Is Power BI a Gateway to AI, or Vice Versa? w/ BARC US CEO Shawn Rogers

Rob Collie: Hello, friends. Today's guest is Shawn Rogers, CEO of BARC US. Shawn is one of those rare people who has kind of seen it all. On one end of the timeline, he was there when the terms like data warehouse and business intelligence were first being coined, but were not yet even remotely mainstream. And today, he's about as plugged in to the fast-evolving world of AI as a human being can possibly be. On that note, there aren't many people whose jobs make me jealous, but oh my gosh, would I love to be a fly on the wall for some of the conversations Shawn gets to have in an average month. To understand why I say that, first, think of BARC as kind of like Gartner, but focus specifically on the data segment. So whereas Gartner has to cover everything in IT from cybersecurity to mobile device management, all that kind of stuff, BARC is lasered in on data, the good stuff.

And something you might not know about companies like Gartner and BARC is this. The big software companies tell people like Shawn all of their secrets. In the case of Gartner, well, you got the sound bite thing like the Magic Quadrant, but all companies like this, all these research analyst firms make most of their revenue off of their premium paid for research that they provide primarily to non-software companies. It's just that a huge input to that research is the conversations that they get to have with all of the software companies. When I was at Microsoft, I sat in on sessions where we briefed analyst firms like Shawn's on what our revolutionary new plans were for BI, for instance, for Excel. And we knew that those same analysts sitting there listening to us that day were also privy to all of our competitors' plans as well.

Now, they would never tell us our competitor's secrets, of course, but they would tell us, for instance, whether our plans seemed competitive with what they were hearing elsewhere. Now imagine what that's like today with a space like AI, the ground is shifting underfoot much faster than at any other time in the history of software. Now, I might've found Shawn's job boring, let's say a few years ago maybe, but today with things changing so rapidly, I very badly want to know everything that Shawn knows. Well, I still don't know everything that Shawn knows of course. We didn't have time for that, and he's not at liberty to spill the secrets, but he has been taking exactly those kinds of briefings with all of the software players and he does know what they're up to. And even without the specifics, we absolutely can get a sense from this conversation of, for instance, how real this AI thing is.

And the answer is it's incredibly real. It's not going away. It's only going to become more of a focus for Microsoft and other big software vendors going forward. It is only getting started. There's a lot in this conversation to unpack and I'll let it speak for itself, but the title of this episode reflects one specific epiphany that I had along the way. With Fabric, we've been talking about how your past investments in Power BI set you up for AI and that is still true, but it also subtly implies that AI will only be an opportunistic side effect of your Power BI investment. And like I said, I think that's still true. That's still going to be true maybe even in the majority of cases, but I now also suspect that at P3, we're going to be selling and performing a lot of clean sheet Power BI work as a means of getting our clients to AI.

Now, if that sounds like I just said the same thing twice, let me clarify. We've been living in a world where Power BI, for instance, is such a massive value, such a massive payoff, such a massively transformative and high ROI thing to adopt that it's kind of hard to imagine anything else coming along that would sort of require Power BI as a setup. So for the people listening, this is I think a really important first shift in the way that we think. I think it's also going to happen that AI projects are going to be the justification and the reason for why Power BI work is done, at least in some cases. And we should not resist this. We should embrace this, all of us.

If there is business ROI in an AI project, those of us who have been coming from, let's say, a Power BI background should absolutely embrace it because what's the first thing that the AI project needs? Data, data, more data and metadata, which is exactly the collection of things that a Power BI semantic model provides. Don't fight it. Embrace it. So yeah, that changed me in a subtle way. I'm going to be thinking about things a little bit differently going forward. I'm always super grateful for those moments. I'm grateful that this podcast is part of my job and I'm definitely grateful that someone of Shawn's profile and stature shared his time with us. So let's get into it.

Announcer: Ladies and gentlemen, may I have your attention, please? This is the Raw Data by P3 Adaptive Podcast with your host Rob Collie and your co-host Justin Mannhardt. Find out what the experts at P3 Adaptive can do for your business. Just go to P3Adaptive.com. Raw data by P3 Adaptive, down to earth conversations about data tech and biz impact.

Rob Collie: Welcome to the show, Shawn Rogers. How are you today?

Shawn Rogers: I am wonderful. It's a sunny afternoon here outside of Denver and Boulder Colorado, so I have no complaints. Life is good.

Rob Collie: And that's where you live, so you're acclimated to the altitude. You don't have trouble breathing. You don't go to a hotel, you come into town and say, "I'll go around the treadmill for a little bit," and discover, "No, I can't run on a treadmill. I'm in Denver."

Shawn Rogers: I actually use a Peloton and I don't have any problems with it, but I've lived here for almost 25 years. The first six months on new residences is very difficult. Altitude sickness or illness is a thing.

Justin Mannhardt: Oh yeah, they used to take us out there for cross country camp in high school. Why are you doing this to us?

Rob Collie: That's mean. I love it.

Justin Mannhardt: Acclimating? Nah, we just go for it. They figured, "Oh, they're 15, 16-year-old kids."

Shawn Rogers: A lot of people don't know. Several of the Tour de France bicycling teams live and train here.

Rob Collie: And then you discover that Bogota is 9,000 feet. Denver is crippling. You don't notice it. You just walk around, you're fine. As soon as you start to exert yourself, you're like, "Oh my god, where's the air?" Then you go to Bogota and it's like, "Yeah," they're like, "Denver, sea level."

Shawn Rogers: Even if you live here, you notice it when you go to ski country, right? When you go up the hill bit, I get an instant headache when I've been in Breckenridge for two or three days.

Rob Collie: So Shawn, you are a long time industry insider and observer and speaker. One of my original Twitter followers, like when I was told in 2009 that I needed to get on Twitter for self-promotion purposes and I checked in on the Boulder BI Brain Trust.

Shawn Rogers: Yeah, I was a member of that group. That was a fun group.

Rob Collie: I discovered you at one degree of separation from Donald Farmer.

Shawn Rogers: Yeah.

Rob Collie: And he was still at Microsoft and then it was years later that we connected in real life and I'm like, "Oh, right, you're that Twitter avatar."

Shawn Rogers: It's always funny when people connect one to the other.

Rob Collie: Isn't it? So you've held so many roles, what are you up to today?

Shawn Rogers: Today, I am the CEO of BARC Research. We're an extension of a much larger brand of a research firm and analyst company that's based in Germany. Collectively, the company has 60 employees and there's about 25 of us doing analyst and research work and we are what you would probably call a specialty firm. We're very highly focused on data analytics and of course, AI and machine learning. We do a little coverage of the corporate performance management, CPM, BPM space and a little bit in the ESG data space, which is really interesting these days.

But primarily, our core function and focus aligns pretty well with my career because this is the space of data that I've been in for so long and the US side of our business was founded last April, so we're still kind of a new and startup entity inside of the greater BARC organization. We have offices in London and in Germany and Würzburg, Austria and Switzerland and here in the US. And then the last bit of it would be is we're driven by research primarily. So we go to market with research events and what we would call advisory services. So we help end users with strategy and we also work with vendors for their strategy as well. And so that's kind of what I'm up to these days. So it's definitely keeping me busy.

Rob Collie: A couple of household names people are familiar with in this space. [inaudible 00:09:21] Gartner at you. What's the difference between BARC and a Gartner? How do you compare and contrast?

