Are We Now Professional QA? – Ep.537

The volume of work coming out of a BI team is going up while the hours spent typing it out go down. That leaves a role neither host is entirely comfortable naming: you are increasingly the person who reviews the output rather than the person who makes it. This episode is an honest argument about what that requires, where the fundamentals still matter, and how the career ladder survives when the bottom rung is automated.

News & Announcements

• Skills for Fabric — Microsoft’s repository of reusable AI assistant instructions for working with Fabric, covering workloads, APIs, and best practices. They install as plugin bundles through the GitHub Copilot CLI or compatible editors including Claude, Cursor, and Windsurf. The skills cover warehouse work, Power BI authoring, lakehouse development, and migration — which is exactly the “bake senior knowledge into something reusable” pattern the hosts spend the back half of the episode arguing for.

Main Discussion

Topic: The developer role is shifting from writing to reviewing — and what that costs

Mike’s framing is that we still need developers; we are just asking them for something different. The direct work is shrinking, the review work is growing, and the crux of the episode is whether “reviewer” is a real job or a quietly diminished one. Tommy pushes back throughout — not on whether it’s happening, but on what you have to know to do it well.

• You cannot review what you don’t understand. Tommy’s sharpest point: if you don’t know that many-to-many relationships cause trouble, you will never think to ask about them. General questions like “what are you doing here?” don’t surface the problems that matter. The fundamentals stay necessary even when you never type them.

• The Italian analogy does the heavy lifting. Ask an agent for a governance doc in Italian and you’ll say “looks good, looks Italian.” A senior Italian writer asks which region it’s written for. That gap — noticing nuance you weren’t looking for — is what separates reviewing from rubber-stamping.

• Conceptual knowledge beats syntax knowledge now. Mike no longer values being able to write a pure Python function unaided. He does still need to know what a notebook is, how authentication generally works, and enough structure to ask the agent the right thing.

• Your MCP server is your junior developer. Tommy names the problem directly: juniors became seniors by doing the tedious work — writing the measures, building the models, sitting through requirements. If the agent does all of that, the path up disappears.

• Seniors should be writing skills, not just doing the work. Mike’s answer to the ladder problem is to bake senior judgment into reusable skills and harnesses. Knowledge transfer stops being a call between two people and becomes something the whole team inherits.

• This is the Power Query jump again, only bigger. Both hosts remember hand-engineering data in Excel before Power Query landed. The work didn’t disappear — it moved up a level and made them more strategic. Mike thinks the MCP-and-agent shift is a larger version of the same move.

• Hire juniors for communication and discernment. Define a problem well, listen to the business, translate it into engineering terms. Mike half-jokes that the useful degree here is closer to sociology than business school — and that he wants evidence you’re a genuine learner with real AI practice.

• Trust is Tommy’s closing separator. A stakeholder has to be able to trust what the senior tells them. You can arrive at an answer through prompting and review, but you still have to know whether it’s right.

The dark room

Mike’s recurring image for this moment is walking into a dark room and slowly turning on the lights — we can see a little further each week, but nobody knows the shape of the space yet. Neither host will predict which jobs exist on the other side. What they will say is that the upheaval is not limited to report authoring: data engineering and business intelligence as whole disciplines are built on people spending time making things, and that assumption is the one being removed.

Looking Forward

Pick one thing you currently review by instinct and write it down as a skill your team can reuse — that’s the senior knowledge that otherwise leaves when you do.

Episode Transcript

0:33 Good morning and welcome back to the Explicit Measures podcast with Tommy and Mike. Good morning, Tommy. How you doing? doing? Mike, I am feeling awesome today. I am here, but I’m also not here. How do you mean? Well, I’m in New York right now, but I’m actually in Chicago at the time. We’re Yes, so which I’m eating pizza probably by the by the truck loads at this point. Exactly., I’ll have pizza for breakfast, I’ll have pizza for lunch, I’ll have pizza for dinner. It’s going to be wonderful. But dude, how you been? I’m I’ve been really good. Just got back

1:04 I’m I’ve been really good. Just got back from conferences and traveling over the last couple weeks, so that’s been good. good., , work’s been crazy busy right now, so that’s good. I like that. And then there’s just all these been announcements. I’m just trying to unpack everything that happened at Build. This is the I think this is awesome. People asked us I think when we started the podcast, “You sure you’re going to have enough topic ideas as you keep going?” It’s like, “Yeah, we’re If you looked at our agenda like we’re going to be It’s insane how we’re like we keep looking at our schedule and we keep looking at episode

1:35 schedule and we keep looking at episode ideas like, “Oh, we got to save that for July.” So, yeah. It is a We are in such a good space. I love the space we work in because how quickly it updates. Some people get stressed out by that, but I think that’s what makes it fun. One thing I I was trying to give my son some lessons or some,, words of advice around things that you can be things you can work on. Doesn’t really matter what career you’re in, but, general skills that you need. And one of them was just like,, talking about the like emotional intelligence, being able to talk well

2:05 intelligence, being able to talk well with people. Yeah. Yeah.,, communicate well with them. And one of the items that came up was exactly that, Tommy. Be a continual learner. Be curious. Mhm. Mhm. Don’t just show up and get a paycheck and be like, this is what I do and move on. Like, on. Like, think about how can you do it more efficiently., what can you what what value can you bring to this solution or idea or pattern or whatever the, you idea or pattern or whatever the,, add be be curious, continue to know, add be be curious, continue to invest in it, and then,,, take some ownership of that. Be pri- like have some like confidence that you’re bringing something to the table,

2:36 you’re bringing something to the table,, and being able be willing to learn and bring more new information there. Don’t just show up and and just be like, I’m here, I get a check, and I leave. I think a lot of people can do that, but You got to fire the synapses. Yeah, exactly. So, I think I think we try to lean lean a little bit into that and try to give them some advice of like what I found to be useful and, effective in my career. And that’s those are some of those items that we used. Well, awesome. Well, speaking of firing synapses, my friend, what is on the docket today?

3:07 docket today? So, today our topic is,, really kind So, today our topic is,, really a generic question. How do we look in of a generic question. How do we look in the lens of agents, the agentic world? Are we now in the space where we are doing professional quality assurance, Q&A testing? Are Is that what we’re doing these days? Is that what we have turned ourselves cuz most of our code writing has moved over to the agent side of things. Right. Right., and so now a lot of it is, how do we quantify what the agents write? And

3:37 quantify what the agents write? And who’s vetting its quality? Where does that come from? So, I think this is kind that come from? So, I think this is a a good interesting topic., I of a a good interesting topic., I believe the developer role is rapidly shifting into something new,, but I don’t really feel like it’s I still think we need developers. I don’t think I don’t think we’re getting away from developers., We still need them, but I think what we’re asking of them to do is something slightly different than we were asking in the past. Which is kind were asking in the past. Which is nice. Anyways, we’ll we’ll step into of nice. Anyways, we’ll we’ll step into a bit more. What do you think, Tommy? Anything you want to add to that topic? I feel like we have another contentious

4:07 I feel like we have another contentious topic. I was listening to our previous one on cloud design, which got heated. And we’ll see what happens out here, but no, I think this the topic idea Mike came from conversation I had who’s been in the AI space for a while. And we’re just talking about the future, like where are we at?, you hear the thing about we’re 1999, like it was that really in from an industry point of view where the internet is like, oh, this thing called pets. com, AOL’s pretty cool. And really saw what happened obviously after the bubble.

