Trascrizione completa di Ep. 35 - The Easy Button Is a Lie: What AI Actually Does for Construction with Hugh Seaton

all? 00:04:18.031 --> 00:04:22.691 Not to mention you spent, you know, we'll call it 100K on Procore for year one, 00:04:22.711 --> 00:04:24.711 and the team didn't adopt any of it. 00:04:25.668 --> 00:04:31.768 I think the difference between AI and other things is, like it or not, people are using it. 00:04:31.948 --> 00:04:36.408 So a good analogy is probably mobile, where there was a period when, 00:04:36.548 --> 00:04:40.748 whether it was construction or other industries, your people were answering 00:04:40.748 --> 00:04:42.168 their emails on their phones. 00:04:42.168 --> 00:04:45.128 And people lost their minds for a little while because that was super insecure. 00:04:45.128 --> 00:04:47.428 And then they figured out some mobile device management. 00:04:48.108 --> 00:04:52.788 So right now, anyone under the age of 22, you have to assume they just got through 00:04:52.788 --> 00:04:56.488 high school or college with ChatGPT writing their essays. 00:04:56.888 --> 00:05:00.648 So whether you want to adopt it or not, your people are using it. 00:05:00.808 --> 00:05:03.948 They're figuring out what's going 00:05:03.948 --> 00:05:07.168 on with the contract or dumping specs in there to see what's going on. 00:05:07.248 --> 00:05:11.208 So I think the difference – I'll answer your question properly in a sec, 00:05:11.288 --> 00:05:15.048 but I do want to set the stage that not every technology is the same. 00:05:15.048 --> 00:05:20.848 And I think talking about adoption of technology in a broad sense can be a little 00:05:20.848 --> 00:05:26.128 bit misleading because things that are clearly written or created for the office 00:05:26.128 --> 00:05:28.888 really struggle when people aren't in the office. 00:05:29.048 --> 00:05:33.828 And we saw that 10 years ago when, you know, essentially accounting software was put on mobile. 00:05:34.208 --> 00:05:37.908 Other things, though, like find the other industry that other than maybe oil 00:05:37.908 --> 00:05:44.708 and gas that uses drones like construction does or uses reality capture or uses, you know, 3D modeling. 00:05:45.048 --> 00:05:48.188 They exist, but it's pretty rare. And there aren't a lot of accountants that do that. 00:05:48.428 --> 00:05:53.308 So for sure, there are some structural and some kind of individual reasons why 00:05:53.308 --> 00:05:56.808 a lot of software and a lot of technology struggles in construction. 00:05:56.988 --> 00:05:59.528 One of them is that you don't have a CIO just buying for the company. 00:05:59.728 --> 00:06:00.748 Like, that's hugely different. 00:06:00.988 --> 00:06:04.948 But I think AI is a little bit different because people are using the chatbots 00:06:04.948 --> 00:06:08.928 at home. They're using, like, no one uses Procore to figure out what to do this weekend. 00:06:09.268 --> 00:06:12.388 But they are using chatbots. Or how to write it or help write a letter to an 00:06:12.388 --> 00:06:16.088 owner or help try to polish it. you know, a letter to a subcontractor. 00:06:16.148 --> 00:06:21.788 I guess, though, to sort of reorient my question is, do you get the sense that 00:06:21.788 --> 00:06:28.908 companies are holding back on implementing or adopting larger platforms. 00:06:29.864 --> 00:06:34.464 Because of how the pace of every, the change. Meanwhile, yeah, 00:06:34.544 --> 00:06:36.164 here's a ChatGPT account. Go interact. 00:06:36.424 --> 00:06:42.064 Go get super curious because we're struggling doing sort of what we've been 00:06:42.064 --> 00:06:44.184 trying to do for 10, 15 years or whatever. You know what I mean? 00:06:44.424 --> 00:06:48.304 Does that make sense? 100% does. And I think we're definitely seeing that. 00:06:48.484 --> 00:06:52.664 So as I run out and talk to contractors about AI and how they might adopt it 00:06:52.664 --> 00:06:56.524 or what they might do, there's definitely a concern among some of them that 00:06:56.524 --> 00:07:00.704 I don't want to invest in something like a bespoke solution right now because 00:07:00.704 --> 00:07:03.224 I'm afraid the models are going to get so good they can just do it. 00:07:03.684 --> 00:07:08.464 I think that's probably overdone, but I understand their concern because they've 00:07:08.464 --> 00:07:12.484 all dropped money on things that no one used to your point, or that's probably the big one. 00:07:12.744 --> 00:07:18.164 I think that the models are already good enough to do most of what we want them to do. 00:07:18.324 --> 00:07:22.124 The big innovations I think in 2026 are going to be what they're wrapped up in. 00:07:24.291 --> 00:07:27.911 The first part of the year, you've seen these agent harnesses like Claude Cowork 00:07:27.911 --> 00:07:28.811 and there's some other things. 00:07:28.931 --> 00:07:34.111 OpenClaw was another one that really wrapped the models into a useful framework 00:07:34.111 --> 00:07:35.771 to get stuff done. It's super early. 