Shawn Rogers: In the analyst space, there's sort of three tiers. There's these tier one companies, Gartner, Forrester, IDC, and I think when you talk to vendors and end users and you ask them about what their strategy is for advice and research and those guys will almost always be part of the conversation, but then the specialists or the tier two analyst firms that have this high level of focus, we exist quite well in the ecosystem because there's a lot of benefits from doing business with companies that have specialty expertise versus the broader ones. And then lastly, there's a lot of individuals in this marketplace, individual experts and influencers and so on, and then a handful of media companies that are doing more marketing services. We don't do that at BARC. We're very disciplined around making sure that our research stays unbiased and that we stay very neutral when we approach topics.

We have some products where we do discuss the vendors openly and score them accordingly. And then we have a lot of research projects that are end user driven and I have an affinity for those. So I like those quite a bit. There's a lot of good companies out there doing work. BARC has some notoriety because of the age of the company. It's 30 years old. As I mentioned, the US side of the business is much newer, but the company itself has deep roots and has always been focused on data and data management and analytics, and I think that positions us kind of well. So I find myself with my customers, I'll often have a relationship with one of those other firms, but they also have a relationship with us.

Rob Collie: Someone like Gartner covers everything.

Justin Mannhardt: Literally.

Rob Collie: Yeah.

Shawn Rogers: They're very broad in their approach to markets.

Rob Collie: That makes sense to me, the specialization thing. Now, BARC had existed for a long time in Europe. I'm assuming that the motivation for opening offices in the United States is to reach customers, clients, et cetera, in the United States, or are the specifics of data just a little bit different in North America versus Europe?

Shawn Rogers: The core of the first question about reaching customers in the US, I would twist it a little and say we're augmenting the relationships that existed. So BARC has got a global viewpoint, but now with the offices here in the US, we're more global geographically. It allows me and my team here to do things that aren't as easy to do from over the water. We have a much deeper footprint face-to-face, especially nowadays when we're all back traveling and doing things face-to-face and it allows us to interact with our customers on the correct time zones and so on. So BARC had and maintains relationships with some of the biggest companies in the world. This allows us to serve our clients a little better by being on the right time clock and geographically perhaps in the right place. And so far the feedback from our customers has been very positive on that side, it's great to have someone here locally that we can work with versus hopscotching around time zones.

Rob Collie: Time zone has become more important than actual geography.

Shawn Rogers: Yeah.

Rob Collie: If you lived in South America but you're still longitudinally at the same time zone, that makes a big difference. You can be a greater distance away in terms of straight line miles, at the same time maintain that crucial time zone thing so you can overlap with your customers North America executives, et cetera.

Shawn Rogers: I agree. I'm on Mountain Time as mentioned, and I'm still behind my team because my US employees are all on the East Coast, and so they all get a two-hour jump on me, so I always feel like I'm a little behind when I start my day.

Justin Mannhardt: Yeah, you got two hours of questions and updates.

Shawn Rogers: Everybody's queued up.

Rob Collie: Then you can be like Justin who wakes up around the time I'm going to sleep. He's unaffected. Unaffected by time. So we could put him in Hawaii and he'd still be synced up.

Justin Mannhardt: Sure. Rob, that sounds great. Coming right up, Justin to Hawaii. Rob approved.

Rob Collie: Please. A little bit more on BARC. What is the average customer of BARC? How big are they? Do they come from different industries? I'm not asking for specifics obviously, but just sort of give people kind of a ballpark feel for the kinds of people who do come to you for help.

Shawn Rogers: It's kind of a multi-part answer. So again, remember we're highly focused on a few topics, so all customers fit into that coverage area. They want to speak with us or consume our content and our research or come to our events that are about these topics. And then I would split my customers kind of 50 50 in the sense of we do a lot of work with vendor customers and the vendors want to have a presence at our programs and events. They want to attach themselves to some of our content and our research and those customers, those vendors, again, are highly focused in data and analytics, AI, machine learning, and then size-wise, very different. We do a lot of really fun cool things with small startups in the space. And then, of course, we work with the hyperscalers and the really big industry logos that you would recognize immediately. So doing business with Qlik or Tableau or working with IBM or AWS, all of those firms come into our purview.

Now the other side of my customers are the end users. We tend to divide them up by industry. We have people in most major industries that work with us and they work with us in similar fashions. They consume our research and pay for it. They attend our events and generally pay to be part of those as well. And then on the advisory side, they pay us for our consulting expertise. And then to be clear, we don't deploy or install software because that crosses that border for us. And then those customers, Rob, are of different sizes as well. Some small companies come to us and say, "Look, we have 112 people, but we need to dive into the AI pool. Can you help us with that?" And at the same time, a huge pharmaceutical company might come to us with an extraordinary diverse, complicated data environment and they will ask for our input and our advisory services to help guide them to a good path.

Rob Collie: I always forget about the relationships with the software firms, which is weird. I used to be at Microsoft and had relationships with companies like yours, but I'm always really fascinated by who you call, I believe, end users. Those would be sort of overlap at least partially with some of our clients, at least in demographics and the types of questions that they come to you with. Today versus five years ago, it would seem to me that AI is now a very, very, very high percentage of your public work, public appearances, your speaking, your LinkedIn posts, et cetera. We're under this pressure as well. Everyone has to be talking about AI. We, P3, don't want to be known as a company that doesn't do AI because we do. Does the behind the scenes and the front office, do they tend to line up or is there still a lot of "boring non-AI" stuff going on behind the scenes while in public we're all forced to play the game of it's all about AI all the time?

Shawn Rogers: I've not had a business conversation in the last four months that did not include the word AI. It has polarized the world and a lot of funny things have happened. So I get to talk to hundreds of vendors per year. So when I speak to software companies, we get briefings and we understand their roadmap and oftentimes the company will share their roadmap under NDA with us so we privately know what they're working on, where they're heading. And 18 months ago at the end of November heading into December, you could hear the brakes being hit on roadmaps all over the world from tech companies as they screeched to a halt. And then the next sound you heard was them turning the steering wheels directly in a new direction towards AI. In the beginning, early on in 2023 especially, it was mostly people waving slides talking about what they planned to do, where they're headed.

Don't worry, we know this is happening, we're on it. We're going to show you some cool stuff soon. And as an analyst, it was like the standing line at the Boston Marathon. Everybody was going to get into the same race. Everybody was there, but they were all very clumped together. Now 18 months later, we're starting to see it spread out, again just like a marathon. There's some runners that are out ahead and they're delivering technology and a lot of people are still lagging behind on, we're almost there, we're getting there. Hey, private previews soon, look for a beta in the fall, that kind of stuff. So the more agile, more adaptive companies are certainly starting to deliver. But to the core of your first question, has the complexion of my interaction with users and vendors changed? Oh yeah. And to be honest, my day-to-day work has changed.

I've published a 539 person piece of research on AI. At the beginning of '24, we did another one on optimizing your architecture for AI innovation and that was about 340 respondents. I think the last thing I'll say in my monologue answer to your good question is I've been in the industry since the mid '90s and this is the first disruptive technology, and I'll say Gen AI because all of us know AI has been around for a long time, but Gen AI 18 months ago, it jumped the chasm and it quickly became a board level mandate and we have research that proves it.