4:37 happened obviously after the bubble. the person brought up that they really believe that they’ll be looking for professional Q&A as like really the skill that people need and the role. To our point, and we’ve talked about this the past again, a bit of a background here. here. Mike, a lot of what you and I are doing now are relying on the foundation we built with Agent of Solutions, be it MCPs, skills, the harnesses that we’re using. So, us actually doing the direct work

5:07 us actually doing the direct work has gone down. And we’re not the only ones that’s experiencing that. So, really you need people who have skill in that space, but are just reviewing it rather than actually creating it. So, I think that’s for me is the crux of the argument today, be it Power BI with DAX or semantic models or Fabric with Python and pipelines. Are we doing Are we going to be doing less of the work, which means that you need people who can review the code, not

5:37 need people who can review the code, not write the code. That’s where the skill is, both from a junior level and a senior level to know are they doing the right objects right? Are they doing this not so it works, but optimizing it. [snorts] [snorts] And I think those are the big parts of where we’re at right now, Mike. So, with that as our framework today, the groundwork has been laid. let’s see if we can burn it down., let’s burn this one down now. So, I think this is a really good topic. I’m going to go out on on record to kind I’m going to go out on on record to say I don’t think that developers are

6:07 of say I don’t think that developers are in any way being removed. I think developers still need to be here. if anything, I’m seeing the the the work of the developers shifting. Mhm. Mhm. know how you’re if you’re seeing the same thing, too, Tommy, but Yeah. Yeah. developers are used to be ones that are,, would come up with the with the plan, figure out the architecture was, and then would spend a a good amount of time writing the code to do whatever you’re trying to execute against. against. Mhm. Mhm. write write the notebook, write the Python, write the website code, whatever whatever code you’re writing. You would

6:37 whatever code you’re writing. You would spend time there actually physically typing out functions directly. And I think we’ve gotten to a point right now where we moved heavily away from that, and we’re able to get to market much faster. And Tommy, I’ve been showing you some things that we’ve been building at Carl’s Solutions, which is the Power Designer application. Well, we’ve gone from one workload item, called a design, we rebuilt the whole thing. Built initially, right? Mhm. Mhm. Majority of it was rebuilt was manually built, and we’ve gone through another wave of okay, now that we have the

7:07 wave of okay, now that we have the requirements of what’s working, let’s rebuild it. But let’s rebuild it in a way that agents would have better understanding about what’s going on inside that that workload. And so now, within a matter of a couple months, we went from one workload to like eight workloads. Mhm. We have a whole bunch more. Just lots of ideas that have come together quickly. And now that the time from idea to something that’s workable is much is greatly reduced. Really easy for us to get to an answer

7:38 Really easy for us to get to an answer faster. faster. I’m also finding that , breaking your ideas apart is is better. But all this is is all this is Let me Let me wrap this. Let me pull back the idea here. I think the idea is we’re doing We were We were building things one large piece of software at a time. [snorts] [snorts] I think with agents and things, you’re going to get a lot more, let’s call it maybe disposable software or a lot of ideas, or you’re going to be able to build a lot more of it. So, I think the volume of apps and software is going to

8:09 volume of apps and software is going to greatly increase. And I think one of the reasons why I stand behind this is Rayfen just came out, Yeah. Yeah. a very easy way of building an application inside Fabric, but handling the data from the back end of Fabric. Sure. Sure. And also announced, and this is around the spot there as well, Replit at the Microsoft Build conference was announced as a partner with this Rayfen project thing. Yeah. And Replit’s a great program for like vibe coding your apps and putting apps together. And I went by the the Rayfen booth as well, or sorry, the Replit

8:39 booth as well, or sorry, the Replit booth and asked about Rayfen. And they were showing me things that were prompt The prompts they were giving, there’s almost no prompting. Like it was very simple to get really good results out of it. And so, I was extremely impressed with like how this is going to go. So, I think think I think the developer role is moving way more into more into you need to understand architecture, you need to understand the main building blocks and how those building blocks integrate with each other. and I think to your point, Timmy, there’s going to be some level of we’re going to be doing a bit more quality assurance on what the agents provide.

9:11 assurance on what the agents provide. But even then, I think that’s that’s probably a little bit of extra work that we’re getting. I think the larger amount of work that we’re getting is do we really understand the requirements? Are we learning skills to be able to tell the agent or plan things correctly, so the agent can actually build what we expect it to build? build? I keep going back and forth on this because there’s a lot of good points with what you said and I want to make sure I touch on them is we are the work is changing. The actual

9:43 we are the work is changing. The actual day-to-day on my keyboard, on my computer, what I’m doing is changing. That being said, I I’m struggling Mike with this idea of how much do I need to know know functionally or in the trenches to get to that point where I can do the architecture. Because to your point, and let’s break down again the developers role, right? Usually with you got to know the functions, you got to know the code, you got to know not just a single function, but you got to know again how it all relates together. You’re building everything and let’s be honest with Raythen, with every with

10:16 honest with Raythen, with every with what we already have available to us whether it’s a semantic model at this point, to build the relationships, the DAX, the Power Query and MCPs, or with Raythen to build the apps, or what I’m trying to do in

10:30 apps, or what I’m trying to do in notebooks in the CLI, I don’t have to have or it is not required that I know or am competent in the skill, in the language, the coding language that I’m working in. And that’s a fundamental shift. That’s one side of the coin. The other side, yeah, but Mike, really Let’s hang on that one though. Yeah, yeah, yeah, let’s hang on that. Let’s just let’s hang on that that comment cuz I think I think you have a I think you touched on something that’s really useful here, Tommy. Like I agree with you. Like the the ability for me to study what it

11:01 the ability for me to study what it actually writes in Python is probably less valuable now. Asking Tommy to write a pure Python function without any help from AI to do something is probably not as valuable or useful, right? So for anyone. However, one thing I’m finding that is very useful, especially when new people are showing up and asking agents to do things, right? So I still have to understand like so what is a notebook? What does Python code can it do a little bit, right? There’s some There’s some framework around like conceptually you

11:33 framework around like conceptually you still have to be there. However, However, we have an intern who’s asked me some questions around, “Hey, well, I I don’t feel like I know like I want to spend some time writing the code cuz I don’t really know what the agent’s doing in the in some of these apps.” And I said, “Well, the agent knows what’s going on. If you so, spend a little bit more time, slow down on the what did it just build, read through the code that it produced. And so, while you’re not learning the code, Tommy, and not learning to I’m not

12:03 code, Tommy, and not learning to I’m not learning to write it, I am learning to read it a lot more. And the reason I I’m saying this is because anytime I have a question about something that the code did or the agent built or something I question, I go back to the agent and ask it, “What did you do here? Add some more documentation to this. Walk me like if I don’t understand what it created, I’m asking it to build me like a mermaid diagram or something. So, it’s really good at reading my code base and then

12:33 good at reading my code base and then summarizing it back down and reasoning about what the code is doing and producing to me like, “All right, give me a sequence diagram. Give me a an orchestration like tell describe to me the sequence of calls that you’re doing in the code.” Okay. Okay. And when I ask it those things, the things that I’m a little bit more fuzzy on because again, you throw the “Hey, agent, go build me an app.” Okay, it just does something. Like, “Okay, how does the authentication work?” Well, you could If you don’t know how to authentication typically works, you can’t say the right words to the AI to