00:07:35.971 --> 00:07:38.771 And honestly, most people haven't touched it yet. 00:07:38.991 --> 00:07:42.971 But I think that that's going to wind up, that kind of thinking is going to 00:07:42.971 --> 00:07:44.011 wind up being real valuable. 00:07:44.191 --> 00:07:47.711 And your original question is, what do they do now? And I think the starting 00:07:47.711 --> 00:07:52.331 point is to say, let's get our arms around the chatbots we know they're using. 00:07:52.511 --> 00:07:56.991 And everyone's had a policy. Usually the policy is very focused on security, 00:07:57.171 --> 00:07:59.391 which frankly is the least of their worries. 00:07:59.811 --> 00:08:02.851 Thinking they've got an easy button is probably the biggest worry. 00:08:02.971 --> 00:08:07.431 And the second one is, I don't know what it is, but chatbots cannot answer concisely. 00:08:07.671 --> 00:08:10.231 Even when you tell them to, you still get two pages. 00:08:10.611 --> 00:08:14.891 So getting good at prompting so that you're just getting what you want instead 00:08:14.891 -->

00:08:17.211 of, because you think about what happened when mobile came out. 00:08:17.831 --> 00:08:21.591 The difference between at work and not at work dissolved, right? 00:08:21.591 --> 00:08:23.851 And it creates all sorts of mental health issues. 00:08:24.011 --> 00:08:26.631 People don't show up fresh the next day. They're a little annoyed that they're 00:08:26.631 --> 00:08:29.871 trying to have dinner with their family and they're getting texts or calls that 00:08:29.871 --> 00:08:30.711 they know they have to answer. 00:08:31.211 --> 00:08:35.211 That won't happen with AI, but something related will. And I think that giving 00:08:35.211 --> 00:08:39.651 people more stuff that they need to consume and think about is a danger. 00:08:39.871 --> 00:08:43.711 And it's an addressable danger if you know how to prompt. And that's my point 00:08:43.711 --> 00:08:48.371 is, I think a good early step is recognizing those two problems of overwhelm. 00:08:48.491 --> 00:08:52.291 And the bigger one really is the easy button. because AI is not perfect. 00:08:52.431 --> 00:08:56.291 It doesn't know what's going on in your job. It doesn't have any judgment like a human would. 00:08:56.591 --> 00:08:59.951 So it's going to give you a first draft of an answer. And if people treat that 00:08:59.951 --> 00:09:02.011 like a final draft, they're going to find, they're going to get in trouble. 00:09:02.758 --> 00:09:07.858 Well, and I think like what we're all kind of pointing at is what we've struggled 00:09:07.858 --> 00:09:11.458 with as an industry for a long time, which is change management and kind of 00:09:11.458 --> 00:09:14.838 the paradigm of change management itself is changing, right? 00:09:14.938 --> 00:09:18.058 Because I think if you think about how to implement any kind of change, 00:09:18.338 --> 00:09:22.598 you're assuming some constants. And one of them is that the thing you're attempting 00:09:22.598 --> 00:09:25.458 to implement is not dynamic. It's somewhat static. 00:09:25.798 --> 00:09:29.578 And that whole concept is like breaking. The flip side of that, 00:09:29.658 --> 00:09:34.158 though, is the technology itself is packaged in a way, in theory, 00:09:34.478 --> 00:09:37.418 that would reduce friction to implementation. 00:09:38.198 --> 00:09:41.298 Lars, did you have a thing before I answer? Oh, I just, my mind is, 00:09:41.418 --> 00:09:44.938 my mind hurts right now. So, but Stephen, what's crazy now. 00:09:45.238 --> 00:09:50.518 You can see every, every, every emotion, every thought, just little mini rockets 00:09:50.518 --> 00:09:52.138 and trains all flying around. 00:09:52.138 --> 00:09:55.458 Now, Stephen, though, what's crazy, and we've talked to other guests, 00:09:55.718 --> 00:10:02.258 at the end of the day, there's fundamentals in construction that aren't going anywhere, right? 00:10:02.798 --> 00:10:11.158 And my concern, and I've shared this, is are we getting away from those fundamentals? 00:10:11.498 --> 00:10:14.478 Are we injecting more complexity in? 00:10:15.064 --> 00:10:20.624 Or are we on this verge of, no, we're actually really close to simplifying and 00:10:20.624 --> 00:10:27.144 focusing after 20 years of sort of like disconnected distractions and failed 00:10:27.144 --> 00:10:29.544 implementations around technology? 00:10:29.744 --> 00:10:34.344 I feel like there's both of those. I'm glad you added that. 00:10:34.844 --> 00:10:40.864 The change management, a lot of times people think software adoption is change management. 00:10:41.044 --> 00:10:44.824 And that really is, of course, it is a little bit. But I think there's an opportunity 00:10:44.824 --> 00:10:48.724 here to look at how the processes are done. 00:10:48.864 --> 00:10:52.584 Now, it's a regulated industry that has contracts that everybody, of course, loves. 