The boards are asking their executives, their ELT, "What's the plan? What products do we have? How are we going to be competitive? How are we going to crush the competition? Are we behind?" All of those things. 37% of our surveyed respondents said that their boards, and by the way, when we asked the question, we only let ELT titled individuals answer it, and we said, "So we know you talked to boards, what are they saying to you?" And one of the takeaways was 37% of the boards talking to ELT are exploring new funding, higher budgets, additional investment, those types of things. It's really got the attention of everybody. So it's a great question, but yeah, it's AI every day here and occasionally, I get to have a nice constructive conversation about intelligence and dashboard, data management and so on, but in short, it's AI.

Justin Mannhardt: Shawn, I'm curious, you describe the core focus of your firm's research, data analytics, AI and machine learning. Three of those words, at least historically, I would put in the same bucket and say, "Yeah, that makes sense, data analytics, machine learning. I see how that goes." And you said this very well, when people say AI today, the first thing that pops in their head is chatGPT. And so the generative technologies have created a huge hype cycle that we're all experiencing. So when a company comes to you and says, "Hey, Shawn, can you help me with AI?" In the context specifically of data analytics and machine learning, where are the real applications for these generative technologies in the analytics domain? I'm sort of curious, some of your insights are on that idea.

Shawn Rogers: We're seeing a lot of cool work in that particular arena around the BI and analytics vendors putting models into their environment to describe what the chart means, to help with text commentating to help catalog what data management process is doing. Instead of having an individual say, "This particular data pipeline is doing X and it's feeding this particular dashboard or analytic," now you can use language models to pump that stuff out without a lot of effort. And it's also the type of workload that makes sense where even the hallucinations aren't the end of the world right now because, let's face it, there's accuracy issues, but these are the types of workloads where we're seeing just a lot of good value-based kind of work. And I've had the privilege of being with three or four of the leading BI firms or what we would normally categorize as BI firms, and you're seeing sort of continuity in that. There's a lot of great content amplifying what exists for BI and analytics consumers already.

The dashboards and the interfaces are getting a lot smarter and it's kind of cool. I see some that are also using it to automatically give you better transparency so they're building automatic lineage into the view of the insight to say, "Here's where it came from, here's where the data was and so on." I think it's really cool stuff. Outside of that just a little bit, the starting point for a lot of companies is, "Hey, let's get an AI chatbot online." But if you flip the coin over and ask for more advanced stuff, things like risk management and recommendation engines and fraud analysis, which are applications that have been around for a long time out of the end users and the vendors alike are seeing opportunity in augmenting and amplifying that capability as well. So not a new application, but let's look at our risk management work and see where we can infuse AI to make it more automated, smarter, more accurate, and so on where it's early days for the use case stuff.

Justin Mannhardt: So early. The detailed description you gave Shawn of augmenting the user's experience with a dashboard or doing things like documentation, explaining what's going on. I think a lot of times people see the hype with generative AI. They say, "Oh wow, it made this video or it made this song, or it wrote this thing, or it passed the bar exam or it diagnosed these diseases," right? You hear about all these exciting things and then you say, "Well, can't it deliver me wonderful insights from my data?" I don't know if I've seen that technology quite get to that level yet. I'm just sort of curious what your experience has been like.

Shawn Rogers: One of the use cases I'm starting to see more and more what I would call it is causal analysis. So you have this great dashboard and it says sales are down. There are a lot of people that struggle with data literacy. In general, they see the chart, but having a hard time kind of absorbing what the chart is telling them. I do think that AI and Gen AI can play roles there and I'm seeing it that says, "Hey, your sales are down, and by the way, based on the data, here is likely three things that are causing your dip." Now you're getting pretty interesting because I have met a lot of C-suite people. Honest to God, they don't understand the data they're looking at. They get this dashboard every day and they glance at it, but they don't really understand the deep whys of it. And when AI is involved, one of the whys might be shipping latency in your supply chain.

Okay, can I dive deeper on that? And AI is enabling this right now and I'm seeing it in the real world through some of the vendors who are starting to provide, not only can I tell you what's likely causing your headache, but if you click on this, I'll show you some information and I'll allow you to dig down to the fact that XYZ Trucking is having significant issues in the southern region of your company and you have a stock problem which is making your revenue dip. That's harder to do in more traditional dashboard environments. Now, it's not impossible, and we all know that because companies were using more traditional AI and ML to surface that, but I think some of the new things that are happening right now are going to make it more democratized for more users because it's more plain English.

Justin Mannhardt: That's great.

Rob Collie: One of the things we've talked about a lot really for years is what I call the data gene. It's really more of a curiosity thing. You're either wired to be curious and want to dig around and sleuth or you're not. Most people aren't. Most people aren't wired to look at a chart and immediately start seeing the matrix, if you will, and understand it. "Oh, you see what that happened there? Okay, let me drill down there and confirm that hypothesis." Most people aren't wired like that, but a lot of people who aren't wired that are still very, very, very good at their job, but it still to this day kind of blows my hair back a little bit, not as much as it used to when I encounter someone who is a titan of industry who's not interested in the chart the way that I am.

The idea that Gen AI could help install an artificial data gene into important people and help them do the things ... Because you're right, what you and I might look at and go, "Oh, look at that. It almost just jumps off the page." It doesn't jump off the page to everyone necessarily, and even that's a sliding scale because as the chart becomes denser, as the dashboard becomes denser, the human ability to spot the pattern breaks down pretty quickly. In fact, the previous episode of the show, Justin and I were talking about exactly that. I'm looking at charts and graphs and even just tables of Power BI produced data from my personal life with so many variables that I can't tell if there's a sweet optimized solver zone because it's already beyond the four or five things that a human brain can keep in its head simultaneously.

Like now it's at eight and it's exponentially more complicated than my poor biology brain can absorb. Everything's got its limit. You use the literacy metaphor I was thinking about, "Yeah, I'm literate, I can read," but I've tried to read Ulysses by James Joyce like five, six, seven times, never made it past the first chapter. Maybe that's another thing I should do. I just take that whole book and just dump it into chatGPT and say, "Give me the hundred page-"

Shawn Rogers: Summarize this for me.

Rob Collie: I don't want a summary. I want to read a book, but I can actually read it. I want to experience Ulysses, but not in the dense way that he wrote it. So these are all really fascinating things that we can expect to start to see from the software vendors. I love the self-documentation thing because, I mean, oh my gosh, who likes documenting?

Shawn Rogers: Well, and it's the same with coding. We asked a question and a piece of research where we said, "Are you looking at specialized models?" We know that they're looking at the big models. We get that, and we asked that question too, and it wasn't especially interesting, but finance models, models that are full of specific financial information, we're going to live in multi-model environments. We're not going to live in a single, I've got Open AIs chatGPT inside the walls of my business, or we're using granite or what have you.

We're going to have dozens if not more models in our environment, and that's going to help us more finely tune the analytics and give us a much deeper level of insight into things, which I think is a very interesting part of it. It will allow us, along with literacy, my brain thinks about prescriptive. It's going to prescribe to people that aren't quite as literate where to go, where to find it and what's actually happening. The big problem will be accuracy. I visited a leading CPM vendor a couple of weeks ago here in the US and I went to their event and of course, corporate performance management and FP&A people, their level of interest on accuracy borderlines on hysterical.

Justin Mannhardt: Yes.

Shawn Rogers: It's dollars. Come on. And the vibe I got from their really cool users on AI, very different than other businesses.

Justin Mannhardt: Sure.

Shawn Rogers: Accuracy is paramount and they know accuracy is still an issue and it's going to be for a while. My biggest piece of advice has been keep calm and carry on.

Justin Mannhardt: Yes.