13:03 can’t say the right words to the AI to tell it what it you need it to do. Right? So, there’s this whole like I need to have a vocabulary that I understand and the agent understands. And I need to be able to say when I don’t understand what it created, I need a skill for me, my skill, my knowledge needs to be able to You need a different word. Yeah, I need a different word, but I need I need capab- I’ll say it this way. I need the capability to ask the agent to describe what it built for me. Help me understand it. Does that make

13:33 Help me understand it. Does that make sense what I’m saying there? It does, but you I think you had a Freudian slip, Mike, because when you first mentioned it, I made sure I wrote it down. it down. You said study versus write the code. My ability to study the code, which is very is fundamentally different than me actually going from scratch on a blank script or Python or, a formula in DAX from white space to actual writing it myself. It is one skill entirely to be able to architect and write it down yourself. Because again,

14:04 write it down yourself. Because again, take the same example you said with a semantic model. If you do not understand the There’s a problem I have with what you said and if you do not know the fundamentals, how do to ask the right questions, right? It’s one thing to ask a general question like, “What are you doing here?” here?” But think about this with a model relationship. If you do not know the conflict of a many-to-many, would you ask that question? If you did not know that was a not a good practice, right? You wouldn’t know to ask that. You would

14:34 You wouldn’t know to ask that. You would go, “Hey, why do you have that relationship here?” But if I’m a professional and I have the experience like, “Is there any other way we Do we have to do many-to-many? Do we have to do cross directional here? Hey, your DAX function has three filter functions inside the same measure. Can we think of a different way?” That’s fundamentally different than just asking a general question. And this is what This is my struggle right now, Mike, truly, because I’ve had the agent tell me best practices before, though. Sure, but how do Yeah, if it’s a language you don’t know, we also know it

15:05 language you don’t know, we also know it hallucinates. I I There is a There’s a modeling skill that Microsoft has produced. Yes. Yes. And in that So again, this is another This is where things things are getting very not gray here, but things are getting like very wishy-washy. Like I know that there are best practices. Right. Right. Do I know where to find them? Mhm. Mhm. Might be a better question to what you’re describing. To your point, you probably inherently know when you’re looking at something there’s obviously something that’s like people have done this before. So, one of these things I

15:36 this before. So, one of these things I think that maybe acknowledging here is like you have to step into situation understanding that Right. Right. this stuff exists. People are doing this stuff stuff every day. I’m just one of many people doing the same thing. So, where do I find my source of like knowledge and information about what this is? And so, again, again, hopefully part of where you’re getting knowledge from is like the podcast. I knowledge from is like the podcast., actually we’ve had a lot of mean, actually we’ve had a lot of feedback even while at the conference. Had a ton of people reaching out and say, “Hey, I’ve learned so much from the podcast. I’m I’m So,

16:07 podcast. I’m I’m So, one source of information that you could have and again, you’re already listening so that this is a good place to get some information about. Tommy and I are sharing our experiences about like what we’re doing. And so, unless you’re plugged into something that’s like regularly teaching you about things you don’t know or pushing into the boundaries and the edges of like what your knowledge limits are, you’re probably not Again, going back to my conversation with my son, right? You’re not being curious enough. You’re There’s not There’s this hunger for being more curious. So, you have to Invest. Invest. Yeah. Yeah. You have to be curious enough to go find

16:37 You have to be curious enough to go find sources of of good information. And again, I like our community. Our community will definitely share who’s teaching and who’s not teaching in the space. SQL BI Exposed to the Metrics podcast., we’re we’re probably one of the most prolific teaching channels out there now. now. We hallucinate all the time, too, on the podcast. Yes, without any drugs. It’s all clean stuff. stuff. Yeah, yeah, yeah. But,, sometimes we we don’t get it always right all the time. But, I

17:07 get it always right all the time. But, I get it always right all the time. But,, the more you interact with things, mean, the more you interact with things, the more you learn. And so, it’s like a progression learning of things. So, to your point, Tommy, right? going back to the best business practice, right? Yeah. Yeah. Relationships, relationship modeling. You may not know it, and I didn’t know it for, eons until the experts invested into me and, tabular editor built the best practice analyzer and then it became part of the Microsoft product and now there’s actually a button in your semantic models that say run BPA on the model. So there is so

17:39 run BPA on the model. So there is so that knowledge is getting baked in to skills, the harnesses, the tools. You got to know how to go find them. Perhaps a good way to differentiate this too is let’s take a look at a senior versus junior semantic modeler or in the in the database developer. In my head right now, and I want to see how much you agree with this, where you land on this because this is a little off the cuff. If I’m thinking of an ideal senior developer that I was going to hire. to hire. Mhm. Mhm. Well, I’m going to quiz them still on a semantic model and the best practices

18:10 semantic model and the best practices and their ability to know DAX, but I will want them to focus and I’m saying this truly on their ability to validate an MCP. Like they need to know those fundamentals as the question, but my expectation for them in the role at this point would not be for them to write everything you [snorts] know, like by themselves off the cuff. That senior developer though has to know not just the fundamentals, but, what how do you optimize DAX? How do you make things quick? How can we make sure

18:42 make things quick? How can we make sure that, many many tables can work together? What are, what is the roach’s maximum? All those things that but in their role on their day-to-day, my expectation is not that they’re writing it all. Like it really would be to vet existing models or to prompt to get the right models. And seriously though, at that point, if that’s the case that’s a Q&A. That’s a professional Q&A. They still have to know the things, but they’re not writing it. This is where I think mhm,

19:14 This is where I think mhm, I’m not sure if I exactly agree with that statement, Tommy. Like Okay. I said this a while ago on the podcast. This is probably like a year or so ago, maybe a year plus more. I have to go search where where I actually said this at this point, but there’s this this idea of your most expensive people in your team are probably your most senior engineers. Mhm. Mhm. in the soft in the software space is if you’re not building if you’re not manufacturing things, Right. Right. world world Yeah, your assets have feet. Your most expensive assets have feet on them,

19:45 expensive assets have feet on them, right? and so those senior engineers who know, to your point Tommy, most like they’ve been doing the big models, they know your data, they know your company, they understand how things are working. I think the trick now is, you can distribute knowledge of those workers directly into agents via skills. So and when I when I was at the the the the a lot of talk on this when I was at the conference. Let’s understand what a harness is. The harness is software code that wraps

20:15 harness is software code that wraps around a large language model. And then what are the skills and the hooks and the custom agents? What are those things? Skills, hooks, and custom agents, those are all things that are like features, inputs, adjustments to the the hardened software that is the harness. So every skill you bring into the harness is some of your business logic, it’s some of your knowledge from your your workers, and so, I’m I’m looking at this, Tommy, and and where I’m going with this is, think about that senior

20:45 , think about that senior developer. What should they be working on? on? Mhm. Mhm. Should they be doing only model building all the time? Or should they be doing Hold on. Hold on. model reviewing or whatever. Should they be doing that? Or should they be doing some of that and some of

21:00 some of that and some of Okay, senior engineer, I need you to work continue reviewing these models, but keep Now you need to bake into after you’re doing these remodels with an MCP server continually building out skills that are useful to us as a business based on what you’re doing in those models. And so, the reason I think this is important is because that most expensive worker is now building plugins, learned lessons, knowledge, information that has now been able to be wrapped in the harness.

