, this all well work this all works well for the individual that we’re talking about. Yeah. Yeah. But once you start putting that with a group people, that’s where that

62:48 group people, that’s where that confusion happens. So, So, how do you do this, Tommy? How do you do this? You going to start with your answer so you guys can think about it. I don’t I’ve done that. Okay. Go ahead. So, I’ve done this to Mike too many times. I I open question and I go. So, all right. Go ahead, Kurt. The lowest the the the best way that you can start to do this is if it’s ultimately going to be like clearly like a cultural issue

63:18 clearly like a cultural issue especially when it comes to things like improving data literacy or improving people’s competence or whatever like you have adoption problems with a know you have adoption problems with a certain tool or whatever you you got to start building bridges and one of the easiest ways to do that is to facilitate some regular touch point. So to have something like a data cafe or office hours where people are welcome to attend. It’s not obligatory. it can be a remote call where people can dial in, but ideally it’s in person because there are certain behavioral cues that

63:49 there are certain behavioral cues that you can’t substitute for. and people people can’t evacuate as easily. But but there are li there are limitations. There are limitations. And if you’re in a big organization, chances are you will have to do it remote. It’s also more convenient., so and you you just make it very a very open and welcoming space where no question is ever wrong and you just start explaining things from the ground up like walking them through the PowerBI app and where they can find different things and you make it

64:19 different things and you make it targeted. It’s not like organizationwide like it’s targeting a specific community where you’re trying to promote that adoption and you let people bring their own questions and you make sure that that’s acknowledged or bring their feedback and you also credit them. So if someone one week has an idea for some change and it’s a good idea and you make sure that you implement that and then next week you’re like this was so and so’s idea it was a good idea and we implemented it and look this is what it looks like this is how it works. and then you you

64:50 is how it works. and then you you keep the momentum and the consistency going until it becomes baked into the culture a little bit. So that’s that’s how I would start. So and that’s something that I think is can be valuable. valuable. I’m going to agree with this one and I want to take flavors or notes of what’s coming from the fabric adoption roadmap. I know Kurt, you’re heavily involved with that and I I expected your answer to align with the the fabric adoption because you’ve spent a lot of time re processing and thinking through that

65:20 processing and thinking through that data. There’s multiple aspects of what data looks like in your organization. So I I think there’s notes of that same rigor that goes along with this, right? I really do think the the organization needs to come up with a position on usage of AI. Yeah. Yeah. Someone at the top needs to say directionally this is where we go. We acknowledge that there needs to be time spent. People need to spend

65:51 time spent. People need to spend resources on this. And I do think there’s a part of this that not everyone in the So if you if you basically gave every something like this to the whole organization and said, “Oh, everyone in the organization start using AI.” Yeah. Yeah. That doesn’t help. No. No. And there are many rabbit trails people will go down and then they’ll learn that that was a bad idea. So you you basically scatter potentially the efforts of the organization and everyone goes off to building these little siloed

66:22 goes off to building these little siloed things and we just get more proliferation of oneoff siloed non-coordinated efforts in the organization to not collectively move us the right direction and so we we too many and there’s there’s a point of too many side quests where you’re potentially distracting from the main direction. So I think leadership needs to allow so I think this is you need to have clear communication from leadership what you’re doing what expectations look like

66:52 you’re doing what expectations look like you need to give ample time I love these regular touch points I love communities it’s a it’s a center of excellence for data and AI extend that if you already have a data center of excellence start extending that into AI adding that into this I think makes a lot of sense because AI does really well when it has a lot of data next to, and then I I also think there’s almost a almost a I want to say there’s like a black ops team of this a little bit., you don’t know what you don’t know. And you you do

67:24 know what you don’t know. And you you do need to have some people take tangents and random thoughts, but you but in that process, you need to come back with, hey, I learned this lesson. Hey, I built a fabric app and it was totally using this D3. js JS stuff, but I was actually asking it to use plotly and it broke here and I burned a lot of tokens doing this thing and it didn’t

