The Power of Empathy & Why Drudge Work is Dying, with Microsoft’s Scott Sewell

Rob Colley (00:00:00): Hello friends. Today's guest is Scott Sewell who works at Microsoft on the Fabric CAT team. It turns out that Scott is a bit of a fan of this show, which is super flattering. We were grateful to spend a couple of hours becoming fans of him in the process of recording this episode. Just a super thoughtful and kind guy and we wound through a number of topics as you might expect from us. Empathy for one's customers, stakeholders and users, and how important that is to success, is one thread that ran through the entire conversation. That's very much on-brand for our show, of course, and hearing Scott's stories of having come to the same conclusion in parallel throughout his career really helps to cement a lesson that we all could afford to not just learn, but to keep learning at greater and greater levels of absorption over time.

(00:00:47): We also touched on Microsoft Dynamics and what changes in our minds when we refer to our users as our publics. Then we changed gears and discussed the impact of AI on academia through the lens of Scott's wife, who happens to have a very pragmatic and progressive attitude towards her students' use of AI, despite coming from a non-technical background. We also talked about how AI isn't cheating if everyone is using it, and what might happen when drudge work is truly drained out of otherwise nuanced and intelligent jobs.

(00:01:21): There's also a powerful statement in here near the end by Scott about Power BI Copilot, and this statement 100% matches thinking that I've already been sharing on the podcast, which is basically that the version that we're seeing today of Power BI Copilot is really just the tip of the iceberg, and that the world of 12 to 18 months from now is going to be very different from today. I'm going to keep hammering this theme in future episodes because I think it's super important for people to internalize it, but hearing it from Scott might land just a little bit differently than hearing it from me, and on that somewhat weighty note, let's get into it.

Speaker 2 (00:01:58): Ladies and gentlemen, may I have your attention, please.

Speaker 3 (00:02:02): This is the Raw Data by P3Adaptive Podcast with your host Rob Colley and your co-host Justin Manhart. Find out what the experts at P3Adaptive can do for your business, just go to p3adaptive.com. Raw Data by P3Adaptive, down to earth conversations about data, tech and biz impact.

Scott Sewell (00:02:32): I am excited to get a chance to talk to you. I've enjoyed your podcast. Honestly, I love the take that you have on it. So much of the content that I consume on media right now is mostly just somebody explaining a product feature, trying to make it entertaining about there's the product, here's the feature, here's the thing we're doing. You guys are taking a completely different tack to it and I've enjoyed that it. Stands out, not to over swell your head or anything, but the philosophical approach to it has been fun to listen to.

Rob Colley (00:02:59): That's a heck of a start, Scott, I haven't even had a chance to say welcome yet. Thank you so much. Yeah, welcome to the show, Scott Sewell. Let's start here. How about you tell the audience, A, what your current role is at Microsoft, and then B, we'll just go straight into what's your data origin story?

Scott Sewell (00:03:15): Sure. My role right now at Microsoft is a principal program manager in the customer experience side for Fabric, and my focus area is all Fabric pointing back to the Dynamics side. That's really where I started from 21 years ago on early versions of Dynamics CRM. But my responsibility is get some data into it, capture what they had, spreadsheets, handwritten notes, whatever it was, to get it into the system. My favorite part of it was not just building the CRM system, but then prove the value of it and prove the value of what was going into it through reports, really simple SSRS reports back in the day. Then when I got to the Power BI side, when that became a tool for me, just exploded in terms of how I could show off the value of the data that I had to my stakeholders.

Rob Colley (00:04:09): You said 21 years ago with the early versions of Dynamics CRM, I remember when it became a thing and it seems like yesterday. So when you said 21 years ago and I'm cross-referencing that with seems like yesterday, I'm like, oh, no,

Scott Sewell (00:04:23): Yeah.

Rob Colley (00:04:24): That is not good that yesterday is 21 years ago.

Scott Sewell (00:04:27): The CRM 1.0, interesting project to get those out and running the first and early versions.

Rob Colley (00:04:32): It was like right in that same timeframe that Microsoft bought Great Plains and that's when the Dynamics brand was sort of firstborn. In fact, I don't even think we were using the term Dynamics yet. It was like Microsoft Great Plains, but Great Plains wasn't CRM.

Scott Sewell (00:04:48): Yeah, there was a lot of affinity because some of the original work for CRM at that point came out of the Fargo office, so there was a ton of the folks there in Fargo that were working on CRM 1.0, a lot of crossover in that space. It wasn't a port of anything from Great Plains exactly. In fact, I think it was originally one of the first proof of concepts of a product built on .net.

Rob Colley (00:05:10): I think so yeah, that product release of Office, the 2003 Office release, which is incredibly forgettable. 2003 was the first time that we were all screwing around with .net on the product teams for the first time. You were mentioning getting data into the CRM, almost like porting customers from their existing borderline pen and paper or previous CRM systems, getting them on boarded by transferring their data into the CRM system.

Scott Sewell (00:05:38): By hook or by crook. We were trying to find any way we could get their data, any historic data about a customer pulling that together. And then import tools were non-existent, so it was all directly into the SQL tables to get it rolling.

Rob Colley (00:05:54): Oh, I bet that was lovely.

Scott Sewell (00:05:55): Learned a ton.

Rob Colley (00:05:56): I bet. How much easier would that stuff be with Power Query today?

Scott Sewell (00:06:00): Oh, yeah, snap. Wouldn't be nearly as fun.

Rob Colley (00:06:03): You're getting people onto the CRM so that they can track things. It's a line of business piece of software that's meant to make the business go. It's transactional. Your sales reps need to be able to track their interactions, need to know information about customers and what stage they're in and all of that, but then as a side effect of going digital with all of that. You now have the second use of data, which is where people like us come in, where you can start to see all of your activity is going through a digital system. You have the opportunity for supervision, the super awareness, this is where BI comes in. You have to have these digital line of business systems to even make BI possible.

Scott Sewell (00:06:42): Exactly. From my perspective, I was a consultant on these projects, an architect and delivering it, but also consider myself to be sales because if I could find ways to add value to the project, extend the project, that kept me billable and that kept my utilization numbers up. That was a good thing on my plate. So I really saw my role not just to say how to capture the transactional details of it, first and foremost, that's what the first line of business was to capture that. But I also had to prove that we weren't just throwing stuff into a bottomless pit. There was some value being created out of it. I was trying to build some reporting, build something to prove that, yes, we were actually making some progress for this customer or we're making progress for your team and letting them know what we can do.

(00:07:35): That became a virtuous cycle because a sales manager gave them really good reports, even if it was just in Excel that they could hit refresh and pull in the latest numbers. That sort of changed the perspective, so that Dynamics became the system of record. If it became the system of record, the sales manager would push back to the sales people and say, well, you got to get your sales into Dynamics, because if you don't get it into Dynamics, it doesn't show up on my nice sales report.

Rob Colley (00:08:04): Right. I can't see you.

Scott Sewell (00:08:06): Yeah, and so if it's not in CRM, it didn't happen.

Rob Colley (00:08:08): I've been rewatching an interview with Pittsburgh Steelers coach Mike Tomlin who turns out to be just a phenomenal person. He could change careers tomorrow, do something completely different than coach football, and then immediately be effective. It's like a masterclass in leadership.

