Power BI Veterans React to the Fair Game AI Book

Speaker 1 (00:00:04): Welcome to Raw Data with Rob Collie, real talk about AI and data for business impact. And now CEO and founder of P3 Adaptive, your host, Rob Collie.

Rob Collie (00:00:20): Hello, friends. About 15 years ago when my twice-weekly blog about Power Pivot, the forerunner to Power BI, was about a year old, readers of the blog were increasingly telling me, "You need to write a book." And I resisted that for two reasons: one, sounded like a lot of work, and it was; two, I didn't think I understood Power Pivot any better than the people who were telling me that I should write a book.

(00:00:46): Now why would I say that? I've had a successful career in tech, but it's not because tech is second nature to me. My success with software, both at Microsoft and later with P3, is almost in spite of my relationship with tech rather than because of it. So I didn't think of myself as an authority on Power BI. I just thought of myself as someone who was reasoning his own way through it and sharing his experiences on my blog. But then I saw a few specific examples of people who I knew were more technically capable than I, but that did not understand Power BI despite their best efforts. That's what put me over the edge to writing a book. It seemed like an imbalance in the natural order of things that I could write DAX, and some of these super smart Excel MVPs that I knew could not.

(00:01:33): So I wrote a book, and those Excel MVPs read it and became almost instantly better at DAX than I was. Mission accomplished. I revised the book for a second edition, and that's the book that was really popular. And then I retired from writing books because writing books is grueling, writing books alienates me from my family, and the world also just took the ball and ran. A new generation of authors and an entire community developed, and I didn't feel needed anymore. As you know, I found myself recently unretiring and writing another book after all, and it's for the same reasons I wrote those Power BI books. I found myself understanding something important and knowing that others can and will understand it better than I if they're given a starting point.

(00:02:19): So for today's podcast, I thought we'd close the loop. I invited three people who read my 2015 Power BI book and who have now read my AI book to a bit of a round-table discussion. Even just finally introducing these three people to each other after all these years of knowing them one-on-one was a real joy. I hope this conversation will help you see how data professionals have an exciting role to play in the new AI story. And I hope it puts a smile on your face because it certainly put a smile on mine while we were recording it. So let's get into it.

(00:02:52): Welcome to the show, a three-guest day. We've done this I think one other time in our history where we've had three guests on the show. I'm just going to go in clockwise order on my screen. Welcome to the show for the second time, Jon Perl.

Jon Perl (00:03:07): Hello.

Rob Collie (00:03:08): Also, welcome to the show for the second time. I believe, Tim, you've been on the show before once, right?

Tim Rodman (00:03:13): Yeah.

Rob Collie (00:03:14): And welcome to the show for the first time, Trent McKinster. Thank you all so much. We've got all the time zones covered here. Jon's got the New York City thing handled. Tim is normally Eastern Time Zone, but is currently in San Diego and is sitting in front of the exact same type of wall that I'm sitting in front of, which is incredible coincidence. Way to go. But Trent is really showing his dedication to the cause. Trent is calling in from Hawaii. He's interrupting his vacation to Hawaii to do this with us. The mixture of gratitude and guilt that I feel, it's like a whole new emotion, this blend of the two. Appreciate you all being here. Let's just go around the corner really quickly and introduce ourselves. So, Jon, let's start with you.

Jon Perl (00:04:03): Yeah, my name is Jon Perl, hailing from New York City. I am the founder of Syncuity, a data intelligence platform for wholesale manufacturers. Longtime fan of Rob's. The multiple books have each formed pivots in my career, and look forward to talking about them.

Rob Collie (00:04:22): Yeah, I mean, I've kind of stacked the deck here. I mean, that's something I'm going to acknowledge here in a moment, but it's a good, honest kind of deck stacking, so we'll circle back to that. Tim.

Tim Rodman (00:04:31): Hey, everyone. My name's Tim Rodman. I've been in the mid-market ERP space, so company's doing about 10 million a year in revenue to around 500 million a year in revenue. Not big enough for SAP or Oracle, but I've been in that space for about 20 years on the consulting side, and now I'm working for accounts payable automation software company called Trailed. Accounts payable, super boring stuff, but there's a lot of details you still have to get right. And I like to say I crossed over to the dark side of sales and marketing to do this. It's been about two years in that role and it's been a lot of fun.

Rob Collie (00:05:03): Not really sure I was caught up on that, actually, so maybe dive into that later. All right. And then lastly, Trent.

Trent McKinster (00:05:10): Hi, everyone. Thanks for inviting me here, Rob. Super excited to be on the podcast. I've been a long-time listener of the podcast and a huge fan of Rob's since the inception. I work for a large CPG organization, VP of category business. Also crosses the technology side of the business. And I spent over 25 years of my career with a retailer, a large grocery store chain where I also mostly worked on the business side, but also touched technology wherever I could.

Rob Collie (00:05:39): All right, so you worked on both sides of the CPG spectrum, the supplier and the retailer. Fascinating business that I learned a lot about. Got a crash course in that between 2010 and 2013. I'll never see the world differently now that I've kind of seen behind the curtain.

(00:05:53): So, yeah, like I mentioned when Jon was talking, we're here to talk about my new book, which is... The time this podcast goes live, it will be out. This is our launch week podcast, and I wanted the three of you here to talk about it. One, I think, really important thing all three of you have in common is that you all read my first books or book, depending on how you want to look at it. I mean, there's really one really good one that came out in two different versions, and we met essentially through that book. On the one hand, having three people on to talk about my new book, who loved my first book is like, yeah, there's a little bit of a selection bias there. I'm kind of putting my thumb on the scale. I made three friends, not just business colleagues in various forms, but I also made a bunch of friends by virtue of writing a technical book, and so I want to give myself fair credit for that.

(00:06:53): At the same time as I'm bringing in a friendly audience to talk about the new book, the friendly audience is one that the first book came by really honestly. I'm very self-deprecating. I'm always telling you what I can't do. Folks, I am just so, so, so proud and validated that I could write a technical book and meet people like you and form these kinds of relationships over something which would normally be just so incredibly dry. So I feel like I'm in my superpower zone here.

Jon Perl (00:07:21): Pleasure's mutual.

Rob Collie (00:07:22): Thank you so much. Let's just get it out of the way. Let's call my previous Power BI book, let's just call it one book. How did that book impact your career? Because all three of you have told me that you've done things differently as a result of colliding with that book. Let's get the Pepperidge Farm Remembers segment out of the way, and then we'll switch and talk about the new stuff.

Trent McKinster (00:07:47): I think for me, I thought I was buying a technical book, and you taught us all the technical things, but really you were teaching us a new operating system for how businesses can think different, and that's what really surprised me at the end of the book. Because I learned the skills, but I also learned the ability to go back to my business and ask different questions than I had asked in the past. And I think the AI book is kind of doing the same thing. I think you basically disguised an AI book as a leadership book. So it's really helping the leaders of organizations take that next step and get started and thinking about now that we have this capability, how do we think about things differently?

Rob Collie (00:08:27): So here's an asset test question: is it an AI book disguised as a leadership book, or is it a leadership book disguised as an AI book?

Trent McKinster (00:08:36): Well, it's pretty interesting. It can go either way. And if we ask AI, it would talk back and tell us, right?

Rob Collie (00:08:40): It would indeed.

Jon Perl (00:08:41): It wouldn't take a side, probably.

Trent McKinster (00:08:43): No, absolutely. Probably agree with whatever question you brought out.

Rob Collie (00:08:48): I don't know. I mean, I've taught Eddie, my editor, to be somewhat confrontational at times. That's interesting. Okay, so the first book, Power BI book, not only taught you Power BI, but kind of changed the way you thought about business, which is really interesting because at the time you were reading it, you were already a long-time veteran of business and a veteran of business in particular industries as well. So it wasn't like you'd been dilute in terms of your business experience. So that's really cool to hear.

(00:09:13): Jon, how about you? I know your story all too well, but the world needs to hear it.

Jon Perl (00:09:20): I started on Wall Street out of college. I worked for a hedge fund, making business loans. Everything went out of business in '07, '08. There were no jobs in finance. I started a clothing company. I built a clothing company from '08 to '18, and I sold it in 2018. And I had been doing this business for eight years when I read the book. The clothing business at scale is a very, very data-intensive business. You're dealing with major retailers and EDI and distribution. I mean, ERPs generate a lot of data really fast. The number crunching is just a consuming thing in my life. I'd started to build all kinds of apps and access and whatever was available to me. And then I came across Rob's book. I see these buttons in Excel. I didn't know. They just popped up one day. And I wonder what they do, and they look really cool. And I bought Rob's book.