21:30 wrapped in the harness. So, the combination of harness and now skills skills can be given to other individuals on the team. And I think this is the real power of this, Tommy. So, yes, I think to some degree, Tommy, you’re right. Who am I hiring? hiring? I think I’m really holding on tightly to the lead developers on my team. Okay. Okay. Because they have the business context and knowledge of what’s going on. I want them to to this is this is changing. No, no, yeah, I know. I I I have a question for you because it’s a hard

22:00 question for you because it’s a hard question. We continue. That senior engineer has traditionally been there to lead other engineers. Right? Other people. Mhm. Mhm. And what we’re shifting towards now is we’re now saying, lead engineer, I need you to lead a team of agents or agentic things in addition to people as well. It’s not I don’t think it can It’s not going to go full one way or the other. I don’t think it’s going to be like only agents or only people. I think there needs to be a combination of the two because there still needs to be a level of like heavy amounts of direction on

22:31 of like heavy amounts of direction on things. things. but if I look at this, Tommy, I’m thinking to myself the the combination of the agents plus skills plus my team I now need to think about what are we what are we leading? Are we leading just people or are we leading people and agents? And what is the right task for the agent portion and the human portion? So, I didn’t realize this, but just what walked into my room with an elephant, and I think we got to address it here. There’s an elephant in the room. In case you didn’t catch that. Let me even realistic.

23:03 realistic. Why don’t we change it up? So, I think this is a really hard question we have to ask right now. I don’t know if we’re going to be able to really answer this today, but what separates then a junior developer and a senior developer if they both have an MCP and they both have Ray then and they both have all the tooling available to them? them? What is the extra knowledge and wisdom and skill that I’m expecting the senior to have that the junior obviously would not? Because let’s go back to the

23:33 not? Because let’s go back to the initial point here. I would expect a senior developer to know the best practices on semantic modeling,, to ask the right questions, whatever the case may be. But a junior developer wouldn’t, right? So, that that’s where I immediately went where well, a junior developer needs to know that too. Most times junior developers are learning, but now there’s a problem that they may not be able to ask those questions. So, I’ll pose it again. What separates a senior developer and a junior developer in their skill

24:03 and a junior developer in their skill then? Because if they both have all the tooling available to them, I know you mentioned the skill like actually writing the skills, but that can’t be it from a pay range, but also someone has 10 years of experience, the other has two, and the only thing the senior developer is doing is writing skills? Right? So, I think that’s where we have to really figure that I am really struggling with this because I think that’s a hard question to answer right now. This is This is This is a great question that you posed on me. So,

24:34 on me. So, I’m going to argue those senior developers have seen things. They’ve been around the block. Right. So, they know the question to ask. ask. There’s this idea of like the business context. How does your company do work?, every time I step into a consulting gig, it’s always like, “Okay, what do you do? Like, how are you How are you working on things? Like, okay, you’re in healthcare. Okay, what What data do you have? How does it work? Where does it come from?” Like, there’s there’s a lot of business context knowledge that I had to go through to get to like a point where I’m becoming a bit more effective in their business and their units.

25:05 business and their units. so as a consultant, one of the things that you’re trying to shortcut, right? So So I would argue a lot of what we do, Tommy, as consultants, we come in as senior engineers. Mhm. Mhm. The only thing we don’t have I would say architects would probably be more appropriate now. Architects, senior engineers, sure. But when we when we come in, we don’t know every single nuance of your business. And I think that would be very clearly communicated like when we show up, we’re like, “Look, your business best.” best.” And we talk with people that are building the pipelines, understand the table structures, understand the data

25:36 table structures, understand the data structures that are there. I I think there’s this another another area of this is that that business context, right? You can wrap some of that business context into skills. You can produce some of that to a harness, to an agent so it understands some of your business context. And I think this is what we’re seeing across like, like, we’re now starting to see custom agents appear where hey, you’re the you’re the software designer. You’re the design

26:06 the software designer. You’re the design reviewer. You’re the design builder, right? You’re the requirements gatherer. There’s all these like custom agent experiences, which is taking that, “Okay, what would I’ve not normally have done as putting that hat on my hat,, that hat on my head and say, “I’m going to be the requirements gatherer. I’m going to go talk to the the business, ask them what they want, what chart they’re looking for, what data we’re trying to produce, what does that look like?” And then I’m going to,, use skills like grill me and other things that are going to help me understand or get the context out of other other team members. So, I think that that happens there. One of the

26:37 that that happens there. One of the other areas I think I will point out here that the difference between a senior engineer other than just pure experience in the industry and knowing your data, I think there’s a there’s a the ability to solve problems quickly. So, So, whether it’s using an AI or whether it’s using just your understanding and knowledge of things, there is I I’m going to say that senior has is able to unwind or notice when things are

27:07 able to unwind or notice when things are running inefficiently and can address and fix them. Or, seniors are also teaching juniors how to do things more efficiently or better as well. well. So, I think a lot of the seniors are setting some more of the the design principles and things as well in your organization. So, I think those are how I would distinguish it. How would you distinguish it, Tommy? What What would you Where do you draw the line between what would be a senior and a junior? So, honestly, this is the This is the hard part for me, Mike, because I think

27:37 hard part for me, Mike, because I think the junior is the one who’s, they’re in the trenches. They’re the They’re learning on the go. So, they’re doing the tedious stuff. They’re,, they’re You have to write the measures. You’re building the models. You have the requirements. You’re learning what it means to do requirements. But, as I say that, Mike, I realize that my MCP is my junior developer. Correct. Correct. And this is a problem here. This is a big problem if we’re going to keep people on Power BI because how else did you How did you become a junior to a senior, right? It was by

28:09 junior to a senior, right? It was by doing. It was over and over and getting these requirements and like, “Oh, I got to figure this out. I got to learn this.” this.” But, what happens when your doing is all done by something else? Right. Exactly. So, this is Therein lies the question. Therein lies the problem, the conundrum here, because the senior developer already has the skill to review. But, why would I need a junior developer? Because they don’t know a lot., that’s the point of a junior developer. Let Let me Let me pull another analogy, Tommy, out of thin air here and see if this is if this resonates with you.

28:40 this is if this resonates with you. Okay. Okay. Let’s imagine the world pre-Power Query. Or, let’s imagine someone who’s not using Power Query and then someone who is using PowerQuery, okay? Heresy begone. Now, continue. Heresy, yes, exactly, yeah. so, pre-PowerQuery, right? What would I do? I’d go download my file from wherever the system was. I’d go in there I’d I’d make a table, CSVs, Excel files, something like that. I’d go download it and maybe maybe on my computer I would have a template file or a file that I’m going to add the data to. So, I’d open up the sample file and I’d open up the real file and I’d

29:11 I’d open up the real file and I’d copy-paste it between one to the other and then maybe I’d have a bunch of formulas that like pre-format and condition and and do all the data engineering that I would want, right? So, it took a person in place in line to get all that work to happen. I believe the conversation we’re having right now, Tommy, with measures and models models is the same conversation we had when we went from no PowerQuery to PowerQuery. And in power Well, now that we have PowerQuery, I don’t need to do all that. Like, it’s not I’m not going and copying

29:42 Like, it’s not I’m not going and copying and pasting manually. I’m not writing a bunch of functions. I’m reducing the amount of errors because now I’m going into a program that has a user interface and that technology basically removed my need to go in and manually transform and clean up data, right? So, that same leap of technology which is what you and I, Tommy, like pushed us into this whole data space, loving Power BI, loving Excel even more, is because that that technology change shifted how we work.