(00:08:27): But one of the things that he has as an absolute luxury is that every single relevant piece of performance data, performance information, performance everything, the sum total of the performance of his entire organization is 100% visible. It plays out on the field. They've got camera angles of everything. They can replay it at great length. They don't even really need BI. You can see this player is doing their job or not doing their job. How well are we doing? What are our results? Everything is just front and center obvious. Now there's still a lot to do to improve raising standards and holding people to standards and helping people improve, which is what it's all about, once you have all the data. But he starts from total visibility effortlessly and that really hasn't changed ever since they've had cameras and replay. I'm just struck by how masterful he is, but also like, he kind of has it easy in a way. The rest of us actually there's mysteries.

Scott Sewell (00:09:26): It's an interesting challenge. Somebody like that that has access to a ton of data like that, you get the feel for it. Especially somebody who's a master like he is, can look at a lot of data and sort through and eliminate what's not important to him or particular strategy he wants to focus on. He can eliminate a lot of the details and focus on that and go to the videotapes and find that.

(00:09:47): I find the same thing is happening on the data we see coming through the sales processes is you've got massive amounts of data. In one case, I was working with a bank on customer service cases, massive amounts of data. Everyone could see exactly the same data, but they couldn't make sense of it because there's just too much, and finding the way to summarize that and present it in a way that's compelling and actually turns it into some sort of action, some sort of usefulness, is a real process that a coach that can do that in his scope. But then you move into some other scopes and they can't get their arms around all that data or they can't really get the value out of all that data.

Rob Colley (00:10:27): The thing that triggered the Tomlin reference was that these sales managers, once they start to get a taste of what visibility looks like and what it can do for them, they start encouraging everyone to get their activities into the system where it's visible. It connected that thought with seeing this guy, Tomlin, walking through a world where he just takes visibility for granted. He's never experienced a moment where they're not running the cameras during a game or someone isn't going to tell him whether they won or lost. He's never experienced that. You've got to bring it into the light.

Scott Sewell (00:11:02): Very clear outcomes and there's a very simple KPI that he works towards at the end of the day.

Justin Manhart (00:11:07): That's right.

Rob Colley (00:11:07): Very simple. Yep.

Justin Manhart (00:11:09): Scott, you've probably worked on a higher number of ERP projects than I have. I've only been a part of a handful of my career. I think all of my projects had emphasized the data in part or the efficiency of running the transaction part and then we need to see the data, the second use of data that Rob was describing is an afterthought and you just have this holistic mindset from the jump. I am familiar with you, I think we've crossed paths. You maybe did a Power Apps Dynamics webinar with a couple of members of our team maybe a couple of years ago. I know your name from the Dynamics Power Apps space, but now today you're on the Fabric CAT team and so you're at this intersection of ERP and analytics again. How does a guy from ERP where so much of that conversation sometimes is data in transactional efficiency, what's that journey been like for you and what really is the interconnectivity between the Dynamics world, the Fabric world, what's Scott up to there?

Scott Sewell (00:12:08): It's been so much fun to be a part of. I did this background, I knew the CRM product, I knew that backwards and forwards, it's just part of what I did every day. Every time I went to a pre-sales call with my team and they said, hey, we'll bring Sewell along with us and show them something about the product. If I showed them the screens of how to create a record, how to enter an opportunity, how to close a case. Some people in the room, that's what they were evaluating. They were going to check off how many clicks does it take to get to the center of an opportunity. But I also had a second audience in the room of the person who was ultimately going to sign the PO. They're going to sign the contract with us. 10 times out of nine, they really were investing in it to make sure that their team ran better, but they didn't personally see it as something that they would be necessarily using.

(00:12:59): They wanted the results from it, but what I kept finding that I could do was as I got excited about the data and the tools that I could show it off with, I was able to turn it around and present to both audiences, the people who were clicking in the application and the people who were going to consume, who were ultimately my buyer in some way. They were going to be person who was going to choose that. Being able to focus and make sure that they got what they needed out of it. One of those moments was I was in a boardroom in the Financial District back in New York City when I lived there, and I was demoing Dynamics to a team of folks that were using the customer service side of it. It was fun, it was going well, but in the room, the vice presidents over sales was in the office and he said, "Well, let me just sit in and I'd like to see what you're doing."

(00:13:50): He was so much senior to everybody else in the room that everybody else just kind of clammed up. It was kind of an odd dynamic. They wanted to defer to him on everything. But the weekend before we had the first release of the key influencers, Visual and Power BI and I talked to Justina and got some ideas on how to do it. I'd spent a ridiculous amount of time building up a data set in Excel that I could put into Dynamics and build it with Excel that had enough storylines in the data that it would actually light up the key influencer visual. You got to have the regressions in there. I'm back porting the data, trying to figure it out and all this is just in Excel while I'm sitting on my couch watching football.So I'm doing this and I'm having a great time doing it.

(00:14:34): So when he gets in the room, I thought, you know what? I can show him the key influencers part of what he can consume, what's going to be useful for him. Not calling out key influence or anything, but just I can show him how this data will be useful to him as he's trying to evaluate. So I'm demoing and I get to the point where I'm like, I'm ready to show off the Power BI reports that I've built for Dynamics. I get to that point and I'm showing him this. He goes, "Stop," which is terribly awkward in a sales demo from the guy to go just stop, and he turns away from me and he turns to the person next to him and he tells that person that this is the kind of insights I've been looking for. There was a little bit of back and forth and she said, "Well, we don't have that in Tableau" and, "It wasn't my question." He goes, "This is what I need."

(00:15:22): So what I learned from that was obviously the storyline, having to figure out what's the motivation of this individual. His motivation was competing against other managers within the bank.

Justin Manhart (00:15:36): I love it.

Scott Sewell (00:15:37): And he was able to look at that and say, I can see how I can evaluate all the deals that are going through my pipeline, who's closing on this kind of a deal, who's struggling on this product or with this industry, and that's the storyline we walked through with him and it turned into this massive project that we deliver. But ultimately that really was so exciting to me to be able to connect the dots between this transactional system like you described, this transactional capture and processing application and then the ultimate outcome for the business on the other side. Power BI was just mind-boggling. As I did more of that, I moved over from the pre-sales side at Microsoft basically as a result of COVID, to moved over to the Power BI CAT team.

Rob Colley (00:16:26): Why did COVID do that?

Scott Sewell (00:16:27): There was a project inside of Microsoft to say this was early days. The sky was falling. We weren't sure what was going to happen.

Rob Colley (00:16:35): I remember that. Yes, I do remember that. Excuse me while I twitch for a second.

Scott Sewell (00:16:39): That's right. There was these rumors that the hospitals overflowing and all the things that were in that early days of unknowns and Microsoft stepped up and said, hey, we can help you build a system to track the ventilators, track the hospital rooms, track the equipment and people, and we can do this on our Power Platform very quickly. So they spun that up and got something going and pretty quickly they said, hey, we need to have some reports off of this, and a buddy of mine happened to be working on that team. When the team was struggling to get some Power BI reports out of Dynamics, he said, hey, I know a guy, in fact, within the team they had quoted and said, it's going to be a couple of months for and get this out. But he goes, I know a guy.