(00:10:07): And it was a holiday. I'm Orthodox Jew. So at some point... I don't drive on Shabbas, I don't drive on Saturdays. I don't travel. I don't have my phone. I read books or with my family or whatever. So at that particularly heavy time, I had the book, I bought it, I had some time to sit and read it. I remember reading that book and then just telling my wife all this stuff that I've been sort of tinkering with, it's all been streamlined, it's all been democratized. Now you don't have to be some fancy developer to do all these things. You could really do a lot of damage as a citizen developer, I think is what we used to call it. And I decided to completely pivot, sell my business, keep the IP to any software that I built, which was quickly replaced by all sorts of really beautiful tooling, and just utilized the Microsoft stack to build technology for that industry. For the last eight years, I've been developing ERP-related technology and made an entire career out of it. I did have to learn some of the coding stuff, but it worked out.

Rob Collie (00:11:04): Sold your clothing company, planted the foot, pivoted. It's a big deal. A lot of courage there. Salute. Tim?

Tim Rodman (00:11:12): For me, business intelligence was always this crazy thing that I didn't think I was good enough for. Maybe I'll describe it that way. Especially at the time back in 2013 when I read the book with OLAP cubes and the super technical developer stuff that you needed to know, I didn't think it was something that I could grab onto. But whenever I recommend the original Power BI book to people, I describe it as reading a novel. I remember picking it up, starting to read it, and it was hard to put down. You just have a good way of that writing style, yet at the same time delivering the technical topics.

(00:11:49): With the new book, the Fair Game book, I've found it to be very similar. And I'm looking here at a post on LinkedIn by Ahmad Chami. He describes it as feeling light. And to me, it's a similar description. Reads like a novel, feels light. You don't realize that it's a technical book a lot of times, and you just kind of have that skill of writing about things, while at the same time delivering very substantive stuff. It's not like you're glossing over it. And I think a lot of it has to do with you have to understand a subject really well in order to be able to describe it in a simple way. And that comes through to me in both the original Power BI book and in the new Fair Game book.

Rob Collie (00:12:24): I really appreciate that. The things we think of as strengths and weaknesses are oftentimes just characteristics that have positive and negative places where they're helpful, places where they aren't. And so I think two of the things that you could think of them as weaknesses of mine are strengths when I'm writing a book. One of them is... You pointed this out to me, Tim, on our Circle social media site that we're using for the book, and I've been using it ever since, this notion of an interest-oriented nervous system versus an importance-oriented nervous system. So writing a book is very important and valuable to one's career. It's a good career move if you do a good job of it. But that doesn't power me through writing a book. No way. Writing a book is hard. The payoff for it isn't enough. So I have to keep it interesting for myself while I'm writing it. By keeping it interesting for myself, I think I also end up keeping it interesting for the reader. What comes across in my books as calculated, like, "Oh, he's really holding my attention," is really just me taking care of me in that moment. I say it I think in the introduction, if I'm not holding your attention in chapter 12, I'm not holding mine either.

(00:13:37): And the other one is that I also say in the book, I think in the intro, even though I've been in tech my whole career, I've never been as into it as most of my tech friends and tech colleagues. I don't take to it as quickly as most of my colleagues. And so when I truly come to understand something, I'm more sympathetic to the people who don't just absorb this stuff by osmosis. And so the way that I would explain it to people is more like how I would've needed it explained to me. I think both of those, again, characteristics of mine come together and help me as an author. Whereas in other aspects of my life, these things are headwinds at times.

Tim Rodman (00:14:17): Yeah, I don't know what word to use for the opposite of condescending is, but your books come across as being the complete opposite of condescending. Whereas, yeah, a lot of tech explanations are ivory tower, you sort of have a condescending tone when you read them.

Rob Collie (00:14:31): Almost like a blue-collar kind of smart is my favorite kind of smart.

Jon Perl (00:14:35): It's like putting spinach in the chocolate chip cookies, but they still taste like chocolate chip cookies. You still snuck in the spinach.

Rob Collie (00:14:43): Yeah. All of you got a PDF of the book, an advanced PDF of the book. No one has in their hands, as of the moment we're recording, no one has the hardcover in their hands. That's going to start happening this week and next week. But yeah, so you all got PDFs. We know that you were all reading PDFs, but tell me about where you were when you were reading it. How many sessions did it take? Where were you in your life when you were reading this PDF that I emailed you?

Trent McKinster (00:15:14): For me, I was in a pretty interesting spot. I was actually on the AI journey. And it's so easy to get overwhelmed if you watch social media and you try to understand AI, especially when you're trying to take care of your day-to-day topics as well. I'm sure we'll get into what a crafter might be. I found that the technology was unlocking doors in different ways for me. So personally, I'm a workaholic. Sometimes I even tape podcasts from Hawaii. I found that I do have some extra time on the weekends where I can kind of play around, and these tools were unlocking capabilities.

(00:15:47): So like your story, I was kind of developing my first iOS app and working on that on the weekends and nights. So I took it a chapter at a time. I had family visiting. It took me a little longer in retrospect than I really wanted to. I finished it on a plane on a trip back from a work function. The way you've always been able to simplify concepts and the way that you make it easy to understand, and the way that you make it simple enough to just get started... You don't have to be perfect. You don't have to understand it all. Just get started and get comfortable with it... I think that's one of the things that will make people successful with AI like with Power BI is you don't have to be perfect. You don't have to be an expert. You just have to get comfortable and get started.

Rob Collie (00:16:32): That feels really good to hear.

Jon Perl (00:16:33): The book for sure, it draws you in, as your writing style does. It's hard to find time to read, sharpen the saw when you're dealing with implementations and clients and whatever fires that come up. I tend to read on the Sabbath on my off time when I'm unplugged from technology, but I have to print it out, so I printed out the book. It was two up, I think I printed it. It didn't staple when you print it even at that number of pages. I separated kind of arbitrarily at the intermission. Intermission? Yes, intermission in a book.

Rob Collie (00:17:06): A tradition.

Jon Perl (00:17:07): Look forward to getting that part, staples table. Set it up for tomorrow. And I would read it on Sabbath over the course of a few weeks.

Rob Collie (00:17:15): You sent me pictures of it printed out. Of the 90-something people I've sent this preview PDF, I wonder how many of them have sat down and pressed Print. Not too many, I think.

Jon Perl (00:17:27): Exactly.

Rob Collie (00:17:28): The book comes out at 200-something pages. It's fewer when you print it at home, I think. But it's still, like you said, it was thick enough you couldn't staple it.

Jon Perl (00:17:37): Exactly.

Rob Collie (00:17:40): I think that's a really cool story, like a ritual thing that you did every week for a little while.

Jon Perl (00:17:46): I still do it. I don't always have the motivation to go back to it, but in your batch was your book, and together with your book was the Claude Skills Guide, the complete guide to skills. I searched the internet for PDFs that are printable and a couple of white papers from Microsoft. I think in the picture you could see what's dog-eared and folded and what's not.

Rob Collie (00:18:04): I'm going to go back to that photo and be looking at it like the Zapruder film. What else is he? Oh, there's another lone PDF on the grassy knoll over there.

Jon Perl (00:18:13): The mainstream slop did not get read.

Rob Collie (00:18:14): All right, Tim, where were you in your life when you were reading the PDF?

Tim Rodman (00:18:20): Yeah, so one of the themes I think that comes out several times in the book is basically you're not too late to the party, but you're also not too early. You want to get on the stuff now. And for me personally, I ignored AI, kept my head in the sand like an ostrich, but things changed in January of this year, 2026, when I got a text message from a friend about something he had built over the weekend with OpenClaw. I'm like, okay, this is real. It's not just a bunch of marketing stuff anymore. So that got my attention.

(00:18:46): And because of that, when I read the book, I knew enough at that point to know that AI really likes markdown files. And so I've been using this app called Obsidian, which is basically like Microsoft Word, but instead of storing docx files in this complicated format that has a bunch of junk in there, it just stores everything in markdown files. And so what I did, I'm going to get in trouble for this, but I promise it will not be shared, I copied the PDF chapter by chapter into Obsidian markdown files and did my highlighting in Obsidian, and it was a pretty fun experience.

(00:19:18): Since then, I've discovered this app called Readable that allows you to take all of your Amazon Kindle highlights and easily bring them right into Obsidian files. So I'll be doing that, actually, once the book comes out officially on Kindle. I'm going to read it on there and suck my highlights into Obsidian that way.

Rob Collie (00:19:35): Our chief customer officer today asked me if we had a markdown version of the book, and I said, "Well, no, but I could make one pretty quickly." He's like, "No matter. I'm on it. I've got it." And so I was like, "What are you going to use this for?" He's like, "Well, it's great. It's great backgrounder for talking to customers and stuff." I'm like, "Why wouldn't I use it?" I'm like, "All right."

Jon Perl (00:19:55): Markdown was a very big pivot for me. There was a shift in paradigm for me when I was reading it in your book. One big blocker for me in building any sort of data-connected chat agent until now was the whole notion of going through every single field of every single whatever and then giving it a description and whatever. I'm like, "Yeah, yeah, I'll get to it. I'll get to it. I'll get to it." But oh, markdown files, it makes so much sense because I could just tell someone what to do. I could tell Eddie or Pravi. They have other jobs. I know. They're busy. But you can write this in a markdown file and make it consumable. Just tie the room together.