30:13 shifted how we work. And so, we didn’t say, “Oh, no, PowerQuery is gone away. We’re not going to have any more jobs. We’re done. Like, we’re just going to leave the data space.” Instead, we learned the technology and what did what did that do? That allowed us to move to other spaces, which was Well, now I’m worrying less about loading the individual table and now I’m thinking more about like optimizing the loads, how to model all the data, how to pull the How do I get more data into the semantic model? How do I start visualizing and putting reports in front of people? So, our work was like we had

30:44 of people? So, our work was like we had work that we were doing, but now it moved that bit of effort away from our plate and we were able to focus on something new. And I don’t think we got any less work to do, it just shifted what we worked on. It It did make us more valuable because it allows us to be more strategic. Like we can spend more time because we’ve gone in the trenches, also what works. And I’m not just talking the output, like oh the table has the great columns. The value that’s ultimately going to provide to the customer, to the department, to the organization. Sure.

31:14 Sure. And that to that’s where the value add is honestly, Mike. It’s not because I can spend less time doing code. Well, that’s part of it. It’s the fact that you now don’t have to focus of the tediousness of Okay, this function,

31:28 the tediousness of Okay, this function, why is that not date time? It should be date time. You spent 30 minutes doing a Python function to convert the date time from Spark back to Pandas and back again, right? That’s gone. That’s gone from your work from your life. Great, because that 30 minutes now you’re finding out how do we need to use time intelligence?, like why are things the way they are? But think about like if you have a junior going you your eyes squinted. So, I’m going to going to Well, I’m squinting my eyes because I agree with your statement there, Tommy. And I think this is I think this is the point. I think the point is

31:59 point. I think the point is there’s a lot of things now that we don’t have to work on manually anymore. We can really step away from that and we can allow the technologies and so while going from manually editing data and engineering that in the Excel to now using Power Query, I think even now we’re seeing even a larger jump between I’m working with my semantic model myself and I’m using an MCP server to make the changes for me by talking to it. Already I’m preferring even like I

32:29 it. Already I’m preferring even like I know it’s changing me when I’m preferring the MCP server to do even simple small changes. Like even like if I have to edit if I have to edit one measure, I’ll probably do it myself. If I have to edit like three or four measures at a time, I don’t really want to do that. Like now I’m just going to be like I want to tell the agent to do generally what I want. And I’m also finding value Tommy and so one of the workloads that we produce is called Themer. And what we do is now is we have a sample Power BI report where you can tell

32:59 tell an agent to adjust a theme file on a report and I just tell the agent what I want. Hey agent, I had I make all the titles bold. Mhm. Mhm. Hey agent, add the labels to the scatter chart. Right. Right. Hey agent, turn off the X axis label on this particular chart and it just does it. It knows the spec. It writes it. So no longer do I have to go one, I don’t have to dig for the property. I spent so many hours on JSON back in the day. the day. And then two, now I just talk talk to the agent and it knows how to write the

33:29 the agent and it knows how to write the code for me efficiently. So when I look at this I’m going, “Wow, this is a much better way for me to look and narrate.” So I’ve been able to pull my technical view up higher and focus more on like other aspects of things. I’m still doing work with them and designing like there’s still like this,, what is what looks good, what is the art, what’s the what’s the message being conveyed. I think those are things that I still have to articulate and explain to an agent, but I think why this is so revolutionary is it the the jump between like edit in

33:59 it the the jump between like edit in Excel and Power Query was like good. It was a decent size. I think the jump we’re seeing now between, write app and use agent to write app or use agent to edit semantic models versus me at like I feel like the the jump is much farther this time. But the technology is way more capable than it was in the past and I think that’s what’s scaring people. It’s 100%. 100%. how far ahead it’s making it become. Well, and I think we have to take ourselves out of our own experience a

34:29 ourselves out of our own experience a bit here too, Mike, truly because the difference in our role is you and I together have over 20 years of experience in the data space. And we are we got to where we are because we’ve been in the trenches the first 5 years. How many How many models do you think you’ve seen in your lifetime? what? It’s a How many DAX measures did you write? How did you not just writing them for the sake of writing them, but seeing what was effective, what do people like to see?, understanding not just how to write a measure, but what measures to write. We have years and

35:02 measures to write. We have years and blood stains of experience around that. And so, we’re in a position where like, “Oh, yeah, I can have an assistant.” And again, I can look at a pipeline. I can look at the DAX measures and from the many many experiences go, “No, no, that’s not going to work for a dynamic XYZ or,, XYZ or,, that has that Q&A aspect where I would also expect like at this point I don’t want a junior to spend 5 hours building something that I know I

35:32 5 hours building something that I know I can do in 30 minutes. Like, where’s the value there, right? So, Yeah, Tommy, this is where it gets really tricky now. This is where it gets tricky. So, let’s see where I’m going See where I’m going with this here, too. Yeah. Yeah. For [snorts] a junior developer, like, how do you make that jump? Because odds are like for at least you and I, you’re not going to hire a junior developer right out of college with limited experience because I think what we’re coming to terms with is you are in a space now where you need to

36:03 you are in a space now where you need to know what questions to ask and how to evaluate or as you put it in the beginning, to do study what’s been generated. And there’s a pretty problem with that. How do what to study? How do what to look for if you’ve never seen it, if you’ve never seen the problems before? Like, for example, let’s say I gave you something and I don’t know what coding language. Let’s say it was a lot of coding languages, so I’m trying to think Let’s say Let’s say I don’t any I don’t anymore. Like now it’s it’s Oh, no. That’s a perfect example. You

36:34 Oh, no. That’s a perfect example. You Oh, no. That’s a perfect example., but know, but It’s It’s a use it or lose it thing. Like I’m now relying more on my agents than ever to to build stuff for me. me. say this. Let’s say you asked your agent to write a data governance doc in Italian. You’re going to look at it and go go “Okay, looks good.”, cuz you Tell me what it says in English. All right, I guess that makes sense. You right, I guess that makes sense., you’re not going to also stand know, you’re not going to also stand bilingual. This is actually a great analogy., the different the dialect of a different region in Italy. You’re going to go “Looks good to me. Looks Italian.” You “Looks good to me. Looks Italian.”, what thing. I even if you

37:05 know, what thing. I even if you know, what thing. I even if some basic Italian. It’s like, know some basic Italian. It’s like, “Okay, I know that word. I know that word.” word.” Yep. Yep. But depending on where you’re in Italy, it’s going to be either be offensive to someone or not going to make sense because I don’t know Italian. Like or you didn’t know Italian. And we’re finding now for juniors why would I give them all the tooling or, this is this is the hard part now where I’m expecting to come in and roll as a senior to be a professional Q&A. I need them to know the experience to be able to highlight the things that are wrong. The

37:36 highlight the things that are wrong. The reason why I want them to be Q&A is not because that’s just the way things are because that’s the best value of their time. Why I would pay them. It’s not to write and figure out things in DAX. If things can get started with my agents then the best value add of their time is to help review scope, mold, comb, and get the model right to help get the solution down to a customer or the stakeholder.