(00:17:19): So he calls me and so I basically worked the weekend, again, what else am I going to do? Everybody's locked down, so I worked the weekend, brought back some rudimentary reports for them off of what we had and got them started. That really introduced me to the team, to the Power BI CAT team. A few months later they picked me up to say because they wanted somebody from the Power BI side to be focused on the challenges of the needs of the Dynamics community, so that's where I came over.

Rob Colley (00:17:49): That's fantastic. I love that.

Scott Sewell (00:17:51): It was fun.

Rob Colley (00:17:52): There's an admittedly self-serving coincidence that I want to bring up here. I put the word key in key influencers even though key influencers wasn't released until long after I left Microsoft, so Bogdan and Jamie who had been working on the analysis services team on what was called data mining at the time, it was this feature of analysis services that no one really used. They weren't getting good adoption from customers. So they wanted to do an Excel add-in to expose the data mining features in Excel, and they came over and asked me for help designing this add-in with them. I think this also kind of paved the way later on for me to get recruited to work on Power Pivot.

(00:18:35): But anyway, they had all these algorithms that they wanted to cluster in, et cetera, that they wanted to put buttons on the Excel ribbon. One of them was this thing like root cause analysis, it was the regression and what should the button be called. After a lot of thought, I came up with key factors, so key factors was the name of that button. Then years later when I saw key influencers, I'm like, oh, influencers is so much better than factors, if I just kept looking a little longer. So I get to claim partial credit, but whoever came up with influencers instead of factors, they're the real hero. One of the best named features I think in the whole suite. That one's nailed.

Scott Sewell (00:19:15): The challenge with that feature is to show it off, you have to have some really good demo data.

Rob Colley (00:19:21): The real world is going to manufacture all kinds of great examples of it, but you can't use the real world because it's sensitive. So frustrating.

Scott Sewell (00:19:29): Yeah, because all the demo data that I had was you'd use a random function and everything winds up being a flat line. All the columns are the same height. There's no trend, no nothing.

Rob Colley (00:19:40): Probably like five years ago I got mad scientist about generating demo data and I got into all kinds of statistical functions in Excel that would allow me to set different center points for different members of a dimension, set a different center point for this store or this product relative to the other products. Then have it generate demo data using random functions, but using those center points, those medians as a seed, so that you could get stories. The thing that the real world and chaos just does for us, it introduces all kinds of differentiation between corners of our business. Yeah, the random function doesn't do that. You've got to go and author the story you want back into the data and then randomize it. It's really frustrating.

Scott Sewell (00:20:25): I would love to show you my demo data Excel file, but I would be embarrassed to show you how much time I spent building this thing.

Rob Colley (00:20:32): Yeah, same here. This workbook to generate demo data, it ran so slowly it was like the reason to get a faster computer.

Justin Manhart (00:20:39): Out of curiosity, have you tried your hand at any of the GenAI tools for generating synthetic data sets?

Scott Sewell (00:20:48): I have not yet.

Justin Manhart (00:20:50): I have yet to bring this little pet project to life because I keep chipping away at it. It's worth a look. I want a fictitious data set about such and such or subject matter, but to give the prompts of I need it to be interesting, I need there to be anomaly, to tell it these things. Then say, okay, write me a notebook that's going to generate this code in my lake house for me.

Scott Sewell (00:21:12): You also have things you have to be careful of because you never want a sales trend line to look like it's going down when you're positioning your product.

Justin Manhart (00:21:20): Yeah, exactly.

Scott Sewell (00:21:21): You always have to show sales going up in at least a reasonable trend line.

Rob Colley (00:21:26): In that same vein, before we move on too far, what is Fabric CAT? Is there still a Power BI CAT team? Is it distinct from Fabric? What does your role really look like on a day-to-day basis?

Scott Sewell (00:21:37): Fabric CAT is really the evolution of Power BI CAT. There is no Power BI CAT anymore. We're all Fabric CAT. We've been assimilated.

Rob Colley (00:21:45): Rebranded as well.

Scott Sewell (00:21:46): Rebranded. Fabric CAT. So Fabric CAT is pretty much a collection of specialists, generalists, and then we have some folks that just focus on very specific data issues and some AI issues as well as another team that focuses on collecting customer feedback. A lot of that's collected through the Fabric CAT. I'm considered a specialist because I focus entirely on the Dynamics and Power Platform as a customer base, helping folks on the Power Platform customers who are using those tools to take advantage of their data using Fabric. Obviously it starts with getting the data into Fabric through the tools we have now. Power BI is going to be the first and foremost piece of it, but then going beyond that into pipelines and gather additional data from the rest of the organization together, visualize that back into the application. I laugh and say if somebody says dataverse three times, then I suddenly appear in the chat.

Rob Colley (00:22:49): We should clarify it for the audience that CAT stands for Customer Advisory Team, right?

Scott Sewell (00:22:53): Correct. It's interesting because to be clear on who's advising, and a lot of times we're asking the customers to advise us. Tell us what you're doing. Tell us where you're going. Tell us what your challenges are. We're more that than we are a consulting organization trying to tell somebody how to implement for their organization.

Rob Colley (00:23:12): Microsoft has a very strong enterprise focus. Microsoft software is useful at all levels. But when it comes to Microsoft allocating human beings to their customers, in that sense, Microsoft has a pretty strong enterprise focus. I tend to think of Dynamics as a mid-market tool. How much does Dynamics come up in an enterprise context? I'm used to Microsoft not allocating people unless it is an enterprise account.

Scott Sewell (00:23:45): I'm an arms distance from the team now, in terms of where they're pitching it. But I know the customers I'm dealing with are pretty decently large enterprises using some of the tools like Dynamics Finance and Operations. Those are some large customers as well as even some large customers are using F&O as well as Business Central for their satellite organizations, so there's pretty wide distribution of it. Dynamics CRM year over year, constantly growing in terms of people using it, not just for traditional CRM contact management but for other use cases of business data that they have, processes that they want to automate into a system without having to build something from scratch, something custom. The implementation, when it sort of transitioned over to Power Apps, Power Platform, and it became disassociated with the traditional that it was a sales or service first product, the breadth of where people have taken advantage of it just kept growing and growing.

(00:24:48): I had, actually, a Fortune 100 customer many years ago and we were implementing it for sales, 100%, that's what we were focused on. But about a third of the way through the project they got a new CEO and this new CEO had a focus on connecting with every customer and doing an inventory of what business this customer did with everybody. Happened to be a railroad and so they were going out and measuring how many boxcars could fit at the customer's site, what type of products did they load or unload. Even to the point of before a train gets to that customer site, do you have to call ahead to get somebody to open a gate to allow the train to go in. Really detailed data that they wanted to capture about it.

(00:25:34): Internally it was interesting because there was a team that was spinning up some traditional tools for development and they were looking at about a six or eight month timeline to get to a very rudimentary version of this information capture. Again, it was one of those things where I was stuck in town over that weekend and thought, yeah, I can build this on CRM much easier. I can capture 90% of what they want with just a simple form, so I built it and turned it around and showed it to them and they went live with it within weeks as opposed to having a prototype several months out. It just solved a problem immediately with this low-code, no-code approach. They used it for several years.