67:43 tokens doing this thing and it didn’t really work for me. I think a lot of this stuff is so new that we need a little bit of experimentation time with some experts of your organization or people that can handle this and have them come back with findings and reports and practices that work well for your organization. And from those lessons, then you start disseminating that knowledge into the rest of the team. Like, hey, when you prompting about a fabric app, we need a process to check it. There’s a governance piece that has to get done. And so, someone needs to build one and

68:13 And so, someone needs to build one and then think about what does the evaluation or another agent that has a skill that comes back and evals the fabric app to say, here’s things that our company says is is not acceptable. We’re not going to do with apps. Someone has to make that decision. so much to think about that. You just can’t wide open everything. So, let me pause there. I’m saying a lot of words. Just jumping in quickly, but it’s it’s about the executive sponsorship and I think you’re right. And it’s about that person. It’s the same like with the PowerBI,, in the fabric

68:43 PowerBI,, in the fabric adoption roadmap, like they need to lead by example. So, you can’t have an executive that’s just like slinging slop everywhere and posting on LinkedIn to something that’s fully AI generated, that’s even complaining about other people’s AI generated stuff. Yeah. And then having having like it needs to be demonstrated. So because people see that and then it it just it signals the wrong thing like times 10%. Whereas if you have an executive who’s

69:13 Whereas if you have an executive who’s communicating carefully and is acknowledging both the benefits and the risks and the downsides and saying like look this is what we value, this is what we want to do, this is what we don’t want to do. then that’s that can really motivate people and there’s nothing more motivating from what I’ve seen like in the last year or so than someone in a position of relative authority standing up and showing look at this cool thing I made with AI this is my like in really humbly

69:43 with AI this is my like in really humbly and simply like walking them through the process and not evangelizing it as the way and acknowledging the stuff that’s bad and what they learned and then just stepping down and saying okay this is what I did this is what I learned. And people just look at that and they’re like, “Wow, I should do that.” So, and like that is something that right now is for these organizations that want,, effective AI adoption or whatever. Like that is huge. That is huge. So, So, all right. I I love the Tiger team and I

70:15 all right. I I love the Tiger team and I love what you just said, which goes on to my Yeah. Yeah. It has a real name. I keep calling it black ops or like the super stealth team or like I keep making blacks. It’s like it’s like the tiger team. It’s like the focus team of effort like your focus is to figure out how to use this stuff effectively for our company. Yeah. I sorry I keep using the wrong term. No, and there’s there’s a few elements here. I just want to touch on what you said, but the phase approach I would have would be to start with the proof of concept stage where I would just want

70:45 concept stage where I would just want people to build that not is or things are not necessarily going to be in production right away, but just let’s build what’s possible, right? So whether you’re an executive or a leader, you want to put your money where your mouth is. You want to see AI then show that you can demonstrate a proof of concept even if it’s just a design or working application. Again, that’s not within your limits or that’s not that’s not past your limits with what we can do. The second would I I second phase would be I would be building in my

71:16 would be I would be building in my organization a agentic team. Their role is to build skills and agents for the organization. And when I say skills, I’m saying there’s going to be a shared directory for their skills and for the agents. Whether you’re for marketing, whether you’re for sales in the same fashion that BI worked with the company where what do you need? Where’s your data? What report are you looking for? Right? So like, hey, what are the common processes that you do? Oh, well, you processes that you do? Oh, well,, all the time we’re doing these know, all the time we’re doing these these flyers. Okay. Maybe there’s a

71:47 these flyers. Okay. Maybe there’s a skill there that you can use. We’ll follow process. It’s like data discovery, but then for AI assets. Exactly. Exactly. And these are dedicated team. From there I would actually have and again not this too much thought out but I would love to implement what I would call levels where in the same my brother-in-law he is a legit he is in logistics and there’s like level he’s like logistics level three or level 17 each level is a different pay salary but also