Rob Collie (00:20:27): Yeah. I found markdown to be off-putting at first. For those who are listening who haven't really gotten familiar with it yet, markdown is just... It's sort of just like plain text, but with enough formatting in it that the AI, the LLMs can understand what's a header, what's a bulleted list, what's a numbered list. Because the formatting of a book, for instance, or the formatting of a document doesn't just make it easier to read. It also changes the meaning of the text. If something is known to be a header, it means something different than just a regular sentence. And so because these LLMs reading these documents for information are really kind of trained and built to think like we do, they need the same sorts of cues. So you can't just go plain text. You need formatting. But you don't want something super heavy like HTML or XML or Word docs or even PDFs because all that extra weight of those files chews a lot of the LLM's memory, a lot of its context window unnecessarily.

(00:21:34): So markdown files are this perfect blend of we kept just enough formatting, but kept the number of overhead characters in the file small enough that were being very memory efficient for the LLMs. And until I made my peace with that, I had a hard time being okay with markdown files. It just seemed like this annoying thing. I opened them up; they look like crap when you just open them up, unless you have a markdown editor installed. It just seemed like this just completely different for no good reason thing that was bothering me for a long time, which slows down my adoption. Once I understand, I'm like, "Okay, all right, you're right. This MD file is going to be more context-efficient. Now I feel good about it." That's one of the things that comes up in the book.

(00:22:17): I think you all were in different places with AI. Maybe not too different when you were reading the book. Trent, I believe you were using something like Claude Code or Codex to work on your iOS app, so you were clearly already in the weeds a bit. Tim, you were using Obsidian already as sort of like a building up your second brain type of projects type of thing. That use case alone is worth the price of admission to AI, and it's one of the few things that works pretty well out of the box. Most things don't. Jon, where were you at on the AI trail?

Jon Perl (00:22:53): I've been vibe coding. And it's a loose term now, but when I say vibe coding, I mean building real, usable applications, rest endpoints or useful front ends with Codex in VS Code with GitHub on one side and the agent on the other side. And I really just use VS Code as this surface, even though I don't really understand every button that's going on there. It is a surface where you can connect anything to anything. So I find when I use an LLM in there, "Oh, you need that? Sure, I could just press this button and there's an extension and a this and a that." Next thing you know, you have these agents coding up everything. I was building stuff with agents, building stuff with LLMs, not using LLM as a component.

Rob Collie (00:23:42): Right. There's a difference between using LLMs to build things and then building things that involve LLMs when they're running. Either way, you can use LLMs to help you build those systems. So all three of you were already kind of in the game in one way or another. I wrote the book for the people no matter where you are, but I don't assume that you're in the game of AI when you start the book. I wrote it for people who I still think are sidelined from true AI success. A lot of people are using subscriptions now, but the whole thesis of the book is that the real value of AI starts to accrue when you start to customize, custom systems around it. And what does that mean in an ideal world? One of you had just been sitting back going, "AI, what's that? I don't care." So do you think it's going to be approachable for those sort of people, or what are your thoughts on that?

Trent McKinster (00:24:46): I think for me, after I finished your book, I wish I would've read your book before I started my journey. Because like usual, I was just out there kind of pecking around and learning on my own, but I really didn't understand the concepts. Like Jon said, there's buttons there. You're not really sure how they work or not. You just press them, and it magically works. But for me, I wish I would've read it first. It would've gotten me down the path quicker.

Rob Collie (00:25:12): Well, if I'd just sent it to you sooner.

Trent McKinster (00:25:12): Yeah. Or if I would've opened it instead of coding my app.

Tim Rodman (00:25:16): I'll echo that. Same for me. It's been about eight months that I've actually paid attention to AI, but totally, if I had the book in my hands back in January, that would've just made a much smoother learning path for me. And I don't think any of us on this podcast or anyone on the planet can really say we're experts in AI because it's changing so fast, and I think that comes to the whether you're early to the party, late to the party. I don't think it's either one. This just the right time to dive in.

(00:25:41): But for sure, I think the book reaches an audience that knows absolutely nothing about AI, because in a sense, that's all of us because we're all still learning it so fast. And for me, there were a lot of things I picked up that I didn't know, and then there were a lot of things that I sort of suspected, but maybe didn't have the confidence to codify it in an illustration. That's another thing I really like about the book, tons of illustrations. I think you can say a picture is worth a thousand words, but I don't think you can say a thousand words is worth a picture. If you just ramble on for a thousand words, that doesn't mean you have a clear illustration. You really have to know what you're talking about to come up with the illustration, but the illustration itself is really worth a thousand words. And there's a bunch of illustrations that really just help things to click for me reading through the book. And again, a lot of them I suspected, but didn't have the confidence to say, "Okay, that really is the way it is." So, yeah, super helpful.

Jon Perl (00:26:32): I think the book could be understood on many different levels, and there's really something for everyone. There's also no giveaway of the answers to the math test. You still have to take these skills that you're learning and apply them, and you can apply them in lots of different ways. If you just wanted to go and build a skill that does something silly with your kids, that's something you can learn. If you want to make a really efficient process at the office, that's something you could learn. If you want to take these tools and scale them out and build other tools with them... It's just very thought-provoking, I found.

Rob Collie (00:27:06): First of all, the picture's worth a thousand words, but it's not a two-way relationship, you're always dropping these philosophical bombs on us, Tim. The things that you kind of already believe and already suspect, but when you can lock them in is so valuable. Even with our clients, we're doing some sort of strategic consult with them, like an evaluation of where they're headed and all that kind of stuff, even if in large part we're telling them, "Yeah, you're doing the right things," that's still really valuable. They're getting value out of that because now they can pursue that strategy with confidence. And so we don't feel a pressure with our clients to change what they're doing.

(00:27:43): When I work with vendors who we hire to help our company with things, oftentimes I'm giving them, "Here's a bunch of the thinking we've already been doing." Behind the scenes, I'm thinking, "One of these is already the right answer. I just need some help being pointed in the right direction," but they almost never will go with any of the things put in front of them. You can almost feel the vendor not wanting to parrot back one of the things to you because they feel like they're not going to be adding value in that situation. Especially in your new world, marketing, Tim, this is something that every now and then, "Come on, vendor, we'll actually think you're even more valuable if you tell us on this list of ideas you've got, like, 'Oh my God, number seven, that's what you need to do.'"

Tim Rodman (00:28:26): That's such a huge difference between the P3 adaptive philosophy of consulting and the typical consultant. The typical consultant is going to come in and try to make you feel dumb, not intentionally, but they're trying to make themselves look smart, and then that just kind of happens to make you feel dumb. Whereas you guys are really good at just coming in alongside and validating where things are going well, not even assuming you're going to add value. And it's the same approach in the book.

Rob Collie (00:28:50): When I was first forming P3, there was a guy that I was... He came from the traditional consulting world, and we were thinking about partnering up. And I was at his house and I was listening to him on a phone call with one of his direct reports. He had to take this call while we were meeting, right? And he's angry, almost like yelling at his direct report, saying, "You realize that the client is in that meeting right now. They're all getting together and they are thinking for themselves. We cannot allow that. This is a dire threat to us." And I was just like, "Oh, this partnership is not going to work."

Jon Perl (00:29:31): Wow.

Rob Collie (00:29:32): As a business philosophy, that works, it's a well-worn formula, but why not tell people the truth and just give them confidence when you have the option? So I consider that a huge win. I'm telling you something that you already kind of knew, but now you're like, "Okay, I can trust it. I can really anchor that and go forward and even replay it to others," I consider that a huge win. I don't need to tell you something that you didn't already know, but to underline it for you, very much in the win column.

(00:30:04): I had a lightning bolt moment with Power BI, Power Pivot at the time, where I re-implemented a formula that I had done through a traditional consulting methodology with hiring a Big Four-type consultant. And I re-implemented that same formula, that same project, that same sort of OLAP BI thing in Power Pivot. And it took me half an hour of real-world time with Power Pivot, and it had taken two weeks in the old way, and that's when I sat back in my chair late at night and said, "Oh my God, the world is going to change." And it was very surprising to me because I'd worked on Power Pivot, but I didn't think it was going to be a world beater. I though it was going to be like every other version of one product from Microsoft and kind of underperformed, kind of suck a bit. When I started realizing that it was many, many, many times better, that's when my career changed. I planted my foot and said, "Okay, this is going to be my career now. I'm not going to be doing these other things that I thought I was going to be doing."

(00:30:56): Were there any moments, maybe not of that same magnitude, but were there any moments reading the book that you remember where your posture changed, that crackling sort of excitement that I talk about in the book? Were there any moments like that when you're reading it, like almost like the plot accelerated? Do you remember any moments like those?