Rob Colley (00:26:17): I love that story. In fact it reminds me of something I recorded a few years ago, this notion of don't accept long timelines. Whenever someone sort of quotes you whether internally, externally, whatever, someone says that a data project is six to eight months, there is a much faster way. If it's an ERP implementation, no, not going to be faster, it's going to be slower in fact. But most data projects, if you've got a long timeline estimate, there is a pivot you can take in your strategy that's going to get you there much, much, much faster. Every time you see something like that, just go, okay. It reminds me of the stories of Steve Jobs, like his engineering teams telling him, yeah, we need eight months to do this, and he's like, oh, that's great, you have three weeks, and they always got it.

Scott Sewell (00:27:07): But that pressure adds to the focus and the clarity a lot of times.

Rob Colley (00:27:11): Yeah. I don't think you need the tyrannical component that Steve Jobs brought to the table. Look at six to eight months and go, no, we can do better. It's more of a positive thing than a negative pressure. But time and time again, I have so many stories like that. We can't even get started for a year and a half and we're live in two months, that kind of thing.

Justin Manhart (00:27:29): Long timelines mean you either have the wrong approach, the wrong tools, or both.

Scott Sewell (00:27:35): Or lack of clarity.

Justin Manhart (00:27:37): Yeah. In the approach, sometimes a project is made to seem like a big project, but it's really just lots of little projects. You get stuck on the starting line. It's like imagine if the railroad example, Scott, you gave, was tied up in committee being this other big ERP project. The CEO has the wherewithal to think about this as a priority. You have the wherewithal to be like, oh, I can solve this problem. That's so great, even to hear that on the data capture side, so much of what we do at P3 is like we talk about the second use, displaying data, displaying insights. There is this important feedback loop of creating the data, enriching the data. With your work on the Fabric team now and your focus on Dynamics, are you focused on leveraging of the data on the second use? Are you still focusing on feedback loops? What types of problems are you spending most of your time on?

Scott Sewell (00:28:29): If you boil down what I focus on right now, it's going to be on the education side. I am first and foremost an end user in a lot of ways. I want to know how it works for a user in their environment and I want it to be simple enough that I could explain it to my wife, who's a smart person, but just not technical. That's just not her gig. I want to be able to help somebody pick it up and I wanted to see them become successful. The most fun parts of this has been seeing other partners, other customers, come back and go, Hey, look at what we did or look at what we've done for this other customer around the line.

(00:29:08): Some of the material that how to use it is actually a little bit obscured because we say, here's 14 different ways you can use it, but we don't give you a clear guideline on point A to point Z. This will get you there. There's other variations off that, but let's show you a simple route to get there. I built a few dashboards that I'm really proud of, probably a little too proud of them. But I try to keep them simple enough so that when you look at it behind the scenes, you pull back the covers, you go, oh, well, that's not nearly as complicated as I thought it was going to be. That's a win for me. Some of the first reports that I got were these incredibly detailed convoluted processes that covered every use case and they were so complex in the query, as soon as you say, well, I want to add this additional field to it, it all blew up.

Rob Colley (00:30:00): Six to eight months.

Scott Sewell (00:30:01): Yeah. So I wanted something that was like anybody could pick it up and go, that's easy. I can add two rows to this query and now we're running. So that's what has been the focus area. Also trying to resolve challenges between working with some of our preview customers as they evaluate it. It's always rocky points and learnings and things. So I'm also trying to build up my data sets to stress test it a little bit. I try to be my own worst user, but it's fun. It's a good time.

Rob Colley (00:30:33): One of the things we talked about a little bit backstage was the P3 faucets first philosophy.

Scott Sewell (00:30:39): Love that.

Rob Colley (00:30:40): We were talking about long timelines earlier. I wanted to circle this back to this. You mentioned that you're in agreement with the faucets first philosophy. You also told me that even when you're a proponent of this approach, and by the way for those of you've never heard, faucets first means we start from the results we want and work backwards, as opposed to going infrastructure forward. We've got to go build a bunch of infrastructure before we can get around to answering your questions or delivering your business need, and that's when you end up with the six to eight month timeline. The infrastructure forward project is almost doomed to be six to eight months minimum. Whereas starting backwards from like, let's see if we can get something going just to test whether we're actually even on the right track and then build the infrastructure behind it to feed just that as you get closer and closer and more and more zeroed in on what you actually want. That's the thing that gets you the two-week delivery or the one-month delivery by comparison.

Scott Sewell (00:31:34): I love the idea of starting with that end in mind and working your way back from it as opposed to trying to build the world to solve a simple problem. Many years ago, I nearly blew it in a role that I was in. I was working for really great guy, very talented entrepreneur, and I nearly blew it with him because he had said, I want to have this for a demo for a customer. So I would turn around and fiercely work away and punch up all the forms and make it all nice and clean and organized. Then when I would turn around and hand it, show it back to him and he would go, that's not really what I'm looking for. I need this. Oh, well, maybe I just misunderstood. So I went back and fiercely fixed it and I kept falling flat in the deal and I was kind of getting frustrated. I was getting irritated. I almost had a little bit of a pity party for myself.

(00:32:26): But the difference of the pivot was as soon as, in my head, I thought if he would just tell me exactly what he wanted, then I could go build it for him, but I thought, well, if he could tell me what he wanted, he doesn't need me to go build it. He could just offshore it real cheap. But in that process I kept looking, okay, let me go back and talk to me about how you want the demo to go. What do you want the customer to feel? As soon as I got on the same side with him and understood the problem, not just what he was telling me what he envisioned it to be, but also I had to go in and ask a few more questions and really listen to what the problem was. When I pivoted, we got on the same team and we went gangbusters after that. I was a little hard-headed and it took me working to listen to him and pivot for that.

Rob Colley (00:33:13): It's so easy to get distracted.

Scott Sewell (00:33:16): Yeah.

Rob Colley (00:33:16): Adopting your customer's motivations is both critical, it's crucial to success, but it involves both a choice. You have to make that mental shift, even if it's a mental shift you've made in the past, you need to make it again, you need to remind yourself. But then there's also a process of it. Once you've made the decision, now you have to do the work to really understand what those motivations are. They're prereqs to success. You have to both be in on it and remind yourself of what you're supposed to be doing, which is hard, but even once you've done that, now you got to go do this work that you wouldn't necessarily think you have to do. You're focused on, I got to build this solution, blah, blah, blah. You don't really understand as well as you think you do yet.

(00:33:59): Years ago I read an essay by Charlie Munger, who, until recently, until he passed away, was Warren Buffett's right-hand man. He's talking about this one thing he believes, it doesn't even matter what it is that he believed, he said, for my entire adult life I have been in the upper 5% of my age cohort in believing this thing, and yet every year I'm confronted by evidence that I didn't believe it enough.

Scott Sewell (00:34:25): Interesting.

Rob Colley (00:34:26): That's how I feel about this aligning yourself with customer motivations and really understanding and working backwards from there. I feel exactly like that. I'm in the upper 5% of believing it and yet all the time I'm discovering that I'm not in on that thing even as much as I should be.

Scott Sewell (00:34:44): At least for me, it's always a process of getting my ego out of the way. Going back and saying, let me stop and I've heard this song before and so I know how this story goes, and if you jump to the end of the story, you miss so much.