72:17 level is a different pay salary but also like the different role he has. I would love to put people into different levels in terms of what they have access to what they can do. clearance levels. Yeah. So once you demonstrate the red door Exactly. Exactly. Once you’ve demonstrated that you can accurately build an an application or something that can be used that can be trustworthy. We’re going to give you access to these agents or we’re going to give you access to the MCP server., I I think you’re right. Like, and I know that it’s something that a lot of people are

72:47 something that a lot of people are uncomfortable with, like putting other people in boxes and stuff like this, but the simple truth is that there there is to your point, like there’s a certain amount of trust that has to be provided to work with these things that are extremely dangerous. And even if it sets up all of these like restrictions and boxes and stuff, doesn’t matter. I’m sorry. It doesn’t matter. like you can do it. How many doors you lock, the windows are getting broken. So there you have to make sure that the right person is in the chair to not want to or

73:19 is in the chair to not want to or prevent windows from breaking. And I think that having some like,, segmentation in place as well as an understanding of how people can move up that ladder. I think that that is important. So, is there are you touching on maybe the shadow AI area? There’s shadows and there’s everything’s in everything’s lurking in the shadows potentially here a little bit as well. So, how many times I’ve had conversations with people like my

73:50 conversations with people like my friends or like random people and they’re like, “Oh, we’re just using Copilot.” And I’m like, “Okay, that sucks, but are you like you haven’t tried Claude or anything else or whatever?” And they’re like, “No, no, we don’t use that.” And then I’m like, “Okay.” And then we meet again like two or three weeks later and they’re like, “Actually, I found out that a lot of people are just using their own subscriptions and like stuff like like this is so pervasive.” And then they like there’s this frantic like,, we got to shut it down and especially because it’s a huge compliance risk. People shouldn’t do that.

74:20 do that. Huge. Nope. Should not, right? But I also agree like this is the same thing of like what we saw with like PowerBI and giving people like people were like locking down everything PowerBI. Well, then everything all the work just moves over to analyzing Excel and and and drop it out to Excel and do everything offline. So, there is this idea of like if you don’t give people the tools that they’re seeing and hearing other people use, you’re getting left behind and you’re clamping your team’s hands away from actually being effective with these other tools. Not saying you just

74:50 other tools. Not saying you just open it up wide wide open, right? And I love the gating. I I felt like I was as you were saying, red key, red door. I really felt like there was a Dungeons and Dragons like analogy showing up here that was gonna here,, around you don’t you don’t give a level,, 36 sword to a level two user because they they just can’t handle it. You actually have to earn a couple levels before you can wield the more powerful weapons. So, I I feel like this is

75:20 weapons. So, I I feel like this is another The AI is potentially a very powerful weapon we’re handing you and we need you to level up enough and will gate you into some of these things that move you on to these other levels. All right. right. It is though. It is like it is it is like this really powerful thing and you need to be aware because the in in organizations where they’re like, “Oh, yeah. We have,, we let people use cloud code.” And I’m like, “Okay, so like but do you have like limits and stuff like this or whatever?” like you stuff like this or whatever?” like we’re thinking of doing clock code know we’re thinking of doing clock code and it’s it’s that stuff can

75:50 and it’s it’s that stuff can get really dangerous and expensive really quickly but at the same time like you you also don’t want to overdo it right like it’s it’s exactly like PowerBI where you have these organizations that are so hyper restrictive and paranoid that it ends up damaging them but then you also have the other way around where it’s like too open too open use PowerBI yeah and every and now there’s like thousands of workspaces and capacity is through the roof and we’re spending tons of money. Like what the heck just happened? We just gave open doors to everything. So I think it’s a balance.