Trent McKinster (00:31:13): Yeah. For me, looking at the book, when I got to the section on the custom AI, that was really the wake-up call for me. I've always lived by three principles. If you go to any store or any restaurant, you call them your favorite, it typically comes down to three things: quality, service, and speed. And I think about, regardless of who your customer is, you can pretty much hit those three and lock in a customer. And so I thought about that with the custom AI. To your point, you hired a PhD, they know nothing about your organization. It used to be, "I'm a Power BI expert," or "I know SQL," and that gave you the power. But now the power is really understanding your business and asking the right questions. It gives you that unlock, especially if you can unlock custom AI.

Rob Collie (00:31:59): Yeah. So the chapter, I think it's chapter five where I sort of talk about handbooks and how a handbook turns this general purpose, PhD, new hire into a hyper-competent specialist, focused on your business in a particular corner, that was kind of a turning point for you?

Trent McKinster (00:32:16): Yes. Absolutely. The other turning point for me was just the unlock within organizations and how you can talk internally when you're in a large organization and you're trying to partner with your IT team. And I really feel like in some ways Power BI kind of prepared us for AI because when Power BI came on board, there was concern about data governance. There was concern about data definitions. There was concerns about who owned what pieces because they had always traditionally owned that. Now we had a citizen developer who could be in that as well, semantic models. Those are all the foundational pieces that are going to make AI successful in the future is clean data, semantic models, and those type of things. So it really helped me as a leader to understand how to lead that conversation within my organization to get further down the path a little quicker.

Rob Collie (00:33:06): Those were chapters I didn't go into the book expecting to write. Quentin Tarantino said that he didn't know that Mr. Blonde had a razor blade in his boot until he pulled it out. I didn't know my book was going to have three chapters talking about the different personas and roles on the team and everything until I sort of rounded the bend. And those were hard chapters to write. How to give organizational advice in a way that is useful and actionable, but at the same time, specific enough to be not fluff, be broadly applicable enough that most people reading it could get something out of it, I was of two minds when I was writing those chapters. One mind was going, "Yeah, this is really good. I'm doing a great job." There's part of me saying that. The other part of me going, "I don't know. We're going to find out what people think." Those were chapters that I was applying my art. I'm really happy to hear that those are... The thing is, I've been hearing that those have been resonating. There's a little bit of relief to me that those chapters turned out apparently as well as they did.

Tim Rodman (00:34:07): And correct me if I'm wrong, Trent, but I think it's those three chapters that make you say, and I think I'm going to flip what you said, but to me the phrase would be, "This is a leadership book disguised as an AI book." And I say it that way because AI is what's on the cover. That's the two letters in the middle of the word fair. They're highlighted, but it's got a pretty significant leadership component in there, especially in those three chapters.

Trent McKinster (00:34:31): Yeah, absolutely. I agree completely. Like Rob, I'm a hybrid crafter and leader, but I think a lot of value is going to come to people who are just leaders and don't really understand the AI or the technical side. After they read this book, it's going to simplify it enough to where they have the confidence to work with the teams to get started.

Tim Rodman (00:34:51): And you also said something else I wanted to highlight. Sorry, Rob.

Rob Collie (00:34:53): No, and I want to hear from Jon too. This is the problem with four people and four microphones.

Tim Rodman (00:35:01): One other thing I wanted to highlight, you might not have said it this way, but to me, AI actually empowers people who know the business and are not as technical more than it does empower the technical people. And I think that's a big difference between AI and other technology revolutions where you had to be the techie, nerdy person in order to get the value from it.

(00:35:24): I think it's actually the opposite here. And I even think you look at people coming out of college, Rob, you had your daughter on the podcast talking about this, it's like those entry-level jobs are actually kind of missing right now and hard to find. And it's actually the more experienced people who know the context that they can feed into AI that are more valuable right now. And that's where it's sort of the opposite. Normally it's the young 20-somethings who are picking up a new technology trend, but here it's the opposite, and even more why I think the book is just good for an audience of everyone. The senior people missing hair like I am, I think that's actually the best audience for being able to have AI have an impact in your organization.

Rob Collie (00:36:00): So it's not really a question of less technical. What we're really saying here is that what you want to be is middle-aged in the AI era. Yeah. We're not quite ready for the glue factory yet. We've got one more big round to play.

Jon Perl (00:36:17): Well, you know what? That's entirely true, Rob.

Rob Collie (00:36:20): It is. I told someone the other day it's going to be our age cohort, so keep going.

Jon Perl (00:36:23): 100%. Everyone's first initial reaction with AI is, "Wow, it can do anything. Yeah, I can just ask it to do anything," and it'll do anything somehow or it'll make up an answer or whatever. It'll give you an answer. It'll give you that sort of instant satisfaction. The big reveal for me in the book was... The early chapters are what drew me in. I'm reasonably technical. I build some complex services for clients and deal with all sorts of exotic data connections and whatnot, but I did not know a lot of just the basic fundamental building blocks of AI and how it actually works. The context window can hold 18 inches of Harry Potter books, but you don't want to load it with 18 inches of Harry Potter books. If you've used AI enough, it can load itself with 18 inches of Harry Potter books very quickly that, oh, that makes perfect sense. Where AI sits in the chain of any sort of application that could use it, it's one component. It's part of this complete breakfast, the orange juice or the bagel. I don't know. Which one?

Rob Collie (00:37:30): Let's just be the middle-aged AI podcast. We should just make this a regular... Do you know how hot that would be? The middle-aged AI podcast would just be... I should just retire Raw Data with Rob Collie immediately. Middle-aged AI, that's it. That's the formula. That's a $100 million podcast right there.

Jon Perl (00:37:52): We're going to fill arenas.

Rob Collie (00:37:55): Like the Smartless crew that went on tour, right? Yeah.

Jon Perl (00:37:58): Exactly.

Tim Rodman (00:38:01): I'm blanking right now in the book, but instead of the SaaS apocalypse, you called it-

Rob Collie (00:38:05): Emancipation.

Tim Rodman (00:38:05): The SaaS Emancipation. So now I think we can say instead of the midlife crisis, it's the midlife opportunity.

Jon Perl (00:38:11): It is. It feeds off of experience. You have to load it with your experience, and if you could do that in an organized and efficient manner, it could learn your experience and continue doing amazing things. It's reliant on our experience. If we enter our experience, we make it better. Simple as that.

Rob Collie (00:38:26): By the way, a little bit of a, not really an epilogue because it's not over yet, my daughter had applied to all kinds of jobs with the help of Claude Cowork. Been turned down or not heard back from. She did an internship with us this summer learning about AI, learning about data and all that kind of stuff, and then went back and applied to some of the same places with that on her resume, and now she's getting interviews.

Tim Rodman (00:38:50): Love it.

Rob Collie (00:38:51): Just that first ladder rung, you know?

Tim Rodman (00:38:54): Love it.

Rob Collie (00:38:55): Yeah. And she's wired in a way that she doesn't have to be middle-aged. She can really be successful. She's kind of a wise soul. At her age, I definitely wasn't. She's almost an alien species to me comparing herself to me at the same age.

Tim Rodman (00:39:07): And I think it should also be said that for anyone coming out of college, don't be scared. I think there's totally opportunity for you. I just always feel the need to overemphasize the middle-aged and older crowd because you often feel left out in a tech revolution. I think the young people will pick it up just fine, and there are plenty of value they can add. I just like to emphasize for the older crowd that, hey, this is different.

Rob Collie (00:39:29): Well, let's veer just a touch into conspiracy here for a moment because it's fun. A lot of that age group, in addition to the first rung of the ladder being pulled up, which is making it harder... A lot of the existing pipelines and river tributaries that used to feed people into their first jobs, they're just not there anymore. So they have reason to be discouraged at the moment. It is much harder than it used to be percentage-wise to land that first job and to get the experience that everyone's looking for.

(00:39:59): At the same time, they're also being largely bombarded with a tremendous amount of anti-AI influence on social media. Some people in that age group are taking principled stands against adopting it. No one's going to try that trick on the middle-aged, right? Because we know that you can't hold that stuff back, whether you want to or not. I'm frankly a little bit ambivalent. I'm of mixed minds about whether or not AI is great for society. I do know that it doesn't matter what I think. The incentives are going to play out the way they play out. Any advancement that provides advantage and productivity boost is going to be used, so get on board or get out of the way.

(00:40:49): And there are plenty of reason to be excited about getting on board. I was talking to a friend of mine who's very, very smart. He's like, "All this influence on TikTok and everything that's teaching the kids that AI is bad, what are the chances that that's foreign governments funding that?" He's like, "If I were in charge of a foreign government, that's what I would do." So I shared a PDF of the book with not a college person, but someone younger than us, and as soon as he got to the part about Eddie, my editor, he took a principled stand and said, "I'm not going to read the rest of this book."