38:09 I agree with that. But I would also argue the senior should be the one doing that work in lieu of making a skill that does the same thing. 100% Okay, totally fine. And and so I think we I think we agree on this. And the reason I’m saying this is because that senior understands that knowledge, we want them to bake that into skills and things that we can reuse in our organization because then instead of the knowledge transfer happening at the I’m now like, Tommy, you’re the senior and I’m the junior. We’re not sitting down in the same phone call now

38:39 sitting down in the same phone call now together. Like there may be a little bit of this like to explain what’s going on so you understand conceptually what’s happening. Right. Right. But now you Tommy and I are not sitting down together. You’re not explaining to me things. You’re not working through the problems with me. You’re Tommy, you’re saying, “Here’s a skill I wrote. Here’s some optimization patterns that we’re using. Here’s Here’s what I’m going to do. I’m going to build this into an agent experience. And then hey Mike, here’s the skills that we’re using as a company to help us optimize this. And the goal of this is to distill Tommy, your knowledge of what you’re

39:09 Tommy, your knowledge of what you’re tuning that down because again, if our output is more semantic models for more business users, we should be building these best practices practices holistically Mhm. Mhm. or using agents to generate the skills and these custom agents around this and then giving that to everyone else. And so, I guess where I go to this Tommy is,, what does the onboarding look like for new people? And I think there’s going to be this new There’s just There’s a new way of doing this. Yeah. Before we get to that, before we

39:39 Yeah. Before we get to that, before we get to junior because that’s where I want to really dive into, I just want to to give another focus to the seniors because everything you’re saying about the junior is exactly where I want to go here. here. The senior though, just to touch that point and again to use that Italian language analogy, if I had a senior Italian writer, if they reviewed what AI generated, they’d go, “What region of Italy?” They’re going to know the nuances. Right? They’re going to know, “Okay, if I write this to Rome, they like sentences started this way.” Those are things that you don’t learn out of college,, those are not the basic things.

40:09 basic things. But where should that knowledge ultimately land? And they can write a skill to do that. And I think therein lies therein lies then what does the junior do? Because I think you’re touching on something here and make to make sure I heard you right. I think there’s two things to what you said that I heard, but let’s make sure we’re on the same page here. Mhm. Mhm. Are juniors being cheated out of the learning curve? And then like where does a junior what should a junior developer where do they fit and what do they learn first? So, I don’t know if that’s what you’re saying. There’s two points there,

40:39 you’re saying. There’s two points there, but those are the things I wanted to touch. touch. Yeah, that’s a good point. I I think the juniors are going to be actually I’m actually looking for juniors that have different skills, honestly. honestly. Okay, interesting. Right? My senior my my senior engineers are more seasoned around writing code and handling models and doing those things. things. I’m finding newer people to this system are more adept to just working with agents from the get-go, right? It’s it’s the same thing with like my kids, right? I grew up in a

41:10 with like my kids, right? I grew up in a world, Tommy, where we didn’t really have computers when I was young. Like I learned them later on in life and and now I’m using them a lot and and more proficient now, but like our kids are growing up in a world where the very first device they’ve ever touched is touchscreen. Okay. Okay. We never had touchscreen for the longest time. So, their whole mental framework of how things should work is really being framed around like the technology and the things that they are seeing and interacting with right now. And I’ve even seen like

41:41 And I’ve even seen like funny little videos of like little kids trying to scroll random things. Yes, because because that’s what they’ve seen or that’s how they interact with things. So, So,, our perception of what a junior will be doing coming into this new world is going to be totally different. And I honestly think, Tommy,

41:59 different. And I honestly think, Tommy, we can’t explain exactly what juniors will do because they’re going to come in with this idea that I’m agent first. Okay. Okay. I talk to agents to build things. They’re going to they’re already going to have really good skills around how do they prompt the agent to get what they want out of the agent? They’ve been Again, they’re they’re already learning how to prompt it in various ways in other forms. Yeah. Yeah. Even now, I’m working on like I I do a lot of Suno songs. So actually this weekend I wrote I wrote like three or four more songs and they’re not they’re

42:29 four more songs and they’re not they’re now out on Who I Who I Carlos Solutions. So there’s more songs Okay. Okay. I just submitted an MVP Summit song. So that’s that’s out now. So you can go So if you want to go find my music, it’s just fun data, work music that I like to listen to like I build music that I like to listen to while I work with AI. Go look up Carlos Solutions on Spotify and YouTube Music. But I’m trying to build a song with my son and he’s like, “Dad, you’re doing it wrong. You’re writing the wrong prompt.” And he’s telling me, “Let’s go make a new

43:00 he’s telling me, “Let’s go make a new prompt that does this.” And he went He went to another room and he goes, “Use this prompt exactly.” And he was figuring out with another AI how to get the prompt written so he can get the right answer. So the But this is the skill though. The skill skill Oh, I got you in a I got you in a corner, but continue. But now he he’s coming into this whole world of like knowing how to prompt an agent differently than I am. Okay. Okay. And so at the same rate, I’m looking at this world from like I know how to write,, some of the songs before it previously. But now as he’s

43:30 before it previously. But now as he’s being exposed to this new world of agents, agents, that junior like you’re going to want to hire juniors that can come in and say, “Look, I understand what a harness is. I know how to wield skills. I know how to make custom agents.” Okay. Okay. Because your seniors are going to be more resistant to that initially until they get it around their head that they can do these things. So some of what your juniors may be doing will be listening to the seniors, interpreting their information, incorporating their knowledge into skills that can be rolled out to more juniors across the team. So

44:01 out to more juniors across the team. So I I think there’s something there where you want these new juniors to actually explore. And what where I think we’re stepping into Tommy is we’re walking into a dark room. We’re walking into a dark room where we’re not sure what it looks like yet and we’re starting to turn on the light and we can only see as far as we can see right now into how agents are going to manipulate or change what we work on in the future. And so I think that’s my perception there is those juniors are going to build different things

44:33 going to build different things the different things are unknown right now about what they need to build in order to make things happen. All right. Man, I the reason why I’m jumping here because I got you in a corner. I have you’re trapped. You’re trapped. Because brother, allow me to turn on some lights and you’re not going to like what you’re going to see in the room. room. What have we got? So based on what you said, I want you to remember what you said here that the junior developer is prompting agent first. first. Yes. Yes. You said nothing about data or semantic model. So who’s to say you hired junior

45:04 model. So who’s to say you hired junior developer developer and then two months from now they’re like, “Oh, I got a great opportunity at Apple.” Apple.” Apple? What are you doing at Apple? “Oh, I’m building Mac OS now.” So where’s their skill then? Do you see what I’m saying here? If they’re primarily agent first, AI first, harness first, that’s their primary skill, then they can do anything. It’s not just semantic modeling or data. They can now work at Apple to build Mac OS. So where’s anything fine-tuned for your company around business intelligence is

45:35 company around business intelligence is there? there? So that’s the room that we’re illuminating here., I don’t I don’t think it really is., and did So this is one of my one of my biggest here. No, No, data engineering’s about to get wrecked, honestly. It’s about data engineering. I said business intelligence. Even that is about to get wrecked. All this whole this whole space is going to get upheaved because traditionally this has been very people and heavy driven by people spending time making things. I think Raven is a huge change in direction here.