76:21 everything. So I think it’s a balance. It’s going to be a balance between governance and fully open. There’s something that’s there. So let’s let’s give me wrap here on some final thoughts. This was a great topic. Went extremely fast. We burn through an hour very quickly here today. So awesome topic today, gentlemen. Thank you very much for for talking about this one. Love the conversation here. Do you have any like final thoughts around a data and AI literacy? Tommy, I’ll go I’ll start with you. Final thoughts. No, honestly, I think the biggest thing here is you need to treat your AI

76:52 here is you need to treat your AI literacy the same way that you took took excuse me, you need to treat your data AI literacy the same way you treated your data literacy, but with the added understanding that it built, right? It’s not just the consumable consumable literacy. It’s very much on what the output is and where it goes. I would tie even more the importance of AI culture than to AI literacy here. However, that you may want to understand the trusting what you’re doing with AI, but it’s also understanding you and the

77:22 but it’s also understanding you and the context of your team or the organization. Yeah. Yeah. All right, Kurt, I want to roll for perception on data and AI literacy. Kurt, so rolling for perception here. Oh, man. I have lots of other stuff. So I don’t have any dice. I got some tumor stuff going on. There we go. All right. All right. So no. So it’s for me again like empathy I think is pretty key. And I like you have to have empathy in order to facilitate

77:53 empathy in order to facilitate good data literacy. And I think you also need empathy for good AI literacy. not just for the people who are adopting it but also to be able to see things from the point of view of an agent and to be able to know what context it needs and how to interact with it etc. not anthropomorphizing, but critical thinking is valuable in both cases and it’s it just needs to be like baked into the culture of your organization. And it’s important to keep in mind that this is stuff that can’t be forced. and it also can’t be like

78:23 forced. and it also can’t be like done in a very short amount of time. Like it’s something that requires tackling things like with organizational change management. it needs to be done incrementally over time, making a few small right steps in the in the correct direction., and being consistent about that., and I think that’s ultimately,, what’s what’s most important is knowing what that right direction is and defining it like this is the culture we aspire to have and how can you incrementally start

78:53 and how can you incrementally start walking there, while having a plan to be able to tackle the various challenges on the way. I think for me the statement you said in our last episode, Kurt, really resonated with me and I’m really internalizing the words you said. You said the three C’s were important. Be critical. Be be skeptical of some things. Continue to be creative. And I think PowerBI for me was the blast radius of like reporting and data and analytics. That was a a certain blast

79:23 analytics. That was a a certain blast radius we’re talking about. I think when you introduce AI, your blast radius just gets larger. We’re now talking apps and data and process and workloads and workflows. Like there’s a whole the things that I can influence now with AI is just much larger. The blast rate has increased. And then the last one is curiousness. I think I of the three C’s that we talk about, I think I heir very heavily on the curious side. I want to understand. I want to dig in. I want to learn about it. it that’s things that are just out

79:54 it. it that’s things that are just out of my reach of understanding or knowledge really appealed to me. I love going after that direction. So I really like that curious piece. So I I think these are all things that you need to be considered of as an organization. Use the three C’s to your advantage. Leadership take these in effect. and I think the message has to come top down. We got to have leadership communicating that this is a direction. We have teams working on things and we’ll report back. This has to be a learned skill that goes along with the organization. Awesome.

80:25 organization. Awesome. Gentlemen, thank you so much for the talk today. Super great conversation. this is a lot of fun. 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. Helps us a ton. You have a question, idea, or topic that you want us to talk about in a future episode. Head over to powerbi. tips/mpodcast. Leave your name and a great question. And finally, join us live every Tuesday and Thursday, a. m. Central on all PowerBI tips social media channels. We’ll see you all next week. Thank you so much for listening and watching. We

80:56 so much for listening and watching. We appreciate you and we’ll see you next time. time. Take care, guys. Explicit measures, pump it up, be be lighting up the sky. Dance to the laughs in the mix. Fabric and I get your explicit measures. Drop the beat now. Pumpkins feel the crown. Explicit measures.

Thank You

Want to catch us live? Join every Tuesday and Thursday at 7:30 AM Central on YouTube and LinkedIn.

Got a question? Head to powerbi.tips/empodcast and submit your topic ideas.

Listen on Spotify, Apple Podcasts, or wherever you get your podcasts.