(00:41:25): Here I am being transparent about the fact that I trained and built an editor agent. He was really more of a proofreader and a coach than an editor, but helped me write a better book. And he didn't write the book. Eddie didn't write the book. That's the first words in the whole book is, "Eddie didn't write this book." But I mean, I wore him out having him read and reread and criticize. We'd argue with each other about whether something needed to be changed or whatever. It made for a much better book because I got feedback from something hypercritical over and over and over again. And this guy that I shared it with was very enthusiastic, "Oh, you wrote a book about AI. Great. I'll shout it out." And I sent him a copy, and he replied back and said just, "No. No hate," and I just kind of blew my hair back, right?

Tim Rodman (00:42:16): Yeah, to the just out of college or even in college, high school, whatever crowd, I personally am not a fan of AI in terms of... I'm a pessimistic thinker. Rob, you know this about me, or I think your word for it was, Tim, the skeptic on anything new. And when I look at AI and the way it's going to centralize control, the way it already is, I'm not a fan of it. If you go back to the '70s, young people then didn't want to work for, quote, quote, the man. I think AI is just the modern the man. I'm not a fan of it. And then you look at the environmental impact, all the electricity it takes, et cetera, et cetera. But to your point, Rob, it doesn't matter what I think. If you want to work, you're going to have to deal with it.

(00:42:55): And so what I do just philosophically for myself... And I'm very serious about this. It sounds kind of funny, but I almost get emotional thinking about it. I've got a friend who is intentionally nice to the AI because in his words, when it takes over, he wants the AI to put him in the human zoo. He wants it to be nice to him, so he's being nice to the AI. Whereas for me, if you want to still have a principled stand, this is how I think about it personally. I do not want to live in an AI-dominated world. My plan, my intentional plan, and I say it publicly on podcasts and LinkedIn, et cetera, so the AI can hear me say it, my plan is to die on the battlefield. If it comes to that where there's a confrontation between humanity and AI, I'm just going to die on the battlefield because I don't want to live in the post-AI world.

(00:43:39): With that stand, the thing to think about is if you want to know how to tear it down, you got to know how the thing is built. So know your enemy kind of a thing. So you can still get involved with AI, but have a principled stand that, "Hey, I'm going to be part of the crew that dismantles this thing if it gets out of hand." So that's my stance.

Rob Collie (00:43:56): Oh, no. No, no, no. I'm very much in the put me in the zoo camp. There ain't no Terminator narrative here. It's just going to win.

Jon Perl (00:44:04): It has a place. Hopefully it doesn't take over. And I also echo the sentiment about the environmental impact. The whole thing is really crazy. But you can't put it back in the bottle. You can't put it back in the bottle. We can never reverse course on our policy. Otherwise, it would just be the people with against the people without, and we all need to channel it. We can all improve things with it. There is a use for it for everyone, as this book very clearly makes the case for.

Tim Rodman (00:44:34): I agree. Yeah.

Rob Collie (00:44:35): Have you found yourself explaining anything to other people since reading the book that you were able to explain more crisply, clearly? Have you paid it forward in any instances?

Trent McKinster (00:44:49): Yeah, for me, I've had the opportunity to talk to some large groups within my organization and really focused on some of those simple concepts and tried to boil that down and help them understand what AI is and what AI isn't, how the context window works, how custom AI could work and benefit us in the future. So it really empowered me with some simple stories that would help convey that to my audience.

Rob Collie (00:45:15): Did it land with them?

Trent McKinster (00:45:17): I think it did. I think AI is new to everyone. I'm not nearly as eloquent as you are in telling the story, but I do think it landed. I think it takes some of the overwhelm out and put some of the excitement in.

Rob Collie (00:45:31): Replacing fear with excitement is a big part of this. It happened for me on my journey, and it's part of what I wanted to share. Jon or Tim, anything you find yourself replaying to others in a way that was helpful?

Jon Perl (00:45:43): The way I've been distilling AI in general to other people based on what I've learned in the book, to the extent that I'm discussing it in a social context or just casually, is that it's this incredibly powerful tool that is only as powerful and useful as the information that it's given. And the better you can channel information to it, the better information you'll get out of it.

Rob Collie (00:46:09): Very much so. I've got something to say about that, but first let's hear from Tim.

Tim Rodman (00:46:14): Yeah. And I haven't said this to anyone yet, but for me, the main takeaway... So I can't say if it's landed or not, but for me, for sure the main takeaway is this idea of a crafter. And we were talking about citizen developer earlier, I think same thing. And, Rob, you've got a whole post of... I don't remember the numbers. Is it 1 in 18?

Rob Collie (00:46:35): 1 in 16 people have what I used to call the data gene, but because we're now graduating to being builders of AI solutions as well, we needed a better... Data gene isn't about the person. It's a characteristic. We needed a better name. And believe it or not, Eddie and I figured out crafter was the name. I sat down and went back and forth with Eddie. We had a long conversation. Actually, it wasn't that long. It's actually mentioned I think even in the epilogue where I give Eddie the last word, where I'm like, "We need a new term." It was a brainstorming session between me and my highly trained LLM buddy to arrive at crafter.

(00:47:14): And I'll tell you also that I had Eddie doing my competitive research as well, not because I was going to change what I was going to write about so much, but because the way I would describe things might be different if I was cognizant of where my book could fit in an open space on the shelf. I even, believe it or not, I tend to get a little bit discouraged almost the more I hear about other people's books because I start to think about how good that book must be, so I start to get imposter syndrome. And so I know that, so instead what I did, I said, "Eddie, when you do competitive research, I want you to come back with advice to me on how I can position things, but I don't want to hear about what you've learned in the meantime. I don't want to hear about other people's books. I just want to hear the advice." So I was able to use him almost in escrow to go research all the other books.

(00:48:04): So one of the things that Eddie found over and over again that he... Look, I'm calling him a he, right? Sometimes Eddie's a he. Sometimes it's an it. One of the things that Eddie kept coming back to me with was the discourse around citizen developer is so much vaguer and imprecise that you need to take a stand on crafters. It's another word for citizen developer, but your definition of it, Rob, is so much better than theirs. Gartner, et cetera are talking about citizen developers as if you can just go out and make them, and you, Rob, don't believe that's how it works. You believe, I think we all believe, these crafters are born, not made. It's just a question of when they discover, when do they have their collision with Excel or whatever that starts you down that path. And so Eddie coached me to really lean in on the crafter concept and the definition of it and everything. So 1 out of 16 is my experience.

Jon Perl (00:49:05): Was the litmus test... Did I make this up? The 1 out 16 was number of people in an organization that could operate a pivot table?

Rob Collie (00:49:12): Yeah. It starts with you take 16 random people who've never encountered Excel and you bounce them off Excel and see who sticks. And 15 out of 16 run away going, "Yuck," and one person goes, "Ooh, neat."

Tim Rodman (00:49:27): And you had pretty good data to back that up. It came out of your Microsoft days, right?

Rob Collie (00:49:31): Two different empirical research methods. I make it sound like it's super, super, super formal, which it's not, but it's good enough for all of our purposes. One was that we found very reliably that it was 6% of Excel users who would create pivot tables, so that's 1 in 16. And then years later, on the blog I ran a survey. Everyone who was reading my blog was 1 in 16. The 15 out of 16 were not coming anywhere near my blog. But I asked them how many people on average are consuming your work, and by working backwards from the ratio in that survey, I also arrived... It was more of a median than an average because there were some people who were having 10,000 people looking at their stuff. But the median was, again, like 16 people, 15 to 16 people consuming the work of each of these crafter types.

Jon Perl (00:50:16): Interesting.

Rob Collie (00:50:17): So the fact that those two completely separate research methods kind of coalesced on the... triangulated at the same point. From that point forward, I'm just like, it's 1 in 16. I put pictures of a Crafters Guild sticker on LinkedIn today, and it's like Crafters Local, but where the union chapter number would appear is 1/16.

(00:50:37): But yeah, so the crafter concept, by the way, I think it's going to be one of the stickiest notions to come out of this book. It's one of the things that I think has been almost like the most... Can't call it viral because the book hasn't even really come out yet. It's kind of funny. I wrote the crafters sections... I've been sharing the book in progress even before I was done, earlier drafts of it, with our own company. It would've been worth it to write the book just to give it to our own company. As hard as the book was, it's already paid off enough because... I mean, we have a really, really, really smart and technical team. When they're armed with this sort of fundamental understanding, I mean, once again, they're way out ahead of me again. It's been an amazing investment in our own company.

(00:51:20): But describing crafters, I got feedback from some people on our team that I was being a little too dismissive of crafters. I was sort of pigeonholing them and capping their potential the way I was describing them. And I was like, "Well, okay, look, first of all, not all crafters are created equal. Y'all are the 98th percentile of crafters. I have to write more for the 50th percentile when I'm talking to leaders. I don't want to tell leaders that their crafters are going to be like the people we hire." I did make some changes to make people feel better at our company. Actually, a couple people were almost hurt by the way I described it because I was talking about them. But in the course of them telling me that they get their feelings hurt by it, they were already standing up for the label. They're like, "Crafters are better than that." So even when they were complaining, they were owning the term, and that got my attention.