46:05 is a huge change in direction here. all right, the community is already buzzing about Rayffyn already. There’s a lot of people trying to buy code apps. And I was when I was at the Fab Con in Atlanta and I did my full day pre-con session around building MCP servers with data models. That’s that’s where our emphasis was. One, everyone was enjoying it. Two, there was a problem getting tokens cuz it’s going to be continually a problem. So, cheap efficient use of tokens is going to continue to be the issue. so, that’s going to be something we’re going to figure out. but everyone in that

46:35 to figure out. but everyone in that session was asking for, “Okay, great. Now we can edit the model with a agent.” Everyone wanted to immediately start editing reports with agents as well. And now we’re starting to see skills show up. up. Yeah. Yeah. And now there are skills for Power BI semantic models. So, if you go into the there’s a link that I’ll put in the description here. Or no, that’s Skills for Fabric. Yep. And inside skills for Fabric, there’s now a report authoring skill, a report design skill, report management skill, report planning skill. These are all skills that

47:06 skill. These are all skills that Microsoft is making. So, that that senior knowledge is now being baked into skills that are starting to appear across multiple places in multiple areas. So,, the new senior should be doing the the seniors that are we have now should be able to look at what does report authoring, design, and management skills look like? Where are these coming from? How are we becoming more efficient? So, the new senior role is not is shifting from can I use an agent to how do I optimize said

47:37 I use an agent to how do I optimize said agent 5, 000 skills and things. 5, 000 5, 000 we’re just sh- again, we’re we’re just shifting what we’re working on. The work is not changing. So, let me go let me go back to what my point earlier about the juniors again, right? Juniors need to come in and need to know how to use agents. agents. Okay. Okay. I’m not going to assume they’re going to know how to use agents efficiently, but that’s [snorts] where my seniors have to come in and say, “Okay, look, how are you building things with agents? Think about not only I’m going to give them obviously more responsibility cuz they’re getting paid more money. Like,

48:08 they’re getting paid more money. Like,, when you use prompts to get something done, I expect your token usage to to provide the same level of quality. Interesting. higher higher thing, yeah. higher amounts of like, quicker, right? And do it with less tokens. So, if your senior isn’t doing those things, that’s what seniors should be doing. A junior will come in and say, “Oh, I know where these skills live.” And they just prompt and prompt and prompt and prompt, and they’re going to burn more tokens because they’re just not as knowledgeable of what the senior’s doing. So, there’s this really

48:38 senior’s doing. So, there’s this really interesting we’re And this is why I said the dark room analogy is is what we’re talking about here is because we’re walking into a space right now. We don’t even know what new jobs need to exist. I agree. I agree with that. We don’t know and the AI is changing physically what we work on. Yes. Yes. How do we How do we Like, Tommy, who’s going to set all the government’s policies and the the process around what app what fabric apps we should build? I agree. I agree. Where do they go? How are they governed?

49:09 Where do they go? How are they governed? How do we deploy them? Who maintains them? Like, this is a whole brand new thing that we’ve never had in touch before. before. Right. Right. We can build with D3. js. We can build with Vega. It doesn’t matter what language you use. You can build with anything you want. Like, so, how are we talking about So, what has happened is our our ability to create has just greatly expanded, expanded, but but we don’t know what parts of that will be most effective for our company and our business. business. So, that’s yeah. that’s where we need to apply the seniors. That’s where the seniors are

49:39 seniors. That’s where the seniors are going to move into. I I completely agree with that, but I think the biggest thing in I think that the big part here is we don’t know what a junior’s going to be because again, using your logic, you guys may hire someone who built the calculator for Apple on their iOS app. They’d be the great at prompting. Right? Because I don’t think you will. I don’t I don’t I don’t think I think there’s I think there’s generally like an interest, like an area that is like intriguing to the brain that people will naturally just fall into. Like so Tommy, you and I will probably stay close to data and

50:09 will probably stay close to data and analytics and visuals because that’s what drives us. We we’re interested by that. That’s a that’s a subject matter area that we’re interested in. yeah. yeah. Yes. Yes. No, 100% but I’m saying who’s to say? Right? Cuz if and again, I’m just using I think the I’m trying to connect the logic that you’re saying. If the primary role of a junior developer is and they’re going to be using MCP, they’re going to be using EI, they’re going to be using the harnesses and they’re very good at that, then who’s to say they can transfer from building the iOS calculator to all sudden building

50:41 iOS calculator to all sudden building, things in Raefen and fabric data apps. They It would translate, no? I would argue you’re going to want to I would argue most of what we’re going to do is going to move more towards Raefen and that building. I Right. But that’s what I’m saying, right? Then it’s not business intel So and again, a junior developer is not just developer and I think we have to This is where This is This is a huge part Mike of what we do. We don’t just build apps. We don’t just show things on a report. What have we talked about so many times

51:11 What have we talked about so many times is you need a junior What have we always looked for junior developers? It’s not great DAX. Is their ability to, problem solve? Is their ability to work with the business Those still We haven’t talked about that. We’ve been only talking about the coding part of this, too. And And we’re we’re running into a lot The dark room is very dark in here. It’s very dark. There’s a lot of There’s a lot of unknown things. There are no windows right now. Try to figure out what’s going on here., , yes, this is all part of the this is all part of what experience comes from this.

51:41 part of what experience comes from this., there’s there is a experience. there is how to, how do you define But, again, yes, Tommy, I understand these things. Like, we’re defining requirements, right? The same thing we’re asking juniors to come in and act like a senior engineer, right? Because of the the way they they it, yeah. it, yeah. Because the the the agents are able to give so much more capability to these lower-level items that we just don’t need to do anymore. So, to your point, Tommy, Tommy, if we’re hiring juniors, what should we be looking for? We should be looking at

52:12 be looking for? We should be looking at juniors that understand how to articulate requirements, Mhm. Mhm. right? Who can basically be that liaison between the business and the AI. So, our our thinking as leaders in our companies need to be thinking about what

52:27 companies need to be thinking about what tasks do we need to get done that can be done by agents, and what tasks we need to get done that are needing to be done by people. And so, our thinking of this has to be broken down now into two lenses of the world. What’s agent-based and what’s people-based. And so, in the people-based side of things, it’s going to be high It’s It’s emphasizing more ability to identify clear problem definitions, because the better you can define the problem, the better the agents will do. The better you can define a problem, the

52:58 The better you can define a problem, the better you can articulate said problem to the business or the challenges that come from there. The better you can listen to the business and interpret their problems and articulate into engineering terms, the better you’ll be. So, like, the skills are moving much more towards good communication, and high emotional intelligence, having empathy for the business. Like, those have always been there, but now more than ever, the technical barrier to get to do those things is now being removed, and you need to have more of those skills. Higher amounts of emotional

53:28 Higher amounts of emotional intelligence, because that’s that’s something that the agents aren’t necessarily going to be able to absorb. 1, 000%, and it’s funny you say that because I’m just starting reading the book about the founder of Anthropic, who we Surprise, surprise, why we like Claude so much? He’s Italian. His parents are from Italy. So, I didn’t even know that. I didn’t even know that. I’m like, of course, that’s my But, no, but in in I have been thinking more and more like, what would be the best degree for an undergraduate who wants to get into this space?