Shawn Rogers: Everybody is afraid they missed the boat already, and I do give a lot of talks and I've talked twice in the last couple of weeks at big events, and that's one of my opening slides. It's like, everybody take a deep breath. You did not miss the innovation hype curve. It was 18 months ago this chatGPT thing started. It's okay, relax. But for here the US phrase is FOMO, this fear of missing out and a lot of people walking down the hallways at work and they're getting collared by some ELT member going, "Are we almost done with AI?"

Rob Collie: Yeah.

Shawn Rogers: We're not quite done with AI yet.

Justin Mannhardt: That's funny because there's a parallel even with BI. BI isn't something you finish, it's a part of how you operate your business. It becomes integrated and it's always evolving and it's always changing. There's this mystical finish line with AI all over the place. I find that fascinating. But yeah, I was thinking about Shawn as you're explaining some of the places where generative AI is sort of penetrating the analytics space. I had a sense early on it's like, "Okay, what are the easiest things for this technology to infiltrate?" It's that end user experience, producing charts, producing graphs, directing traffic, so to speak. I love the way you described that. I think if anything, it further emphasizes the importance of getting your company's data models clean, accurate, correct, useful, enriched so that that technology can actually do something useful with it.

Shawn Rogers: So that one's funny. So I was asked for a summation the other day, "What's your parting piece of advice, Shawn, for companies that want to innovate with AI?" And I answered it with data, data and, well, data because if you don't have the data foundation in place, it's not agile, it's not clean, it's not high quality. If you have turned a blind eye, which many companies have to their data projects and what happens oftentimes with data projects because they're not nearly as fun and exciting as perhaps AI or doing big data or cloud or what have you.

What companies often do is they'll get distracted by the bright shiny toy of whatever new tech is out there and they'll go, "Well, let's put that MDM project on hold. Let's move budget from here to there and let's go chase this one." And a lot of companies have now come face to face with that demon realizing that, "Our data house is not ready. The only data we could actually feed into the environment is standard structured data. We're not ready to deal with semi or unstructured. We don't have an architecture that supports streaming or time series data. Are we in trouble?" And my answer is, "Yes. Yes, you are. Figure it out." You can't do the fun things without doing the hard things. You can't have one without the other.

Rob Collie: What role does business logic play? I thought your joke might've ended data, data and metadata.

Shawn Rogers: Thank you. I'll steal that moving forward. Thank you.

Rob Collie: The third data did in fact surprise me. In our space, this notion of a semantic model is more than structuring the data. It's more than just cleaning the data. It's also infusing it with meaning.

Shawn Rogers: Context.

Rob Collie: Yeah. The relationships between tables, the formulas that calculate the metrics. There's real meaning and enrichment happening at that level that isn't just data. I don't know if we would call it metadata. Maybe we would. From my perspective of having come up along the path that I have, to me what Microsoft now calls the semantic model, like all that model that was actually buildable, that actually worked and that actually would evolve and move at a tempo that actually met the business's needs was a miracle.

Honestly, Shawn, it's the only miracle that I've personally witnessed in the software world that I am really, really grounded on. I am such a hardened skeptic and seeing what power pivot could do that turned into today's SSAS, Tabular, et cetera, was one of the few times that this cold, dark, skeptical heart warmed ... My Grinch heart, grew three sizes that day, I decided to bet the next chapter of my career on the whole thing. How much do you, in those conversations with end users in particular, do you get into the notion of the business logic and that semantic meaning layer of structuring the data estate?

Shawn Rogers: So I believe, much like you, that it's a conversation that should be taking place front and foremost. That's kind of what I mean by data, data, data. Unfortunately, I have this phrase, immaturity dictates priorities. The priority right now isn't so much about what do I need to have to do it? It's where do I start and how can I do it right away? By the way, what does RAG stand for? I can see it in my research on a lot of different levels. I'll give you a couple. We asked at the top of our research after qualifying the respondents, and I want to make that point, no one got to take our research unless they were able to inform us that they had a strategic view of their strategy. And we asked them seven big questions. I'll try to rifle through them real quick, but it starts with security standards, data access and use policies, legal considerations, AI program standards, project governance and enterprise architecture requirements.

So I did it fast, but do you know which one came out at the bottom? Project governance and oversight because no one's super interested in figuring out the deep way to do it right. Everybody's just excited about getting in the game. And so I think, to your point, Rob, the conversations that you're talking about because you've been in the industry so long, you know that this is where the problem and the opportunity lies. And we'll get to talk more about that later in the year and into '25 as everyone starts asking the next level of questions, which is I can't get it to do what I want it to do, or it lacks the context or it's not accurate, or we don't know where it's coming from. We ask another set of questions around responsible AI, which you guys know that ethical mandate that you put in-house, how are we going to do it?

How far is too far? Where are the guardrails? So on and so forth. And at the bottom of the priority list, the bottom third, accuracy and model bias. Are you kidding me? How can that be a low prioritization? And it's because they're worried about other things like PII and staying out of trouble with the EU AI Act, and I think collectively as leaders in the industry, whether it's analysts or vendors, we're supposed to be two steps ahead of everybody and are thinking, and Rob, your question is three steps ahead of everybody where they are today. They'll get to you and they're going to arrive there and they're going to need help.

Justin Mannhardt: Yeah, it's sort of ironic too, right? Because talking about the prerequisites to setting yourself up for success with AI.

Rob Collie: Accuracy in the bottom third. Oh, nothing chills my blood. A tragic and perhaps unfair example, but this reminds me of the Boeing MCAS debacle. More important to get it out there was the number one concern not having to recertify, right?

Justin Mannhardt: Yeah.

Rob Collie: The bottom third accuracy. No, no, no, no.

Shawn Rogers: And the damage that's done to their brand is remarkable. I had a subconscious thought watching the rocket launch the other day after having two scrubs, mechanicals, and I kept thinking, "I'm not sure how I'd feel sitting on top of that stack with the way they've prioritized certain things recently."

Rob Collie: You're going into space, you're expecting there to be some risk. What you don't want to be doing is thinking about their civilian airliner track record.

Shawn Rogers: Well, it's like the military joke where you're jumping out of an airplane with a parachute that was made by the lowest bidder. It's a scary proposition. I will come back to that big list of seven things, and we had this moment while we were doing the research, we were looking at how everybody answered the question, and my research partner, Merv Adrian, a recently retired Gartner analyst who's a very good friend of mine and a partner at BARC, he and I did this research together and Merv says, "I wonder if anybody actually said they had fully deployed all seven of these concepts." And I went, "I doubt it." We went back and asked the research team to do a little cross tab for us, and we found out that there is a sub cohort of advanced users, people that are leaders that are latched onto this a little earlier, had a bigger foundation and it comes out around 20%.

Justin Mannhardt: Interesting.

Shawn Rogers: 80% of us are in one basket and there's a small cohort, 20%, and we actually didn't believe the finding when we found it because we thought it was highly improbable. So we crossed it with every question in the survey and we could see that they behaved differently on just about every question and they were asking in a more mature fashion. So the people that we'll get to talk to about metadata and semantic layers and contextual data and bringing that into models, that's the 20% and you'll get to talk to them before the other 80 for sure. We call them a high readiness group and they behave differently. I'll make another point the audience might find interesting.