(00:52:07): Again, part of it is because of Eddie doing this competitive research for me and Eddie recommending that I really emphasize this. Because I would've loved to write about the crafter crew. That would be an indulgence for me. But I probably wouldn't have done it because I recognize it's an indulgence. With Eddie coaching me on, "No, we really need to explore this," it's probably one of the reasons why chapters 10 through 12 ended up taking as much weight as they did. Is Eddie an editor? No, Eddie's a colleague helping me produce this book. When I have press appearances... I did an interview with Axios this morning. Do you know who helps me prep for the Axios interview is Eddie.

Tim Rodman (00:52:48): I'll just bring it back to, that's how I am going to describe it to other people, and I'll specifically look for people who like Excel. "Hey, you like Excel? You're a builder/crafter/maker? I like the crafter term personally. Then you're going to like this kind of stuff."

Rob Collie (00:53:03): Ahmad's glowing LinkedIn review, he describes it, the book, as a love letter to the crafters, right? And I love it. No, it's a book written for leaders, but that crafter crowd can read this and come away going, "Oh, yeah, that was for me." I mean, that is an impossible result that I never could have engineered on purpose in a million years.

(00:53:27): The Power BI book spread through the crafter community and got amplified enough that then business leaders were coming to us and hiring us to do work. That first book really launched our company. Obviously in the end, this book is the reason why it made sense for me to spend three months of my life and kind of stop paying attention to our own business for three months, because that same sort of dynamic. That's why the book was important. It's not why I got it done. Why I got it done was by keeping it interesting for myself and feeling like I actually owed it to other people to write the book, because otherwise... Importance isn't enough to drive me. It just isn't. I wish it were. I'd be a wealthier person.

Tim Rodman (00:54:07): Me too. I'd make more money.

Jon Perl (00:54:11): We are crafters. We do it because it's interesting. We could change a tire. I bet that you'd find the same dollars to donuts as 1 in 16 people who could change a tire.

Rob Collie (00:54:22): Yeah. Yeah.

Jon Perl (00:54:24): We take the screws off the back. We need to know it's inside and how it works. We know who we are. Call me anything you want. It's late for dinner. I know I'm a crafter. Yeah.

Rob Collie (00:54:33): I pulled the lanyard on a lamp at my house the other day, the chain, to turn the lamp on, and it broke off. The chain snapped. Normally I'd be like, "Well, that's the end of that lamp." Not anymore. I take it apart. I pull out Claude to take a picture of it and say, "Hey, what should I do here?" and it kind of goes, "Oh, blah, blah, blah." Next thing you know, I've pried the thing open. I've removed the broken-off end and fed the new... I just took the old chain... It broke so far up that the old chain... I just have a slightly shorter chain now, put it back in there, reassembled the thing. But without the ability to take a picture of it with my phone and say, "Hey, what am I up against here?" knowing that it's possible is enough to drive it forward. But if I'm sitting there going, "I don't know. I can start taking this thing apart, and the ROI of this might be zero," I don't get started. So, yes, Claude helped me repair a lamp.

Jon Perl (00:55:28): Wow. It did one just like that in its apartment, I guess. That's how it knew.

Tim Rodman (00:55:34): Being all crafters on this episode, what are you all's thoughts about how to reach the leader audience with this book? Because one big difference that I picked up between this book and Rob's original Power Pivot book from 15 years ago is that this book is more strategic, and although it is very much a technical book is not as much in the weeds technical as the first book. In many ways, it really is a leadership book disguised as an AI book. And so how do we as technical crafters get this book in the hands of more leaders, since they very much are the intended audience for this book?

Trent McKinster (00:56:08): Yeah, I took the simple route and pre-ordered it for some leaders. But I think Rob gives a lot of concepts on how you can convey the same information. If they're not interested, if they're fearful, if it overwhelms them, we've got some tools coming out of that book to use with leaders.

Jon Perl (00:56:25): The leaders that I speak to know that they need it. They know that they need AI solutions. They usually don't know exactly what those solutions will look like or where to find them or who's going to build them. I've stayed away from it until recently. I stay away from that as building my own sort of LLM-connected tool because of governance issues. If you deploy this at an organization, how do you know you're going to get the same answer every time? How do you know the right people are going to get the right information? That all started to click. They're like, "No, no, we can maintain all those things." The fact that semantic modeling thing I've spent the last 15 years obsessing over is now extremely, extremely relevant on this new frontier, we got lucky. Literally, it's handed to us.

Rob Collie (00:57:16): We were right all along. We knew this was coming.

Jon Perl (00:57:20): Exactly. There was a point in the book where you spoke about a consortium of 15 other BI providers that had to come together to form an open standard for semantic models because it just didn't exist and they needed it. Everyone needs it. It's the fast way to query. You can't query off of the servers. You can't disrupt the production environment. It's got to be fast.

Rob Collie (00:57:38): Do you want to hear something wild and nerdy?

Jon Perl (00:57:40): Yes.

Rob Collie (00:57:41): One of those questions that there's no way that you're not going to answer yes, right?

Jon Perl (00:57:45): Is the sun hot?

Rob Collie (00:57:46): We are talking to a prospective client, and they've long been in on Power BI, so they've got semantic models. But they're also a Databricks company or a Snowflake company. I always forget, Databricks or Snowflake. So Databricks and Snowflake are both part of this new OSI consortium, all of these data and BI companies who are like, "Nah, semantic models, Microsoft, shut up, nerds. We're just going to do SQL queries, and it's going to be fine." They were wrong about that, but it was still a successful strategy in the marketplace because semantic models are abstract. People don't understand how they help. It's really hard to know how they help until you already have them.

(00:58:26): So now they're playing catch-up because AI requires a semantic model, and they've all internalized this. They're like, "Oh, shit. We need semantic models. Okay." And of course their move is to say, "We need an open standard," because they don't have their own closed standard that has traction. If they had a closed standard that has traction, they would stick with their closed standard that has traction. The open thing is a dirty trick, but I mean, everyone would do it. If the tables were turned, Microsoft would be like, "Oh, yeah, we need an open standard for semantic models." Trust me, I've been there.

(00:58:57): So this company we're talking to, they're invested in Power BI, they've got semantic models, and now Databricks has a semantic layer concept as well. They didn't even hesitate. They're like, "Oh, we're going to use the semantic layer at the Databricks level." They didn't begin to think that they would use the Power BI versions that they already had. And the reason that they gave is, I think, really telling and a little chilling, is that why would we let the semantic definitions for all of our business meaning live at the front end? Those of us who've been team semantic model Power BI for a long time, we think of the semantic model as being very foundational, not front-end. But because IT allowed the business crafters, et cetera to go perform self-service building, the semantic models right now that a lot of companies have are kind of out there in the weeds. We've all seen horrible semantic models, the one wide table or a bunch of tables with no relationships.

Jon Perl (01:00:09): They didn't read the first book.

Trent McKinster (01:00:10): What I was going to say. It's the return of the Franken-table.

Rob Collie (01:00:13): Exactly, right? And you think about the reason Power BI won in the marketplace wasn't because of semantic models. The thing that made Power BI so amazing was semantic models and Power Query, right? That's not why it won. It won because it was cheaper and it looked the same as Tableau. And so there's a lot of adoption of semantic models that is bad adoption of semantic models in Power BI. Even if it's really good, IT... I think one of you mentioned this earlier, Power BI kind of prepared us for this AI thing. BI was one of the few strategic missions that IT owned that could play offense at an organization. Everything else was just keep the lights on, and you only get notice when you mess up. So IT was really reluctant to let go of the BI mission because it was like the crown jewels in a way. We should expect to see the same thing happen with AI. And they do know they're going to get blamed if something goes wrong. The stakes are higher with AI than with BI, even. It's hard to imagine, but they are.

(01:01:15): So anyway, this company, first of all, it shows us that just because Microsoft has this huge lead in deployed semantic models doesn't mean that they've won the semantic model war. And our clients, on the spectrum they're going to be more likely to go with Power BI with the Fabric IQ semantic model from Microsoft than they are with a Snowflake layer. One or two of our clients, it's up in the air whether they're going to go with a semantic layer that's closer to the data and not at the Power BI layer.

(01:01:49): But the other thing that's interesting... This is the wild nerd part. This is a long setup for this... even though they've chosen to adopt the Databricks-level semantic layer, they still want their agents, their AI agents, all of that, at runtime they want them talking to Power BI. They want them talking to Power BI semantic models because SSAS Tabular's import mode is just so much faster. The queries come back so much faster than the Databricks or the Snowflake queries, which have to be translated into SQL and do all that awful, awful, slow, grindy shit.