Justin Manhart (00:35:00): It's amazing how often the substance of what a customer is saying at the beginning. It says, Scott, my problem is X. Great. I have a solution for X. Let me show you, right? You realize, no, it's like X to the fourth power is the real problem, and you don't get there without putting your ego aside, being curious, asking questions. Customer problems seem very similar when you see lots of customers. You've gotten roles like we've got. But they're all very nuanced in different ways and you really got to make sure you understand what's going on and if you can get there, put yourself that other person's shoes, really embody that. I love that ego aside thing because it's easy for us as consultants like, ah, we know the best way and we know where the real ROI here. It's, Scott, what matters to you? What's valuable to you if we did this, is it worthwhile for you? Does it make a difference for you? That's a really important thing to have in a role you've got.

Scott Sewell (00:35:58): It's also fun as you get into it. Sometimes the need isn't quite as clear to them as you find you're helping each other clarify and you're cutting away and getting to what the motivation is. The case of the VP, his motivation was bonuses against his teammates. He didn't care about external banks. His competition was against his fellow managers and wanting to get the biggest slice of the bonus pie, and so figuring out what that motivation was becomes a superpower.

Rob Colley (00:36:28): In some sense, in that situation, your customer is the bank, is the broader org, but now you've got this sponsor. By the way, their personal motivations are for their personal success, and these always line up to varying degrees, like if the sponsor succeeds and the project succeeds, the company succeeds, et cetera. Was there ever a moment when it's dawning on you that this is what his personal motivations are, it's to win against other people at the same bank? Is there ever a moment where you're like, maybe I shouldn't help him?

Scott Sewell (00:37:01): Maybe there would be in another situation, but in that situation, his motivation was to be better and better and better, and if he was getting better and better, the bank was succeeding. He wanted to succeed even faster than his other counterparts, but him succeeding was not holding them back.

Rob Colley (00:37:20): Yeah. Obviously I was leading the witness with that, but you could imagine another person in that role going like, well, this doesn't feel right, but the broader bank wants that. If that manager succeeds and outraces everybody else, that's because that manager's producing better results for the overall organization and now the others are going to be incentivized to follow suit. His advantage isn't going to last forever. If it's actually an advantage, it's not going to last forever because everyone's going to adopt it. Or they're going to be replaced by people who do if they're stubborn about it, right? In the end, so often, even a broader organization's motivations, your only on ramp, your only window into it is through the individual's motivations that you're working with. So leaning into their human motivations is kind of a godsend.

Scott Sewell (00:38:09): I don't know ultimately in that situation that it made a big difference in what we delivered or what we sold to the customer. By connecting with him, he unblocked us. We turned him into an advocate for us, and by doing that, we got to the sale. That was our motivation. By giving him a perspective on it, we connected the dots for him and he could bless it and say, yeah, these guys are delivering on what we as a bank and what I as a manager need. When I first brought Dynamics and Power BI together was with a Titanic demo. That really lit me up in a lot of ways in terms of these two products, this is better than a Reese's cup. These two things are amazing, but they go great together. That was a pretty big turn and that has an AI flavor to it. I was almost infamous for that demo because I demoed the Titanic over and over and over after a while.

Rob Colley (00:39:05): Is this the who dies and who lives demo?

Scott Sewell (00:39:07): Combination of that as well as using key influencers and finding who sank or swim or whatever you want to say. It's a well-known data set that people have used a lot of times for machine learning. But the funny thing was I demoed it and it was a group like the three of us. First time I showed it at Microsoft, I got feedback, well, is it too soon to talk about this? So at that point I pivoted to saying I always referenced the movie Titanic and that was a little safer to talk about the movie. Any humor I tried to inject into it, I would reference Leonardo DiCaprio. Well, fast-forward a year or so ago, I was over at the school where my wife teaches and I was demoing Power BI and I referenced the Titanic movie and whatnot. I realized it came out before anybody in that room was born, that I felt extremely old. That was fun.

Rob Colley (00:39:57): It's no longer canon, certain pop-cultural references that you could assume that there was a critical mass of humanity that's aware of them, a lot of my references are now out of date. They don't even show Bugs Bunny cartoons anymore.

Scott Sewell (00:40:10): I know.

Rob Colley (00:40:13): It's kind of like radio carbon dating, right? You freeze, at some point your relevance just decays with the half-life. They can measure your age by like, oh yeah, this one's still talking about Fight Club like it happened-

Scott Sewell (00:40:24): Exactly. Another favorite movie.

Rob Colley (00:40:27): Yeah, no doubt.

Scott Sewell (00:40:28): That was also one of those data gene moments for me. The other part of it was just the learning to tell a story and how a story can educate in a way that doing a technical demo couldn't. I would get into demo rooms and I would never lead with it. We'd start talking about the products, here's this feature and here's Power Automate. Here's this and here's Power BI and show that off. I said, let's take a break for coffee. We'll grab a cup of coffee. We can come back. We'll continue. By the way, I got this crazy thing over here. Let me show it to you, and I would show the Titanic demo and never once mention a product name. It was just a story of how you set up this passenger tracking system and handheld welcome app on Power Apps that would welcome you, how to track customer safety, ensuring customers are safe, doing all that.

(00:41:17): Then at the end of it, I would come back and go, by the way, this was CRM that I was showing you there. This is a Power App. This is Azure Machine Learning. This is Power BI. This is Power Automate. Everybody was so disarmed by just looking at a silly story that we could get through that and all of a sudden people go, oh, I'm connecting all the dots now. I know what all these tools do. Titanic's going to be on my tombstone.

Rob Colley (00:41:42): I've had a similar experience with an old demo that I used to do with UFO sightings.

Scott Sewell (00:41:46): That's a great one.

Rob Colley (00:41:47): Just to drive home the notion of being able to cross-reference multiple data sets in one place leads you to conclusions that you would never reach looking at individual reports from individual silos. So we took the UFO sighting database and spliced it together in a data model with USA based UFO sightings and then with various drug use trends in the United States as well. It's crazy how you cross-reference these things, you splice them together into a single line chart and you index them each against their relative mins and maxes so that they fit on the same chart. Just how parallel UFO sighting trends are to various drug consumption trends in the United States. They just go in lockstep. Some percentage of the audience that I was working with, trying to teach them how to write DAX or whatever, some percentage who had not gotten it before seeing that demo suddenly would get it. Oh, okay. It is rare to find these data sets that are public so you're not dealing with sensitive customer data.

Scott Sewell (00:42:46): Yeah. One of the fun learnings out of the Titanic data set that I haven't seen other places, but I found it through the key influencer, was that obviously the first class, and second class, third class, there's a ratio of people who survived in each of those classes, diminishing as it goes. If you look at male versus female or age group, there's all these different parallels. But then you start adding layers of attributes on top of each other. The group that had the highest survival rate on the boat, obviously if you're a musician, bad news, but the group that had the highest survival rate was female staff members assigned to passengers, and they were the staff members that came along with the passengers. They didn't work for the ship, they worked for the passengers, so they were like the nannies and 100 percent of them survived because they were carrying the children of the first-class passengers.

(00:43:43): So they wound up in the boat, and so there was like this aberration. They stuck way out. The other one was funny was if you're a young man, your odds are really, really low, especially if you're a young man that's not a passenger. If you're a crew member of the boat, with the exception of the deck crew because they're the ones that have to row the boats. If you just put a percentage across the whole group, that's a very low percentage, but there's one group pops out of that as long as you start crossing over some different attributes, which was a lot of fun to discover each time as we demoed it.