We always ask, "What's the big challenge?" And like 39% generically speaking of the answers came across as we need new skill sets. It reminded me of finding a data scientist a few years ago. It was impossible. They're overpaid. If you want a raise today, change data scientist to prompt engineer and just wait for your phone to ring. You're going to make a lot of money. But what we found with that high readiness, that special cohort, they weren't concerned about that as a challenge anymore. They had moved on to cost. They're concerned about how expensive this is. Whether you're doing it with a hyperscaler or you're trying to fine tune models in house or what have you. This is suddenly presenting a stark cost issue for people. I think that's interesting. So that 20% has arrived at that number before everybody else, and I think we'll hear a lot about that this year too.

Rob Collie: That's fascinating. I just realized that I'm now going to be changing roles in the conversation that I used to have, so the conversation used to go something like this. We have it at P3, like this concept we call Faucets First. Ignore AI for the moment, just talk about BI. Instead of sitting down, first, let's get the data warehouse and then someday, eventually we'll start getting to charts and outputs and things that you can actually see. We go the opposite direction. We sort of start from the business requirements and try to get something in front of the stakeholders as quickly as possible, regardless of how temporary the data connections have to be. It's okay to dump a bunch of CSV files or whatever just to even prove out that there's value there, and then eventually your back end plumbing improves over time and iterates to meet the needs as opposed to the plumbing first.

That's really core to our philosophy. And I give talks like this at places like Professional Association for SQL Server, and there'd be ETL people, there'd be data warehousing people in the audience that thought I was telling them that they were obsolete. I would be instead saying, "No, it's just the sequence is going to go away. The sequence is going to change." And what I would tell them is wherever we go, wherever P3 goes, ETL follows. And in fact, people are much more likely to sign up for it, that boring ETL data warehousing work because they now can understand directly what the ROI would be.

You say, "Oh, I get the same chart that I get, but now I get it five times a day, updated as to once a week." Or you can always give people a really clear business improvement speech as opposed to the data warehouse, which was a project that just sort of wandered on its own as if it was its own mission forever, and for years just occurred to me in something you were saying like, "Oh no, now I'm that person," that where AI goes semantic modeling follows. Oh, no.

Shawn Rogers: Well, it aligns with that comment I made about maturity dictates priorities. It's the same sort of idea. It's let's not see how much we can do. Let's see if we can curate what we have to be more valuable or more current, like you said, five times a day versus an overnight batch and that type of thing. I think the end users are starting to figure that out and the sophistication levels are there. Plus it's not as financially startling as it used to be. We used to write articles. I remember the 5-50-5 rule of data warehousing from way back in the day. Forgive me people listening if I have these out of order, but the five stood for it takes five years to build a data warehouse. The 50 stood for 50% of them will fail and not find their way to production, followed by the five, which was it was a minimum of five million bucks to go do it.

Bottom line is it took a long time, it cost a lot of money and it didn't always get over the end game. We've moved beyond that, but here I am on this great conversation today and my answer for doing well with your AI was data. We still haven't fully accomplished that we're not leveraging the data the way it could be. We don't have it in systems where it's ready for AI. The best models are going to be the ones that are influenced by enterprise information. When you start feeding enterprise data into your models and not exposing that publicly, your chatbot gets smarter, your analytics gets quicker and smarter. You're going to be able to get a lot of advantages from sharing your data into these smaller models like finance or medical or what have you. So the data conversation is kind of old.

I mean, I owned a magazine. We talked about my career a little bit in the beginning. I was a co-publisher and co-owner of DM Review Magazine, data management review, and it was a print publication that about 75,000 people here in the US read and we had some of the world's greatest, biggest, coolest experts in it, the biggest names in data warehousing. Bill [inaudible 00:44:03] was on our advisory board and you open that magazine and there was a column from everybody. And I am sad to say that a lot of the conversations I have today, I could hand someone a 1996 version or '93 version of my magazine and go, "You should read page 12."

Justin Mannhardt: That's so fascinating.

Shawn Rogers: And I believe strongly that the tech is really caught up. There are vendors out there right now where you can go spin up a data warehouse in a very short time. You're not going to fail and it's not going to cost you five million dollars, and we know who those guys are and they're making a lot of really cool noise in the industry, but now the discussion is, is how do you leverage that for your AI work?

Justin Mannhardt: And some of the conversations I've had with customers or just other people seem to be thinking about leveraging AI in two major categories, but one would be let's infuse AI in some way into our product and service that we provide to our customers. The other is let's infuse and integrate AI with the way we operate our business in-house. And I think about some of the customers we have here at P3. I've described some of them in the past as like they're just regular old companies. They make stuff, they make steel beams and boards and houses. You're not going to infuse AI into a steel beam, right? So I'm curious the 20% that are having success, what are some of the lessons learned or insights about how people are deciding specifically where to invest their AI efforts in terms of either internally with their product space and maybe even some cool things you've seen?

Shawn Rogers: It's a good question and the points you made along the way and the question are pretty valid. I like the way that you spoke about both, right? We want to optimize our internal processes and functions with AI and then we want to optimize how we work with or integrate with or interact with our customers, and I do think that both of those paths are the two main paths that we see. I'm not seeing clear strategies for a lot of specific use cases other than the basic ones. I'm seeing more strategies towards those are the two forks in the road that we want to follow, and now we have to have this conversation about how do we augment what we already have technologically speaking and what else do we need? We're going in those two directions. I think there's basic things like recommendation engines, so a great inner way to interact with your customers. It's been around for a long time, a lot of machine learning and traditional AI behind them.

Justin Mannhardt: Sure.

Shawn Rogers: I think they'll do what I said earlier. They'll look at their recommendation workflows and logic and they'll see if putting a model in place is better or worse. They'll see if they can score higher with this process. What we're going to see a lot of change around I think is vector databases being in environments, but I think those are the things we're going to see initially and then we're going to see complete oh my goodness brand new monster innovation thinking. We asked a question about what parts of your architecture are you going to maybe overhaul or add or amend, and it's everything from model ops to knowledge graphs to better auto ML environments. A lot of companies are investigating, again, data fabric and data mesh to address that other problem.

Justin Mannhardt: Data cashmere honor roll mention.

Rob Collie: Let's go with data polyester.

Shawn Rogers: Yeah. We asked a question, in the last piece we said, is this the moment in your time where you're going to blow up what you've bought or are you going to retool? Is this the moment? And the answer was no. Resoundly no.

Justin Mannhardt: That's very reassuring insight.

Shawn Rogers: I kind of thought the hyperscalers had a compelling argument. Move everything into our data lake. Use all of our highly integrated tools and environments. By the way, we have foundational models. We own them. We'll make them available to you. They are getting a lot of customers that are doing certain things with their environments. All the hyperscalers are reporting increased revenue through AI solutions, but the user community has said that might be part of it. That might be part of our strategy, but we are going to augment and we're going to fill gaps, and that opens up this really cool conversation for companies that have built an environment that can be plugged into an existing architecture without overly disturbing it and bring in all the toys. We're starting to see a lot of really cool things. Take a look at what DataIQ is doing today. They're doing that and they have a lot of governance built into it because we haven't bored the audience talking about the EU AI Act and compliance yet. The penalties are pretty treacherous if you screw up with AI.

Justin Mannhardt: Interoperability, if you're going to play in the data platform space seems like table stakes. If you can't play nice with other platforms and other products, you've created a barrier more than you've created any sort of concept with the customer here. I like what you said, Shawn, about things like vector databases or graph databases. We haven't really had, my experience anyways, like a mass in-market technical capability that made that attractive for a lot of people, so I know we're using technical terms.