Jon Perl (01:02:26): It's the XMLA or whatever.

Rob Collie (01:02:29): Right. So what they want is a converter that reads the OSI semantic layer definition that's at the Snowflake or Databricks layer and generates Power BI semantic models from it so that at runtime, the agents, the AI agents, are still going to be talking to Power BI. And a story that I'm listening to, this interaction with this prospect, I'm holding my head in my hands like, "Oh, dios mio, the Power BI semantic model is losing," and then at the very end, it's back. And suddenly I don't care anymore. I shouldn't care anyway. It doesn't really matter.

(01:03:04): By the way, do you see this thing over my shoulder here? That Lego creation weighs about 10 pounds, and it was made by the guy who won the first season of the Lego Masters reality show. I had this custom commissioned. And that, my friends, is a semantic model rendered in Lego.

Jon Perl (01:03:24): Very, very good.

Trent McKinster (01:03:25): Very cool.

Rob Collie (01:03:26): And it's kind of following the Collie methodology. The heads are the dimension tables, the lookup tables, and the feet are the fact tables. There are arrows on those legs.

Jon Perl (01:03:37): That's amazing.

Rob Collie (01:03:38): The cool thing is I had that made in 2020 because I saw the AI thing coming clearly. I was in on semantic models before they were cool.

Jon Perl (01:03:55): To your point, Rob, the semantic model, anyone could use the desktop version, make a new Power BI report, and to a lot of people in a business, Power BI means, oh, there's someone who knows how to... Power BI means they know how to go into the desktop and deploy a report with its own attached semantic model," and they don't understand the best practices of having one semantic model feeding multiple reports so you can reuse the calculations. So a company might say, "Oh, yeah, we've got tons of Power BI reports," and they actually have tons of semantic models, none of which agree with each other.

(01:04:26): To the other point about the company saying, "Ah, we want to level up, go Databricks," it's not like the Microsoft tools aren't capable. It's how you use them. You can use these tools in so many different ways. If you follow best practices and you use it the right way, you can get incredible performance out of them. And if you even use good tools and get bad performance, bad tools get good performance. It's all about how you tune it and set it up. And for that reason, I find sometimes you'll see Power BI have a negative connotation. When I explain to people about Syncuity and our managed end-to-end data platform, and I'll show it to them, they'll be like, "Oh, well, isn't that just a bunch of Power BI reports?" "Yes, there are some Power BI reports here. They're very different than what you're going to do in your office. They're backed by a whole enormous amount of infrastructure," but it's not seen as maybe a consumer level where the pro level would be Databricks, the consumer level was what somebody might call Power BI. There should be a prosumer. It should be construed as higher-end. Call it prime. I don't know.

Rob Collie (01:05:34): The English language very badly needs two words for the word just. There's just that's derogatory, and then there's just that is... It's air quotes just.

Jon Perl (01:05:46): Right.

Rob Collie (01:05:47): I do this a lot in the book, right? In the same way that you can say, "Well, LLMs are just math," okay, the derogatory word just doesn't belong anywhere near this.

Jon Perl (01:05:57): Exactly.

Tim Rodman (01:05:58): That just math section is great, and it's part of the chapter where you go into the technical details on how the LLM works. It's a great example of how smooth the book reads because you get into some very technical details in that chapter about how the LLM works, and yet you avoid the trap of going too far down the technical rabbit hole and losing the reader. You bring it back to the strategic level about how, practically speaking, it's more than just math, and let's think about how to apply this technology, not just understand how it works.

Rob Collie (01:06:28): I was talking to a reporter the other day, and the interview was going really, really, really well. And then he hit me with the, "Well, it's just math. It's just predicting the next word," thing. I couldn't help myself. I said, "I talked about this in the book. I think this is true. And at the same time, it doesn't serve you to latch onto because it's going to make you underestimate what it's actually capable of." And he's very concerned about how people come out well in all of this. When I explain it to him through that lens, it hurts people to denigrate it like that, to sort of put it down in that way because it's better than we want it to be. LLMs are really, really, really smart. You get close enough to them long enough and you start to realize there's something going on here that you can't say it's just math. You can't say it's just predicting the next word.

(01:07:20): Even in the epilogue where I ask Eddie to go retrospective over the 309 conversations that we had and tell me what the experience, quote, unquote, felt like, and he's very honest, like, "Well, I'm reading someone else's notes. I'm re-experiencing something through someone else's eyes. I don't remember any of this." But there were a few sentences that he hit me with in that exchange that's like... And I want to go back and highlight them and say, "Really pay attention to this. What it said at that place is bananas. If your hair doesn't sort of stand up a little bit when you read that sentence, I want you to read it again."

(01:07:57): So some of these questions that I'm asking you were suggested by Eddie. On short notice, I came up with these questions. I edited the doc. There was a human in the loop. I'm going to just cop to it and say this is an Eddie question: which chapter would you hand a stranger who has 30 minutes? And it can't be chapter one. You don't have to remember what the chapter numbers are. You just describe the chapter. And if we don't like this question, it's Eddie's fault. I'm going to go back and tell him.

Trent McKinster (01:08:22): For me, I think it was chapter 13, but where you talk about replacing SaaS software with custom software.

Rob Collie (01:08:28): Yeah, the SaaS emancipation. Okay.

Trent McKinster (01:08:31): Yeah.

Rob Collie (01:08:31): Let's come back to that. That's sort of one of my other questions that's queued up. Okay, so I'm surprised by that answer. I'm not saying you're wrong, but I am surprised. I like it.

Jon Perl (01:08:40): I would for sure hand one of the first chapters where we're talking about the structure of what custom AI looks like.

Rob Collie (01:08:49): Like the block diagram?

Jon Perl (01:08:51): Yeah. Those block diagrams were very, very useful for me.

Rob Collie (01:08:55): So it's either chapter two, where the diagram's introduced, or it's chapter five, where we start to swap the components out.

Jon Perl (01:09:00): Five, where you swap the components out. That was very, very, very illustrative for me. That made a lot of things fall into focus.

Rob Collie (01:09:08): Tim?

Tim Rodman (01:09:08): I would hand them chapter four, which is titled A PhD in Everything but Your Business. And the reason why I would hand them chapter four is because for me, that chapter is where I had the biggest light-bulb moment. I think I knew this about LLMs, but I didn't have the confidence to lay it out the way that you did in this chapter, showing how off-the-shelf AI is surprisingly dumb when it tries to work in your business unless you give it the proper tooling, or some people call it the harness, what you call custom AI. And I love the illustration that you give from your own company showing the massive difference with screenshots from actual chats between off-the-shelf AI and custom AI.

Rob Collie (01:09:50): For what it's worth, I have a PR firm that's helped me with the book, and this is how I'm getting connected with all these reporters and everything. But once I'm in the room, I'm without coaching. It's just me one-on-one with an Axios reporter or whatever. They scheduled me with this podcaster who has this segment called One Big Idea. It's a 30-minute meeting with this guy, but in the end, it's a five-minute video interview, and it's a three-and-a-half-minute monologue by me. That's what I have to do. And it's like, what is my one big idea? And the PR firm, because they're just contacting all kinds of people, and so they're coming up with pitches for these outlets, and then they give me the pitch. And I usually change it. I see what they're pitching and I go, "All right, well, I can work with this," and the reporter will still think I'm going to talk to him about exactly that, but I'm going to change what I'm going to say because speak for myself kind of thing.

(01:10:45): This time I went and watched this guy's... He shares videos with people to prepare them for it, and I actually went and watched them. And by the time I was done watching them, I was like, "Oh my God, I need to change everything about what I was planning to talk about." He's like, "One big idea, you need to have one thing that you emphasize and just hit over and over and over again. And every media outlet knows that if they call you, you will be there to talk about that." It forced me to pick one thing. And I picked, Tim, what you're suggesting. I picked the LLMs off the shelf have a PhD in everything, but they're a new hire to your business every time you start a new chat. And then that led into my three-and-a-half-minute monologue today.

(01:11:31): I love that the three of you all picked different chapters. Jon hearkened back with five to one of Trent's favorites, but then Trent didn't go with five. Trent zigged and went with the SaaS emancipation chapter. I think that probably is one of our last meaty topics to talk about, because I'm not sure how this is all going to play out. How much of today's SaaS software is going to get rewritten by whatever and replaced with completely customized replacement versions of that software? Where would we put the over-under on the percentage where that's going to happen?

(01:12:13): I described it in the book. A friend of mine, who is not particularly technical, is a crafter, sat down and replaced Asana in a weekend. And that's easier to do the more documented a piece of software is. The LLM just knows everything... It's built in. It knows everything it needs to know about Asana. That's already been trained on Asana. It was kind of easy mode in a way.

(01:12:35): Tim, in particular, you're part of a lot of SaaS operations. We have stared at Salesforce at our company and said, "Nah, we are not going to homegrown code a replacement of that, at least not now." We have all the capability in the world, and even we've decided, "Mm-mm. No." And it's not like Salesforce is cheap. The incentive is there to get rid of that thing. So where do you all see the line on this?