53:58 into this space? And And maybe it’s a bias, but like you just spoke about understanding human behavior in a group setting, human behavior, you in a group setting, human behavior,, understanding why people behave know, understanding why people behave the way they do, so to speak. I don’t I don’t Are degrees even relevant anymore, Tommy? Tommy? I would say That’s a That’s a discussion for another day. That’s that because I have opinions here. But, if you were to get a degree, if you were to get a degree, because I I think you and I already feel the same way about that. thing. You’re not building a bridge, flying a plane, or checking my

54:29 bridge, flying a plane, or checking my tonsils, thing. So, like I I think you can get away with it. That being said, if you wanted to go to college,, it’s not business school. It’s I feel like it’s like sociology or something in that space, where understanding how people interact, to have to learn empathy, to learn what what causes people to move in a certain direction. And then it’s like I want I want to know that you have that skill, and I want to know that you have practice with AI, and that you’re a good learner that way. Because I don’t need

54:59 learner that way. Because I don’t need you to have a statistics background. Because you need to know what num- what presses someone’s button with a number. That does not mean you need to know algebra. You need to be a hot taker, but I don’t I’m going to go after this at a different angle slightly, Tommy. I’m I’m going to say the the main So, college education, I think it’s still very valuable. I think it’s a lot of other things here. here. No, I think it’s still valuable. Yeah. Yeah. But, the pure knowledge part of it,

55:30 But, the pure knowledge part of it, Mhm. Mhm. like information is so easy and cheap to get anymore. Like it it used to Quickly changing. Yeah, and and so the ability to get access to this information is just so fluid right now. And, I I look at this Tommy going For a $300 a month subscription, I have basically all of human knowledge distilled down into some model that I can go use. Mhm. uncomprehensible, like what this is doing. The the the amount of capability

56:01 doing. The the the amount of capability that that single thing provides to us is incredible. You could take a $100, 000 scholarship and say stop. You could pay $300 a month for your entire career and still not come up to like $100, 000. Now I know, right? tokens and things may get more expensive, like it it could change it, but if you think if you think about like that amount of investment, where’s the better part of your investment?, I think the college experience right now is about like emotional intelligence and finding people that are different than you and how you work with people and like there’s a lot of other emotional

56:32 like there’s a lot of other emotional parts of this that is like the human side of things that, kids and and young people have to go through to get through that college stuff, but anymore the knowledge to do the work is now baked into AI is all over the place. So So what does that look like moving forward? And And this is where I’m going to argue like I’m not sure what what this future looks like, Tommy, but AI has greatly changed it. Information’s gotten super cheap. YouTube has changed how we learn. so I’m going to argue like yes

57:03 learn. so I’m going to argue like yes you need to understand statistics. But the AI already knows how to run all the statistics for you. what I’m saying. That’s my argument. Yeah. Yeah. So what I’m where I’m going to go after this one is I think we’re we’re moving into a world and I heard I heard this on YouTube. This is not my idea. This is someone else’s. I don’t remember who was who said it, but we’re moving into a world where everything is becoming whitelisted as opposed to blacklisted. In the past, I would accept a lot of information in and I would blacklist a source of data and say that’s not good.

57:33 source of data and say that’s not good. That was telling me lies. I’m not I’m going to ignore that source of information.” Right. Right. On the flip side now, I don’t know what’s really real. Every video with scrutiny. Every Like literally now the comments with my family I’m looking like that’s AI. That’s AI. Everything’s AI. So, discernment in what’s real, discernment around how to evaluate what truth are things are is going to be incredibly important Yep. Yep. important now moving forward. So, just because you prompt an AI,, I I had a an experience.

58:05 , I I had a an experience. someone was asking me help me to work on their their engine and we were going to go do a spark plug replacement. And so, we bought some spark plugs. I didn’t buy them, they bought them. They brought them over and they were like, “Oh, here’s the spark plugs.” And I’m like, “These are the wrong These are the wrong size.” He goes, “Well, that’s that’s what ChatGPT told me.” told me.” Yeah. Yeah. Go look at the manual cuz that’s where the source of truth was, but we just asked ChatGPT thinking it knew the answer and actually it got the wrong spark plug. So, the threads are right, but the the depth was too long. So, like it was literally wrong. And so, you

58:35 it was literally wrong. And so, you it was literally wrong. And so,, that there is a point of this like know, that there is a point of this like discernment element. The AI is super eager to give you answers for everything. How do Like your seniors need to be having having your juniors have better discernment around what is really the AI doing and how do we steer it, direct it so we get the output that we want. we want. Yeah. Yeah. And I think I think more than anything we’re moving into an area of white listing, meaning these are the information and sources that I trust because there’s real information, it’s

59:05 because there’s real information, it’s truth behind them. We have to white list them and say, “That’s what I’m going to allow to to be educate me because I now know that that’s they have experience. They have truth. I’ve gotten proven reliability out of that source.” And I’m that’s going to be close to my closing thought, too. I think what’s going to separate a senior and a junior is trust. For when a senior is speaking to the client, a speaking to the stakeholder, you got to be able to trust them. And it you can go through prompting, you can go through Q&A, but they got to know what they’re

59:35 Q&A, but they got to know what they’re talking about because you will eventually talk to a human. I do want to make a special note though, if you are an analyst now or if you’re not in that senior role and maybe this sounded like there’s no place for you. I don’t think that’s the case at all. I think the big part is what Mike talked about is we are in a dark room, but everybody is right now. How can you position yourself to be in a in a place of value,, in a place where also you’re going to flourish as well? And I think it is going to be a big part

60:06 And I think it is going to be a big part of really leaning in to these the new era that went. It is not going away when it comes to especially Wraith in right now, but that being said,, you can still build reports in Power BI desktop, but utilizing those tools to make you more efficient, utilizing those tools to think more strategically. If you’re junior, you probably don’t need to write a lot of DAX and that’s okay. You need to read the stuff, you need to read the books and you need to know what works. For example,,

60:36 know what works. For example,, go get DAX the fundamentals of DAX. Try things out, make sure things, verify. There is going to be more Q&A in your life. But I think we’re coming to the conclusion here. I don’t think that’s going to be the only role you have. And really you need to be able to talk to the business, understand their problem, translate that to data. That will not change. So that’s really my closing thought, but Mike, love the conver- Is there anything you want to add to that? Anything that Yeah. Yeah. I think I’m just going to say discernment. My my biggest thing is like

61:06 discernment. My my biggest thing is like landing on how do you discern what is really what is the agent really doing? what is really value added for your company? Discernment is where you’re going to let it lean into your senior developers and I think how do you discern what source of truth is really useful to your business, your company? Where do you listen, where do you get your information from? Because the agents are going to make it very useful. and I think the juniors will lack that discernment level, but your seniors should have that discernment level as you integrate more and more with agents. So, that’s the my final thought there as well.

61:37 there as well. With that being said, thank you very much for listening to this podcast. We hope you enjoyed our professional quality assurance conversation. Who’s going to be doing the QA on things? What does this look like? Where do people, developers, and agents all sit in this new landscape that we’re stepping into? and hopefully you find a good source of knowledge in this podcast. Hopefully this has been helping you think through and unpack and then put ideas in your head that are going to help you move your business forward. That being said, Tommy, where else can you find the podcast? podcast? You can find us on Apple, Spotify, wherever you get your podcast. Make sure

62:07 wherever you get your podcast. Make sure to subscribe and leave a rating. It helps us out a ton. Do you have a question, idea, or topic that you want us to talk about in a future episode? Head over to powerdadtips. podcast. I forgot that part. Oh, well. /podcast. /podcast. /podcast. where we’re at. So, and finally, join us live every Tuesday and Thursday a. m. Central and join the conversation on all the Power Dad Tips social media channels. We’ve only done this 537 times. times. Yeah. I had to mess up once. So. All right. Thank you all very much and we’ll see you next time.

62:37 we’ll see you next time. Hey! Express yourself. Pump it up. We out. Tommy and Mike lighting up the sky. Dance to the data. Laughs in the mix. Fabricate a day I get your fix. Express yourself. yourself. Drop the beat now. Podcast kings feel the crowd. Express yourself.

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.