The reason you would use a vector database is because you want to take something like your company's documents or recordings or visual assets or other unstructured data media and process it in such a way that it could be retrieved by something like a large language model, like a chatGPT type thing. People that were doing vector databases before then they had a whole different set of needs and use cases that weren't broadly understood yet. We did something interesting where we took some of Rob's blogs from back in the day when he was blogging all the time. We fed them into some of these things and just little in-house project, right? Those use cases are very in range for people now. I'm not trying to shove that information into my data warehouse. I need something a little different.

Shawn Rogers: All three of us live through the big data craze, the Hadoop thing, and I remember bumping into people who had the opinion that Hadoop was going to replace data warehouses. There will no longer be relational databases in the enterprise. Your data warehouse is going to disappear. Everything will be in this big thing and you're going to distill stuff from the big thing and that's going to run your business. That did not come true. Unstructured data and Hadoop and Spark and so on have found homes within the architecture.

I think the same is true about vector databases and that was the path they were on. When the brakes screeched and the wheels turned, the vector guys were also, you could hear the party they threw because they knew the market had just swung in their direction. All of a sudden you had vendors who were using vector technology but weren't talking about it. Embedded in what we have, we're using a little vector and all of a sudden it's like we have Vector. But yeah, every Vector database company changed their slide deck on December 1st, 2022.

Rob Collie: Shawn, I'm sitting here listening to you describe what you do on an average day and the interactions that you get to have, and I just have to ask you, how much do you love your job?

Shawn Rogers: I get paid to learn. If you distill my job down, I actually get paid every day to learn stuff, and if you're inquisitive, Rob, you said it earlier, there's data people and then there's really data people. We're inquisitive. I've never seen a dashboard I don't want to drill down through. My mantra is bring data, be gone. I don't really want the opinion. Just bring me the data and let me make my own decisions. I get to spend days wallowing around in white papers and academic research and trust me, there are a lot of analysts that are a lot smarter than I am, especially when you get into the deeper technical minutia. And I have a deep amount of respect for those folks that can give you three hour overview on how a vector database actually does what it does. I focus more about helping business people understand not so much how to do everything but why.

And the why to me is strategic and I like getting to do that, but yeah, I get paid to learn and then I get paid to be the center of attention and give speeches and write papers and all that kind of stuff, so it's a very fun job. I went through a transition in my career a while back and I had a really difficult time deciding did I want to go back to being an analyst or did I want to continue down this marketing path that I had had a lot of fun on.

In the end, the analysts work one out and it's because of that core stuff. I gave a talk in Amsterdam in October on AI, and one of the things I said sort of ad hoc from stages, I'm a little angry that this is happening at this point in my career. Damn it. And I really wish that this startling disruption of AI and Gen AI especially had occurred a little earlier because I just think it's going to be a cool ride to be on. I've never had to work so hard to stay current in my entire career. This is morphing and changing so fast every single day. I can't keep up with the velocity of news, but that's why I love my job.

Justin Mannhardt: That's such a great message for people that are early on in their career or in the middle of it. I mean, you've heard Rob describe his experience 15 plus years ago with BI and what happened there. What's happening right now is seemingly so unique. We talk about FOMO and FOBO, embrace it, ride the wave. This is a really exciting time if you're in that stage of your life.

Rob Collie: I agree.

Justin Mannhardt: Wisdom, Shawn, wisdom.

Rob Collie: The reason I asked you how much you loved your job is because I had to turn the thought in my head into a question. The thought in my head was, "Oh my God, am I jealous of what you do every day?" And then in your answer, you hit on exactly, I think, the theme that was so compelling to me is that I'm not someone that enjoys sitting down and playing with some new piece of technology. Let me go build this kind of new database. Let me go write some Python code. I just don't get excited about that.

If I immediately have a real direct applicable purpose for something and I'm already at N out of N plus one steps, I'm more than willing to go the N plus one step, but having access to such a broad cross section of the cutting edge, both from the software vendor's perspective and from the corporate consumer perspective, the things that I enjoy the most and always have about my career is the synthesis of what's actually going on and putting things in perspective. Even for myself, just learning for myself, getting my own model of what's going on straight and then learning how to communicate and share that model with others. That is the core passion, and I'm sitting here listening, I didn't expect this, and I'm like, "Oh my God, Shawn's got the job."

Those habits, those inclinations of mine are how I found that I wanted to do this company P3. It wasn't P3, and then becoming an observer of what was going on. It's always been that drive to understand when I saw that the thing I was trying to piece together was going to be a hole in the market, well, let's go build the company that fills it. I find myself, because I'm not working for a software company anymore, I'm not working for one of the big vendors. I'm not being exposed necessarily to what they're thinking like I used to be. I'm not getting that six, 12 month preview like you are. I'm not getting all the information that I would crave. Just listening to you, I'm like, "Ah, damn it, Shawn." I didn't expect to leave here with a new itch, but here it is.

Shawn Rogers: It's a very interesting entree to the industry because you do get to speak to all of the vendors in the space and you get to hear the future plans, and you get to speak to small ones who are really oddly taking really cool and disruptive approaches, and then you also get to watch what the really big ones are doing strategically, but you get to do it in a first-person sort of way without working for each of them. You get to interview their CEOs and then along the way when you're interacting ... I was at an industry event put on by a software company a week or so ago, and they were kind enough to set up a couple of one-on-ones with their clients. So I got to sit down with private premier clients and go, "Well, what do you like about the software? What don't you like about the software? What's your AI strategy look like?"

So the ecosystem that analysts in general live in is this level of entree is a very cool opportunity, so you don't want to squander it. So it influences everything I do every day. When we write research, it's influenced by my last six months worth of conversations with both ends of that spectrum. I know what some of them want to learn. I know what agendas some of them have. I want to prove that it's right or wrong, and I think the best analysts, and I'm not saying I'm one of the best, but I'm saying I have those two tools. I have a long, long tenured experience coupled with data and I think they balance each other out, which is cool.

Justin Mannhardt: This is a great setup, lots of experience, lots of data, lots of insights. You've got a seat at the table with what the vendors are thinking. You've got a seat at the table with what customers should be thinking about and are struggling with as they prepare for this AI. We showed up a little over a year ago, 18 months ago, and OpenAI was going to steal all of our jobs and then two weeks later, I think Sam Altman even said this, two weeks later, everybody was complaining about how slow it is. And so here we are today. What are the leaps over the next 18 months? The breakthroughs both in technology and companies getting sized up for these kind of capabilities.

Shawn Rogers: I do think in the next six months we're going to focus on failure a little bit as we often do. We build something really big and tall and then as a community we tend to knock it down for a little while. There will be some end users, customers that help us with those stories. I have this running joke about I'm going to write an article about failure this year. Don't be in Shawn's article. Please don't. I just said it on stage the other day. Don't be in Shawn's article because it's going to happen. So we're going to deflate it a little and then we're going to get more serious about it. So the silly hand waving and hollering from stages is going to die down because the separation will occur. The ones who are delivering will continue to deliver. The ones who struggled are going to go by the wayside.

They're going to fail. I'm not wishing that on anybody, but we're going to see some companies completely miss this opportunity. So there'll be sort of a resetting of the table going into 25. I think the use cases will become more sophisticated and then I'll answer the final part of the question that you didn't ask for this. In 10 years, we're not going to talk about this topic anymore the way we're talking about it. I am going to be retired. AI will be part of our ecosystem. It will be part of what we do, part of what we leverage. It'll be a tool in the toolbox that we reach for at the right times, for the right applications. Accuracy, hallucinations will disappear, not entirely, but pretty close and it's going to drive a lot of business and then there'll be this next wave of innovation using something that's really super stable and proven because right now it's neither.