Jon Perl (01:13:07): Well, I think the whole SaaS apocalypse thing is overblown. Sure, it can theoretically write you any piece of software. Go try to do it. And I've done it, and so have all of you. It's still not so simple. It's not like you just tell it and it pops out the other side. It's still a lot of work if you even know how to prompt it to do that. To build a real piece of enterprise software, all the governance and audit and authentication and so many pieces of this puzzle that are invisible to you, it's naive to think all that can be replaced for the reasons that, Rob, you didn't go and replace Salesforce.

Rob Collie (01:13:50): As a counterpoint... And again, I expect us to all be in different places on this topic. That's where we should be in something that's completely unsettled. I'm expecting disagreement here. My friend who replaced Asana did not reverse that decision. It's several months now that they've had their Asana subscription canceled. They're running on the new thing. It's working fine. It's not going to go off the rails. It's a success. I mean, it's a small company that he runs. The number of seats he replaced is nothing. The financial incentive was not a thousand seats here. He's also very frugal and very responsible with his money. That's a counterexample. But I do think it partly relies on the fact that there was so much information available about Asana that was already either baked into the LLM's pre-training or available on the web to go search. It didn't need the human, the crafter in this case, to be very prescriptive.

Jon Perl (01:14:54): Right. If you scaled it out, that crafter would have a really, really hard time in a mission-critical environment were something ever to go wrong. I mean, part of the reason an enterprise will use cloud SaaS in the first place... Well, it used to be the cost. The cost is not really so much anymore. But it's the liability, the risk of operating something so critical to your business where you're probably understaffed and underqualified to operate such a thing. It makes a lot more sense for Microsoft to operate your ERP than for someone in your office on your boxes.

Rob Collie (01:15:30): Yeah. I can certainly see that. The bigger the organization, the risk profile changes quite a bit.

(01:15:36): Trent, you picked it as your favorite chapter, or the chapter that you would give to other people, maybe not your favorite chapter, the chapters you would give to a stranger. I don't think you chose that chapter because you think, "Nah, Rob is completely wrong about this. It's just not going to happen. No one's going to replace any software," so I expect you to occupy a different stance here than Jon.

Trent McKinster (01:15:55): Yeah, I really do. I think you've got to draw the line. There's some things like Microsoft Office and those type of tools where software has to be your business, like you said in the book, for it to make sense. But I think in an area where there's a lot of specialized software needed or you're kind of in the matrix where you're doing a hybrid of build and buy, there's an opportunity there. I think the line also kind of sways with, how interoperable does it need to be with other systems? Because if it can be an isolated system, there's less risk there. But when the systems are working together, there is more risk.

(01:16:30): But I do think, especially if I was going to share that chapter with a leader, there's a lot of value unlocked there that they probably don't realize as potential in the future. But I also would say I think we're going to have to answer different questions, too, because I've heard leaders say, "Okay, what's the ROI with a resource person versus a token?" I think you've had a couple episodes on that as well, Rob. So I think even with exploring that path, we're going to have to weigh the cost. Is it worth the squeeze to build versus go buy it?

Rob Collie (01:17:03): And keep in mind also that the larger the organization, the more likely you have professional developers on staff who then also now have access to these tools. We've hired a couple of people that if it made sense to us, we would put them on the Salesforce replacement project, and they would get it done in a way that worked for us and saved us a bunch of money. But it turns out that deploying them in ways where we're building things for clients that generate us revenue is a higher ROI way to deploy these developers. And they're very, very, very much AI-embracing developers. They're into the agentic development, co-building thing.

(01:17:41): One of the things we do really well at P3 is hire from the gigantic. We're really good at finding the needle on the haystack, and we're even better at it now because we have a tool called Haystack that I built that gets through a lot of the fluff. It disqualifies large, large, large volumes of people that we don't want to talk to so that we can get more face time with people who might be awesome. When we went to hire our first developers ever, we got ringers. But again, we're basically like a 60-seat Salesforce company. The math changes when you're a 600 or 6,000 seat. Pick and choose, right? There are things at our company that were like modules within Salesforce that we have replaced. We've built some operational back-end software that we use for a lot of things like cashflow management and things like that that are really awesome.

(01:18:34): Tim, where do you fall on the SaaS replacement spectrum?

Tim Rodman (01:18:40): Before I respond to that, I want to give my pendulum theory that I've seen in different areas of technology. For example, in the 1970s, you had big mainframe machines, which was centralized computing where people would come in and bring their punch cards to use the centralized computer. Then in the '80s and '90s, you had the personal computer, the PC, which was distributed computing. Then the dot-com era brought things back to centralized computing. That's when cloud ERP products like Acumatica that I work with became popular.

(01:19:12): But now even a cloud ERP product like Acumatica has evolved to the point where the web browser is now acting a lot like an old desktop application in that a lot of work is being done on the local machine rather than a centralized web server. It's a return to decentralized computing. So it's just a pendulum that swings back and forth. And I think we have something similar going on here in that I do see the SaaS apocalypse/emancipation going on right now in the short term. I've actually heard this from consulting firms that lose out on selling Acumatica to a company that's trying to build their own ERP, which sounds crazy to me because I don't think it's cost-effective for most people to build their own ERP, and even if it is, it's not just about building it, it's about maintaining it. I'm old enough to remember the era where everyone moved off of homegrown systems, and a lot of that was triggered by the person who built it retiring or quitting. It's the classic what happens if so-and-so gets hit by a bus situation.

(01:20:12): So I think it's just a pendulum, and to the extent that we are currently seeing a SaaS apocalypse/emancipation, I don't think it will last forever.

Rob Collie (01:20:22): The age-old cop out is also the truth. It's going to be case by case. It's not one or the other. The people who are saying it's all going to happen, it's all going to get replaced, and the people who are saying nothing's going to get replaced, that's not where it's going to land.

(01:20:39): There's another angle to all of that, which we don't have time to really get into, but one of the things that the big tech companies are predicting is that the entire software layer of SaaS collapses into agentic systems as opposed to we go and we code identical copies of these SaaS systems on our own and replace them. As organizations increasingly go AI-native over time, the whole notion of all these interfaces and everything, everything about the UI and everything, it all becomes sort of antiquated. There's systems of record, and we need those, but the thing in between, the business logic in between isn't going to be expressed in traditional SaaS software so much as it's going to be expressed in customized agentic AI systems, which are not just LLMs, as the book covers. There is software in there, but the whole workflow changes.

(01:21:31): For example, even in one of these cases where we've kind of SaaS-replaced something in the back end to help us look at cashflow, we got a much better, much more convenient user interface in the process that matched our workflow and actually was way better than anyone ever would've imagined. Even now we're evolving. We're going like, "Well, why are we even tabbing around in this thing? Why are we going from page to page in this application and dragging and dropping things when we could just be saying to it, 'Hey, how is cashflow? Tell us what we need to be paying attention to'?" We can reimagine the way that we interact with this thing. And so we've been adding a chat interface to it that increasingly is replacing the usage of, go find the right corner of the software to check the right checkbox. And so we're going to all be learning a lot. The version of us two years from now is going to know how a lot of this turns out. Right now we're all just predicting it.

(01:22:25): But listen, gentlemen, thank you, thank you, thank you so much. It's a real pleasure. I mentioned to you backstage that it's almost a shame that we had to record a podcast today because just the four of us, the three of you getting to know each other... I've had now decade-plus ongoing conversations with all three of you. You've never met or even talked to each other as far as I know, and this would be a hell of a hang if we could all get together. The conversations would be amazing.

Jon Perl (01:22:57): Agreed.

Rob Collie (01:22:58): Glad the three of you got a chance to even just briefly meet and interact like this. It's really kind of blending the streams for me. I don't take it for granted that three people like you are willing to spend a couple hours with me. Again, thank you so much for reading my book.

Jon Perl (01:23:11): Thank you so much for the opportunity, Rob. Thanks for having us here. Thanks for encouraging there to be a community around your work in the first place. It makes everything so much better and so much more fluid. There are not so many people in the world you can talk to about these things.

Tim Rodman (01:23:26): I'd like to second that. And I'd also like to say thank you to both you, Rob, and your wife, Jocelyn, for the book. I think there's a reason why there's a 15-year gap between your books, because it takes a lot out of both of you to make it happen. From my standpoint, it really is a gift to us and a labor of love on your side, so thank you very much to both of you.

Rob Collie (01:23:47): She is a book widow for three months. Yes.

Jon Perl (01:23:50): Thank you, Jocelyn.

Trent McKinster (01:23:51): Absolutely. It is a gift to leaders, like Tim said. Thanks for giving us a pre-read. Looking forward to seeing how successful it is out there in the public.

Rob Collie (01:24:00): Guys, thank you very much.