Justin Manhart (00:44:15): Let's face it, you work on a team that is responsible for technology products and these technology products can do dang near anything. You can build all kinds of apps in your CRM. You can build all sorts of data structures and reports and outputs and all this sort of stuff. But the ability to contextualize that, tell it in a story format that makes it feel real and relevant and people can see themselves succeeding or solving a problem with it. Earlier you said both audiences, but it's really like there are always multiple audiences.

(00:44:49): There's the person who has the motivations to succeed that is going to become the champion. There's the technical people that will want to make sure it's fast and can be integrated with and the people that are going to push the buttons and the levers. So you're telling an experience that's so well-rounded. I think it's important for people, whether you're people like us as consultants or you're in a company and you're responsible for these things, that your awareness around all these things, it can be the difference between a project going well and a project not going well.

Scott Sewell (00:45:20): The emphasis that [inaudible 00:45:22] puts on empathy, that's a motivating factor. I've got to lean into empathy when I'm talking to folks trying to figure out what is it, we talked about their motivation, what are they afraid of in some cases? What are they intimidated by? Try to figure out where they're coming from so that I can help them get from point A to point B to help them know that they can succeed with these tools. Not with just the tools, but with the skills that they already have and skills that they can build on. It's been fun to do that.

Rob Colley (00:45:53): With that empathy thing in mind. One of the things that's been tickling at the back of my brain throughout our conversation is, Justin, you just mentioned the multiple audiences that you have to keep in mind. My wife, Jocelyn, has been doing an online master's in communication. One of the things that I've noticed, and they never really explained their choice of this, they don't use the word audience, they don't use the word stakeholder, they use the word public. Keeping in mind your multiple different publics. That doesn't even mean external. Within a company, there are publics audiences within the company, there are publics outside of the company. I'm not talking about the grocery store in Florida.

Scott Sewell (00:46:36): Good sandwiches though.

Rob Colley (00:46:37): It's the plural of the word public. Think about the connotations of the word public. We have to protect the public. We have a responsibility to the public. Audience is someone that you talk at.

Scott Sewell (00:46:51): You're right.

Rob Colley (00:46:52): So I've been always a little bit struck by their deliberate choice of this word public instead of audience and how much it changes your sense of responsibility to them. It helps trigger the empathy.

Scott Sewell (00:47:05): There's times where I've done demos and the worst ones are when you're doing it online and nobody else has a camera on.

Rob Colley (00:47:14): Yep. Those are the worst. I agree.

Scott Sewell (00:47:15): You're just talking at the wall of my office here. I love the feedback. I love somebody looking at me like what did you just say? That's gold, like, okay, let's lean in and we'll make sure that I communicate that in a better way. If it's perceived as a one-way communication from you to the audience, absolutely no good.

Justin Manhart (00:47:34): For your audience, people that are using Dynamics, is there something they should be very alert to? Hey, if you're on Dynamics, you should wake up to Fabric?

Scott Sewell (00:47:44): For the Power Platform communities that are out there, Power Platform really, I think the thing that was amazing about it, going back to my early days and a lot of folks were early days, when we went to implement a Dynamics project or a Power Platform project. Literally I was sitting on server boxes installing a server into a closet and spinning up and installing SQL or IIS or Kerberos or any number of tools that were in there and then doing all the work to get all the plumbing. Then I would go and talk with a customer and configure it for what they were trying to do.

(00:48:16): Fast-forward to 10 years ago maybe, all of that complexity of the plumbing started falling down below the surface line as it came online and suddenly the stuff that I had spent a ton of my project hours doing was now just done by Microsoft for me, and I could spend the rest of my project hours on tuning it for that customer, finding out what are we going to solve for that customer and whatever it needs is. But if you looked a little beyond my project at the BI side of it, there was still this uphill climb of data warehousing, data lakes, and integration, ETL, and tools that were needed for analysis services and things like that that I needed to engage the customer, try to get them to set up so that I could add the configuration value on top of it.

(00:49:05): In a lot of ways, there's a parallel what's happened with Fabric as to what happened with the Power Platform. A lot of that complexity slid down below the surface line. A lot of the complexity then now just goes to being part of the software as a service. So for the Power Platform users and partners, come on over, the water is fine. It's a transition that we've been through and we can do it again now over into the analytics side and gain some great value. The other part of it is in the first iterations of the connections between Dynamics and Synapse and Azure, it was a little bit of a weird science project to set up. It wasn't simple. It worked, it was great, it was fun, but it took some effort to get it up and running, or it took a lot of money with if you were using the DES and you had to run a pretty decent sized SQL server up all the time.

(00:50:02): Now with the Fabric connections back to dataverse, it literally can spin up the thing and have a data lake ready to consume in a matter of an hour, depends on how much data you have in it. It'll start hydrating and you go. So the complexity of the plumbing should no longer be a barrier. The need to be able to just say, hey, let's just go and try to figure out what do we want to represent about your particular needs. That's been the transition, I think, with Fabric link or the connections between Dynamics and Fabric. It's been tremendously accepted and it's growing every day.

Justin Manhart (00:50:36): I could say it back to you maybe this way, so if you're out there, your business is running on Dynamics and you feel like you just could never pursue your analytics that you really think you need and want. Scott is saying, come on over, the water's fine.

Scott Sewell (00:50:55): Yeah. The great thing is it's lowered the barrier of entry to people to be able to start taking advantage of it. What was too complicated for in the SMB space even, now the tools are available that you can start taking advantage of it.

Rob Colley (00:51:09): Scott, in a way, like the Genesis of where we started in our conversation on LinkedIn that led to you coming on the show. It started with one of the absolute favorite pieces of feedback I've ever heard about our podcast was when you told me that our conversations about AI on the podcast have triggered an ongoing series of conversations between you and your wife about AI. So first of all, that just felt really, really good. That's definitely a point in favor of coming on the show, but it's more than that. I wanted to hear more about that and the medium for hearing that story is definitely verbal. It's not typing back and forth to each other on LinkedIn. I felt like this would be a good vehicle and an excuse. Can you tell me a little bit about that relationship and how that's been going, and you describe her as your non-nerd wife, but it also sounds like she's pretty sharp, so let's hear it.

Scott Sewell (00:52:05): So for years my wife worked in communications, she worked in PR, did work on TV, and so she was decidedly non-technical. Did not ever bother opening Excel. Her technology ended pretty much at email and writing a few contracts in Word. That was about it. Fast-forward to when we lived here, she now teaches in the college of business here locally and teaches business communication as being one of her focus areas. As she got into that, she picked up more and more skills. I taught her how to use PowerPoint a couple of years ago and now she's showing me stuff. Dangerous. The thing that was interesting is December of last year or year before last, when ChatGPT came out, it hit the radar of consciousness. Really quickly the academic world where she works, the first reaction was, oh, no, students are going to use this to cheat, and the first reaction was we got to find a way to block people from using it. We got to have a way to detect cheaters using AI to generate their text.

(00:53:11): But my wife, I love for her so many reasons, but one of the things that she said is no. If that's what people are using in the business world, that's what we got to prepare them to use. There's a quote that she stole from somebody that said what academics sees as cheating, the business world sees as productivity. Preparing the students to move into that became one of her focus areas. Not in terms of teaching them how to write an LLM, but how do you use the tools that you find inside the tools, how to use that and use it well.