Justin Mannhardt: Love it. It fits the keep calm, carry on lines that you shared earlier. You described a future, a decade out and it's like, "Yeah, this all is moving very fast and it's going to continue to move fast and it's going to continue to evolve." But what helps, I think me and the way Rob and I talk about it sometimes is I think about the last 10 years of my life and I didn't just all of a sudden wake up one day and start doing everything differently. It was a process and that's a good reminder.

Shawn Rogers: Stay with the process.

Justin Mannhardt: Yeah. And it will become a part of what we do both as a society and as businesses.

Rob Collie: I hear a few questions in ascending level of difficulty. First of all, what album is on the wall behind you today?

Shawn Rogers: I took the album down because I thought we were doing a video and I usually take my albums, so for the folks listening, I listen to vinyl record albums while I work. Rob knows that. I often hang them on the wall in a frame so that the outer jacket doesn't get dinged up. Everybody's about to laugh, but I have been listening to Taylor Swift's newest album.

Rob Collie: On vinyl.

Shawn Rogers: On vinyl. It may not a stereotypical fit the tastes of a guy who's 60. Generally it's classic rock. A lot of Rolling Stones and Led Zeppelin get played in my office, but I'm enjoying the music and I enjoy the rhythm of it and the poetry behind it. I have listened to all four sides multiple times.

Justin Mannhardt: Hell yeah.

Shawn Rogers: I just put it away in the rack before our call today.

Rob Collie: I love that Vinyl has now become this way that artists still get to sell their music.

Shawn Rogers: You and I talked about this once when we were catching up a while back and I said, "I am slowly recollecting the record collection I had when I was 18 at four times the price."

Rob Collie: That's right. Got rid of them all. Got CDs. Now we're going backwards and yeah, and now they're artisanal.

Shawn Rogers: Yes, they are. I did make the choice. I don't buy the new ones except when they're brand new like the Taylor Swift album was just produced, but all of my classic records are original label antiques. All my Beatles albums are over 50 years old, which is-

Justin Mannhardt: Cool.

Shawn Rogers: ... kind of cool when you look at them.

Rob Collie: What about degradation of the surface and don't bits get lost over time?

Shawn Rogers: They do, but I'll tell you honestly, it's the reason I listen to vinyl. As all of us do, I have an incredible collection of a couple thousand songs on my digital devices, on my iPhone, and I was a huge adopter of iPods when they came out. I thought were great and I was listening to Breakfast in America by Supertramp one day.

Rob Collie: Love it.

Shawn Rogers: There was supposed to be this small skip in one of the songs and it's not there because it was there when I was 18, 17 years old and it bugged me and I was like, "I know it's an imperfection, but that's how I listened to that song," and that was the stimulus to buy a record album and the first one I bought was Breakfast in America by Supertramp.

Rob Collie: Strangely, I have a similar experience. My dad's first CD player, like in 1988, the Def Leppard CD.

Shawn Rogers: Pyromania, probably.

Rob Collie: The song Rocket. There was a point in the song Rocket in the chorus that happens multiple times in the song where digital analog converter and my dad's CD player had a problem with it and you get this high-pitched whine that would appear. Every time I hear the song Rocket now, I'm expecting that high-pitched whine to kick in after the yeah in the chorus. It was just some sort of aberration in that one CD player, that first gen CD player. I'll be searching for that, trying to replicate that forever. That's never going to happen.

Shawn Rogers: Yeah, it seems odd to say, but those little clicks and little bits, that's kind of music to my ears. Not to make a bad joke. It's part of listening to music for me and I just find the digital stuff to be a little too sanitized.

Rob Collie: That makes sense to me. I've increasingly found myself drifting towards the live versions of recordings over a studio, which is exactly the opposite of what young me did. I'd hear a live version, I'd be like, "Oh, this sucks terrible. It's muddy, it's whatever." Luke and I went to a concert together in October. I'm actually wearing the T-shirt of it today and, "The studio versions of those songs, Luke, aren't any good, but the live versions are amazing."

Just for human interest, it seems like we don't go a day without some "leak" or whistleblower reporting that Open AI is playing with the future of humanity. They're on the verge of a Skynet level breakthrough and no one cares. Everything that I'm seeing commercially in the marketplace doesn't give me any such fear at all. I do believe that such a thing is possible. I do believe that artificial general intelligence is possible. I do believe that it is very scary. I do not see how these Gen AI models are approaching that. There's just a little bit too much of this noise from the open AI whistleblowers, et cetera. It is still, like deep down in the recessed corners of my brain, it is a little bit unsettling. Do you have any thoughts or reactions on that given your unique sort of perch in the industry?

Shawn Rogers: Yeah, I've read some crazy AGI articles. I think we're a lot farther away than most people think. I'll give you an example. I was doing an image for a slide recently and I wanted to talk about that sub cohort, the advanced people, the high readiness group, and I just wanted a blueprint and I wanted the blueprint to include the words high readiness on it. And I asked one of the big foundation models, one of the really big popular ones that are out there. The prompt was pretty direct. I'd like a blueprint, not a historical one and a modern blueprint that shows the word high readiness. It gave it to me. I didn't like the first blueprint. I asked again for a redo. I got the second one. I liked the blueprint. It spelled high readiness wrong, which I thought was weird. I looked at the prompt because I thought I must have spelled it wrong.

I had spelled it right. So then I adjusted my prompt and said, "I love that blueprint. Let's keep using it, but I need you to spell high readiness exactly how I am spelling it right now." And I put it in quotes. And four turns later it was not able to spell the word. And so I look at stuff like that and go, "Yeah, that's a couple of miles from artificial general intelligence." I tried it on a different foundation model for a different slide yesterday, completely different set of words and it spelled those words wrong. We have a long, long way to go. I've seen some really smart people say the next five to six years for AGI. I think we're closer maybe to a decade to 12 or 15 years, and I think that like you as a technologist, there are some scary things there. It will require curation, regulation, compliance.

Will it get out in the wild and spread and take things over? I think there will be mistakes. It's hard to declare what it's going to be like in a decade, but I do believe that our children's children will live in a different world with AI and I hope it's a healthy good one. I don't spend a lot of time worrying about taking over the world. I do spend more time and I'd love to hear you guys comment on this, I do think that there will be a subset of the population that will find it difficult to find work that will cause a new level of unemployment issues as robots driven by AI and automation driven by AI. Common sense indicates that someday there will be less jobs and we're not all going to be able to sit around and be artists and we're not going to get to do fun stuff. So I do wonder how it'll hit us economically.

Justin Mannhardt: Sort of along the same thought, Shawn, you had about how nothing has really stabilized in market yet, I think makes thinking about what you just suggested, even more difficult because we haven't seen how the technology is stabilizing in the form of products and services. I know there will be disruption, but I have no idea to what extent and specifically where, I mean, some people say, "Oh, this will create a bunch of jobs." I don't know. It's hard to compare even to history and people say, "Oh, you look at the industrial revolution and the digital revolution, all these things and it all worked out." And the good human in me wants to believe it'll out, but I don't have a hot take because I just don't know where this stuff's going to land and really stabilize and really be adopted.

Shawn Rogers: It's interesting, you look at the internet when it started and it wiped out some jobs. Remember travel agents?

Justin Mannhardt: Yeah.