(00:53:47): Her role, what she has to teach, is quickly shifting from teaching them how to write a letter, like start from scratch and how to do it to making sure that they know what a good letter looks like. What does good look like is one of the hard gaps. I was talking to her about this conversation and I was referencing what you guys mentioned last week about knowing what formula to write, and the gap between, this is a real opportunity that she's leaned into now in terms of saying, you're going to use AI, but you need to know how to use it and understand what you're producing that it's going to be effective. She's gotten incredibly excited about it in terms of saying, okay, let's lean into evaluation of students in ways that allow them to use AI but not in a way that allows them to replace thinking with AI, or that leaves the door open to say, you can just regurgitate an answer you found searched on Wikipedia or created through ChatGPT. She's looking for ways to say, how do we turn this into a learning experience for them?

(00:54:57): She and I have had all these conversations about AI and how you position it in the classroom, which bouncing off of some of the things that you guys have been talking about has been fun, knowing what formula is good, what formula works or understanding the problem better. That all leads to some of the conversations that she and I have had about it. And which if you had told me that I would be talking to my wife about AI two years ago, I would've absolutely laughed. She would've laughed. She would've laughed harder. It's become a fun part of the conversation between us.

Rob Colley (00:55:27): Well, first of all, huge credit to her.

Scott Sewell (00:55:29): Oh, absolutely.

Rob Colley (00:55:31): When a disruption comes along like playing ostrich, sticking your head in the sand, it's a time-honored tradition. It's the default response of most of humanity is to either deny it or fight it. There's no fencing this one in. We already can see to say how preposterous the idea is that you're going to be able to prevent them from using it. How preposterous the idea is that you're going to be able to use AI to detect other AI. It's like now it's just playing this hunter killer game.

Scott Sewell (00:55:59): Whac-A-Mole.

Rob Colley (00:56:00): Coming back to this idea of you need this human referee that adjudicates and modifies what these systems are producing for us, but also provides the right information in the first place to get it started. A silly but funny example that I'm reminded of, there's so many scenes in movies that are like this, but here's one from Game of Thrones where there's this trial by combat and these two guys are fighting with swords. One of them fights kind of dirty and wins in the show, kills the other guy. When it's all over, some of the people who are rooting for the loser, one of them says to the winner, you didn't fight fair, and he said, no, I didn't, but he did.

Justin Manhart (00:56:42): Scott, you mentioned two years ago you wouldn't talk to your wife about AI. This happened just a few months ago. My wife came home from work one day. She works in healthcare. She goes, "Oh my gosh, Copilot." Like that's all she said in a super thick Minnesota accent, because she has to write policy documents and prepare presentations. I think people that are onto this have this realization is there are certain tasks that do require a level of genuine human critical thinking, innovation, like breaking new ground. The way you described cheating versus efficiency, let's be honest, business is a competitive environment. Like in the story Rob used from Game of Thrones, that's the reality. There will be people that get ahead and it's going to be finding the right balance between where true critical thinking and innovation meets the need to be more efficient. Major kudos to your wife and her awakening to this and passing this along to students because I think there's a real risk of the younger generation not getting hip to this.

Scott Sewell (00:57:47): It plays into what she already leaned into. She inherited a class that previously taught you how to write how many spaces to put between the lines of an inner company memo or how to format a press release. That was the focus of the class going back a few years, and she was like, nobody does this anymore. Her focus shifted to saying it's not just a matter of teaching a student how to write, it's how to write with impact and write in a way that somebody will walk away and go, I understood what you meant. Or you motivated me or you convinced me of something.

(00:58:27): That plays back into the AI conversation because if the teacher assigned something and a couple of keystrokes, they've got a turn paper turned in, there's no real value there. The other side of that is also that if I produce a press release through a couple of keystrokes and it's a wall of text, goes on and on and on and verbose kind of like I get at times, just keeps talking. The same people who don't want to write that are also going to be the same people who don't want to read that.

(00:58:58): She leans in and says, your competitors in the marketplace, there are competitors who are going to generate walls of text in their descriptions of things and nobody's going to read it. Your goal is to change the perspective and say, how do I write this and how do I summarize it? How do I deliver it with impact so that people will actually embrace it? It's a pivot on the AI side because the AI side is great for giving you some ideas, but at the end of the day you've got to evaluate it with the human side and turn it into something that another human is going to read and consume.

Rob Colley (00:59:34): Even just thinking about it through the academic lens for a moment, I'm thinking about what the end game is for this. In the era when an eight-page paper is the product and there was no AI, there was no GenAI, in that world where they produced the eight-page paper, a tremendous amount of the credit in terms of providing them a grade was did they do it? There's so much labor that goes into producing an eight-page paper. So, in a way that dilutes the standard, if you read this eight-page paper and you're like, oh, this student only about halfway gets it, well, they still did the paper and so you feel like you have to give them credit. They're not going to get an F because they understood the topic halfway, they still produced an eight-page paper.

(01:00:22): Okay, well guess what students? In a world where eight-page papers are now easy to generate, I believe the standards are going to go up. You're going to be held to a higher bar with what your paper says, like actual clarity here. You're also talking about the power of the communication. If this is something that's meant to be communicating to others like how persuasive is it, how clear, how moving? If I was someone who was grading papers now knowing that these tools are available, I'd raise my bar quite a bit.

(01:00:51): The other flip side of it is if my job were to truly grade students accurately and the reason why we want to grade students accurately is to properly incentivize them to learn. By the time we're grading them, it's too late, but if they know they're going to be graded and they're going to be graded well, it's going to incentivize the right behaviors for them to learn the right things. So after they turn in the paper, I'd want to ask them a few questions to see how well they understand what they "wrote". We do this in our interviews. We've done this even before Gen ai. Someone else could have written the DAX for you, so when it comes time for the live interview, we ask you to explain your formula. If you don't know how to explain it, you didn't write it.

Scott Sewell (01:01:33): Right.

Rob Colley (01:01:34): Then you also, in this new world, would not be a good referee for knowing when the formula was correct.

Scott Sewell (01:01:41): Part of the challenge, she spends a lot of thought around the idea of how do I convince a student I would not have written a nine-page paper if an eight and a half page paper would be fine. As a student, I had other things I was interested in. Part of her pivot or what she's trying to do is convince that their objective is not something that's turned in. She's not excited about seeing a paper. She's excited about seeing how they got to that paper. What of their own personal experience could they bring into it? How do they structure their persuasive approach? Can they articulate the challenge that the paper is trying to address in a way that they actually embrace? There's a little bit of a tug of war between that in terms of you got one out of four chances you're going to get it right.

(01:02:28): That's so limited when you get further into it where you get into the oral examination where you say, hey, defend this, explain this, stand up here and present this and convince me of the reality of what you've described. That's both much, much harder on the student and the teacher. Teachers have a ton of work they have to get done, volume of work that they have to do. That's harder on both of them, but it also pivots it to the real understanding side of things. Again, I go back to your conversations about DAX and I was thinking about how would I train a new consultant today? How would I train somebody to step in that was eager, excited? They're fun. There are smart folks and too many things can be done for them by the AI tools to the point that they miss some of the process that it takes to figure that out.

(01:03:22): Can they define the problem? Well, you want them to use the tools, the AI tools, but you also want to understand what's behind it. Can you evaluate whether it's the right formula? Can you evaluate whether it's the best formula? I feel fortunate I got to learn a lot of the tools that I get to work on from a real raw state. Now I get to evaluate them in terms of really having a sense of what's going on behind the scenes. I laugh and say, I know where some of the bodies are buried in these tools. Somebody the other day said, I have generational trauma from early versions of this. Through that process, and of course people before me were like, oh, you kids these days you should have learned punch cards. But to me it's a fascinating topic, I'm more interested in her process sometimes than she is as we talk through it.

Rob Colley (01:04:07): When it first broke, I just couldn't imagine being a college professor at that time. My wife needs a back surgery that our insurance company has done everything in its power to exhaust us on. Go get yourself six weeks of physical therapy that everyone knows isn't going to help. We're not going to even talk to you until you've done six weeks of physical therapy, so we go do the six weeks of physical therapy, we come back, what now? No, you're rejected. So the end of the day we have to write an appeal and they're just trying to wear us out still. Right? That's how I feel about it anyway. I'm like, I just sat down and wrote an outline nine 10 bullet points, and I fed it to ChatGPT and said, hey, write an insurance appeal letter based off of this outline. It's awesome. Now I've got to go and tune it, but guess what insurance company? Appeal letters are now going to be easy. You're going to have to innovate in your denial strategy.

Scott Sewell (01:04:58): It becomes a little bit of an arms race in those situations.

Rob Colley (01:05:01): It is. We finally reached the point where the CAPTCHAs are harder to solve for humans than they are for bots, So I noticed an inflection point in Midjourney recently in image generation. For a long time, if you told it to put a label on something and then you want it to put some text on part of the picture, it would never get the text right, because it doesn't understand text. It's doing this computational intuition thing backing into an image.

(01:05:29): Recently, they must have added a new subsystem to the whole model that specifically handles text and adds the text in as an after effect, as opposed to generating it through the same algorithm that, again, intuitively backs into an image. I don't know what I was doing with this. I was like, generate me pictures of jack-in-the-boxes with the name Rob on it. I think I was trying to convey surprise, here I am. I never used it, but it did a great job at this. These captures with horses, it doesn't know what a horse looks like. It can generate horses. How funny is it? Is it something that could generate a horse but not recognize a horse. You need a different system. The system that generates horses is not the system that can recognize them.

Scott Sewell (01:06:13): The number of fingers on your hands is a constant challenge.

Rob Colley (01:06:17): Yes, but again, they're going to develop five-finger subsystems that go in and correct fingers, but it has nothing to do with the original model that produced the image. It requires post-processing to get these things right, which is just fascinating.

Justin Manhart (01:06:32): AI is being infused into the Microsoft landscape, so a parting thought from Scott around where AI is going to show up. I'm not asking for secrets from the CAT team, Scott, but I'm just like how it's going to impact that world. If you have a secret from the CAT team, that's cool too, but AI is a big thing. It's going to change the way we work. What's on your mind of how that might play out?

Scott Sewell (01:06:54): Even though I work for the engineering team, I'm not on the product decision processes, the product planning side. But you really don't have to be that much of a wizard to look around and say, AI is going into everything that we do, every tool that we have. If there's a way to make it easier for a customer to consume that or anything that we see where customers say, this is difficult, AI is going into that sooner or later, and probably sooner, just because that's where the investment dollars are right now because it's paid off so strongly already. You see all the things that we and our people we work with have said, oh, well, DAX is hard, or designing a decent looking report is challenging. All the things that we do, the tools that we're seeing, even to the point of getting where data modeling is starting to be influenced with that.

(01:07:52): Looking at how do we set these things up to say, let's take the drudgery out of this and make it easier for customers to consume it, because if customers are happy using it, everybody wins. Microsoft wins, the customer wins. There's nothing more motivating right now within Microsoft than to say, can we use AI tools to improve the process a customer. It's not a fully altruistic approach because we know that if we make our products easier to use, customers will use them and we will get paid because customers are finding value there. It is very much a design decision that says there have been barriers to success before, can we unblock them with AI tools? If you want a feature to be improved, tack AI onto it. It'll get funding. It will get funding to be included in the product.

Justin Manhart (01:08:46): I bet.

Rob Colley (01:08:46): It used to be cloud. For a while there was even this term inside Microsoft called cloud washing where revenue was being cast as cloud when it wasn't.

Scott Sewell (01:08:55): Well, you remember, what was it? Office.net? Everything was .net for a while.

Rob Colley (01:08:59): Yep, absolutely. Which was the Office 2003 release that I told you was so unspectacular.

Scott Sewell (01:09:05): We remember. It's an exciting time to be a part of it. I'm excited for customers who've seen these shiny things off in the distance and through a lot of things have been promised or said, hey, look, it's this easy. Sometimes it was, sometimes it wasn't.

Rob Colley (01:09:21): That's the whole model behind Power BI. Honestly, the reason why Power BI exists is because analysis services was such a great product, but it was hard to build models to use it. The software was great, but building the model to feed into the software was really, really, really hard, and that's what Amir set out to do was make it easier to build models so that there could be more analysis services licensing sold. Now it ends up having a lifts all boats component. When you build productivity software, you're going to make more money when people are more productive. This isn't like a gift to the world. If it was, Microsoft wouldn't charge anything for it. That's how you can tell if it's altruistic, is if they charge. When you make productivity software, it's hard to win and have the world lose.

Scott Sewell (01:10:09): It's so much fun to see what people are doing with it. Every week we have meetings with the leadership team and get to sit in and hear customers just saying, here's what we did, here's what worked brilliantly, here's what didn't. Those conversations are happening each week with the leadership team and fortunately, Arun is a guy, he's amazing.

Rob Colley (01:10:31): Really is.

Scott Sewell (01:10:32): I report up through Kim who reports to him, Kim Manus, who's amazing, and then Mark [inaudible 01:10:38] who's hired me onto the CAT team, who I would literally walk through fire for that guy. He's incredible. The interest and excitement about stories from customers is, that is the currency in which this thing that feeds the fire and we hear it. We hear the good, bad and the ugly, but the exciting part is how much is happening.

Rob Colley (01:11:00): It is an exciting time and I'm really, really happy that we had the opportunity to sit down and talk to you. Thank you so much.

Scott Sewell (01:11:08): It's been fun. Again, let me tell you how much I've enjoyed being part of your public or your audience from the podcast side, because when you say the data with the human element, it's a catchphrase, but at the same time I go, that's really the way I perceive what you're doing. The way it comes across to me, I love the fact that you guys have the back and forth, and here's the motivations, here's the challenges, here's the insights, here's what I don't know. That transparency and that honesty and integrity just comes across in a way that really sets what you guys do. There's a lot of other podcasts that I listen to that I really enjoy and I learned something from, but this one is kind of a different lane and I really enjoy that and it stands apart, and thank you guys for what you do.

Justin Manhart (01:11:50): Thanks so much, man.

Speaker 3 (01:11:51): Thanks for listening to The Raw Data by P3Adaptive Podcast. Let the experts at P3Adaptive help your business. Just go to P3Adaptive.com. Have a data day.