Shawn Rogers: You couldn't go on a trip unless you went and saw your travel agents and then they disappeared. They're back. I see travel agents all over the place and what they figured out was is they adjusted their value proposition. The value proposition changed. And I actually am taking a personal trip later in the year that's European based and I've decided I'm going to talk to a travel agent. So technology comes that displaces workers and then the workers change the value proposition and many show up in the same spot. Now, I'm sure we could have the same conversation about some of the jobs that did disappear permanently with the advent of the internet. Real estate agents were supposed to disappear, we were going to buy houses online.

Rob Collie: We're still working on that.

Shawn Rogers: I like what you said that the crystal ball on that aspect I think is really hard to spin. I just don't know. The unemployment thing that I think about from time to time that scares me a little, that might be 50, 75 years from now. I don't know.

Rob Collie: Two part question. Number one, are these large language models, the Gen AI, is that even the path to artificial general intelligence? To me, it doesn't seem like it is. It seems like something that will forever be incrementally improved, refined, et cetera, and isn't going to suddenly one day just cross some critical threshold and become self-aware. To me, there's going to be something else in parallel that would become that, and again, neither you nor I are deep AI researchers. We're not the ones in there designing new hardware and things like that. So you have a position in the industry. You're seeing things that I don't get to see. Does my hypothesis, does my hunch line up with your thinking at all that chatGPT isn't going to be the thing that turns into Skynet? It would be something else that turns into Skynet and we haven't seen that thing yet?

Shawn Rogers: I think it's going to be a combination of a bunch of things. Our large language model's going to be in the environment that's influencing it. Sure, yeah, probably. Is another type of technology going to be in there that we don't know about or we're not reading or exploring? I also think yes, that's likely. I think that's what pushes us over the tipping point. I also think that the hardware and compute aspect of it will change. I do see an interesting collision in the future between quantum computing and AI. I am not a quantum computing expert at all. I know that it will eclipse the coolest supercomputers that we see today. At some point, an AI will love that. So it's a collision of a bunch of things. Well, to the core of your question, will LLMs be part of it? Sure. Large language models probably be in the environment because they will contextually assist with the communications from the AI, but there's still parts that are missing. Are those parts in a military lab somewhere?

Rob Collie: Maybe. We don't know. Right? Somewhere someone might've already broken the RSA algorithm. All modern encryption is broken, but whoever's going to break that, it's not like the British told the Germans we broke the enigma.

Shawn Rogers: Yeah. I've read an article or two that says that has occurred, so there you go.

Rob Collie: Another thing you said in there was we couldn't get the spelling right in these images. With images that we're trying to produce, the hallucination problem isn't a risk because we know it's wrong, but you take that misspelled word and you turn that into a numerical recommendation from a piece of software, now we can't tell. I think it's going to be a much more durable and dangerous problem. I just don't see it getting ironed out in any way that's like we're going to be trusting it in some small number of years. That's just been my hunch with it.

Justin Mannhardt: Yeah, I agree with you Rob and I saw something the other day about spelling specifically that was really interesting for me to have this kind of brain click moment. You really break down what these models are trained to do. It's like, "Well, they're trained to correctly predict the next word in a sentence or a phrase." They don't predict letters. That's not trained to deconstruct a word into its subcomponents of letters, and that's internally someone will shared, it's like, "Hey, how many Rs are in the word strawberry?" There's three, and it kept thinking there's two, right? It's because it doesn't know how to do that. So okay, now there's going to have to be some sort of reasoning model that understands letters. There's going to be to a reasoning model that understands visuals and it really just the capability of the human brain to process information just far exceeds any computational system we have today. Great point about quantum computing, Shawn. There's a chasm there that needs to get crossed for anything like that to come into the picture.

Rob Collie: So last question. Most of our clients would fall squarely in the mid-market category as broadly defined as that is anyway, and specifically, I think a lot of people that you work with and interact with, it's not exclusively enterprises, but you work with a lot of companies that have tremendous resources for the mid-market companies. When it comes to AI and data going forward, what are your thoughts there? If you were a mid-market business leader today, what would you be thinking about?

Shawn Rogers: I think the mid-market is going to be pretty exciting in the next few years. They have this interesting advantage, and I'm talking generically it's not true with all of them, but newer mid-size companies, companies that don't have the legacy of some enterprise companies and also they lack acquisition legacies often if they're a mid-size company, they haven't spent 15 billion on acquisitions in the last 12 years, they have much more agile modern data and technology infrastructure. And I do believe that those types of companies, those ones that have that are going to benefit and be able to take advantage of changes in the market much quicker and in a more innovative way. I'm on purpose paying attention to them through that filter is finding these mid-market players that are in cool industries that are doing things differently because they can. A lot of enterprise companies want to do things, but they are dragging around a long tail of tech that just makes it impossible.

I still bump into enterprise companies almost every week that have very old coding languages within their environments, have servers that are hidden in someone's garage that are attached to an extension cord that haven't shut down in 43 years and no one wants to reboot it because if they do, holy cow. I would want to be in a company like that to take advantage of a disruption wave, especially with AI. So yeah, I think they're in a cool spot. I think they're more agile. Even their teams, you guys know this really well because you interact with them.

Their teams are younger and see the industry differently. Like I said, I've been around the block a few times. I'm prejudiced by my experience. A handful of years ago I was at an event standing in the hallway next to a clearly younger person, maybe a DBA, somebody that was deep in data and he was saying to his buddy that he wanted to go to this next session this afternoon. It's about this. And he said, "I think it's kind of a new database. It's like rows and then they have columns." And this individual was so deep in Hadoop at the time that he didn't fully understand that there was a precursor.

Rob Collie: It was a rectangle.

Justin Mannhardt: Oh gosh. That's amazing.

Shawn Rogers: They had mid-sized firms. They got people in there that aren't tied to the old. They're just looking at what they can do that's different and fun and exciting, and all the coding languages in their environments are different. They've already moved on from Python to whatever, and it's fun. That's part of the reason it's so fun for me to talk to end users. But yeah, if I were going to comment on that sector, keep your eyes on these people.

Rob Collie: I'd like to even amplify one thing that you said there and sort of riff on it just a touch really quickly before we go. You know about the square cubed law where T-Rex is a possible animal. Godzilla is actually impossible. There isn't enough bone structure in the world that could support an animal of Godzilla's size. It would have to be all bone. There'd be no room for muscle, no room for organs, and so it's just not a possible animal. In my experience with mid-market, even if it has as much legacy, its infrastructure is the same age as an enterprise's infrastructure, even if it's team of the same age and came from the same generation. The geometric interconnected complexity of the large organization has exploded to the point where you can't even change out the parts.

Whereas a mid-market, even if you looked at it's exactly the same on a per unit basis, was the same age in every way, it's still going to have the ability to move with an agility and to think with an agility that Godzilla just can't. And it is weird. That's one of the reasons why working with enterprises is harder is because we're able to move at a certain pace, but that geometric interconnected complexity, that's the limiting factor. That's the speed of sound in the organization and it's not fast. So it's really cool to hear you echo some of the things we've been discovering in data in general, which is that the mid-market is just such a fun and exciting place to work.

Shawn Rogers: Well, I agree with you.

Justin Mannhardt: Shawn, thanks so much for being here, man.

Shawn Rogers: It was fun.

Announcer: Thanks for listening to The Raw Data by P3 Adaptive podcast. Let the experts at P3 Adaptive help your business. Just go to P3Adaptive.com. Have a data day.