00:10:52.924 --> 00:10:56.764 So there are limits to how you can move around certain parts of the process. 00:10:57.184 --> 00:11:01.264 But if you can start looking at what people do and how they do it and say, 00:11:01.424 --> 00:11:04.944 all right, there's some things we can't move, but there's some things we can. 00:11:04.944 --> 00:11:11.144 And there's some things we can automate or augment that make it enormously faster 00:11:11.144 --> 00:11:15.504 to do some of the shuffling of papers and finding stuff and all that and summarizing it, 00:11:15.624 --> 00:11:20.044 connecting dots and all that so that you have a little more time to think about 00:11:20.044 --> 00:11:22.224 things or pick up the phone and call someone. 00:11:22.444 --> 00:11:26.424 Your point, Laura, is about things that are just going to be true in construction. 00:11:26.884 --> 00:11:31.344 You're not going to get away from people out in the field figuring stuff out. 00:11:31.484 --> 00:11:35.844 Right. And how many of the solutions and how much of the project is decided 00:11:35.844 --> 00:11:39.044 and figured out when you're not near a computer? 00:11:39.384 --> 00:11:43.184 Of course, you have your phone in your pocket, but I think that's not going away. 00:11:43.364 --> 00:11:48.144 The question is, how do you make that time when somebody's out there. 00:11:48.664 --> 00:11:49.924 How do you make that more effective? 00:11:50.144 --> 00:11:54.044 So, they're not thinking about, when I go back, I have these RFIs to get into. 00:11:54.044 --> 00:11:56.264 Got to write all this stuff and coordinate with this person. 00:11:56.364 --> 00:11:57.724 Got 15 phone calls. Yeah. 00:11:58.184 --> 00:12:01.664 The phone calls probably don't go away. But maybe one of the things that I love 00:12:01.664 --> 00:12:05.144 is that right before a call with some, like a difficult call, 00:12:05.304 --> 00:12:07.724 maybe it's with a sub or maybe it's with an owner or who knows, 00:12:08.424 --> 00:12:11.724 I want a quick briefing because I just got off another call. 00:12:11.944 --> 00:12:15.304 So AI can pull together a quick briefing for me to say, all

right, 00:12:15.364 --> 00:12:17.644 you're about to talk to Lars and we know how difficult he can be. 00:12:18.486 --> 00:12:22.066 And it gives me the ability to just be smarter in that moment. 00:12:22.306 --> 00:12:24.246 That's a tiny, tiny use case. 00:12:24.486 --> 00:12:28.106 But I think that there's an opportunity. And I think the difference between 00:12:28.106 --> 00:12:32.866 adopting AI versus really integrating it. And I don't know that you do this right now. 00:12:32.966 --> 00:12:36.826 I think the first step is what I said before, which is spending the time to 00:12:36.826 --> 00:12:38.146 just get good at what you got in front of you. 00:12:38.446 --> 00:12:41.506 And I think the fact that we know there's innovations ahead of us, 00:12:41.546 --> 00:12:44.106 we don't quite know what they are. There's some things we know people are working 00:12:44.106 --> 00:12:45.686 on, but it may work, it may not. 00:12:46.146 --> 00:12:50.426 Don't sweat where it's going to be. know that the models are as good as they need to be. 00:12:50.586 --> 00:12:54.026 It's how you're surrounding them or how the providers are surrounding them with 00:12:54.026 --> 00:12:55.646 tools and with guardrails and all that. 00:12:55.766 --> 00:12:59.866 And that's getting, that's already, this year, we've already seen a big step forward. 00:13:00.046 --> 00:13:04.186 So the Claude Coworks and the Open Claws is a meaningful step forward. 00:13:04.206 --> 00:13:06.726 And it's not an incremental one. It's kind of a big one. 00:13:07.006 --> 00:13:10.926 So I think you've got the ability now to say, let's just get good at that stuff 00:13:10.926 --> 00:13:13.986 and let the other things figure it out, figure themselves out as they come. 00:13:14.486 --> 00:13:17.306 Otherwise, you're paralyzed. Otherwise, you're like, you know, 00:13:17.486 --> 00:13:21.146 okay, I'm just going to wait till some future moment when I think it's good enough. 00:13:21.326 --> 00:13:25.286 And then your competitor is outbidding you because they can get a bid together faster. 00:13:25.486 --> 00:13:28.586 They understand the gotchas in the bid if they're there. 00:13:28.926 --> 00:13:31.886 And they're winning and you're losing because you're slower than they are. 00:13:32.726 --> 00:13:38.026 So let's pull a thread on that terminology we started with, which is if we're 00:13:38.026 --> 00:13:44.346 kind of at the point of AI where, you know, we're looking at kind of internet in the 90s. 00:13:44.346 --> 00:13:50.846 The businesses that really were able to leverage that moment, maybe they focused on. 00:13:51.911 --> 00:13:57.131 Cable infrastructure. They focused on like some decisions around like data storage. 00:13:57.171 --> 00:14:01.511 Like what are the, what's the analog for, for the, for the companies that, 00:14:01.611 --> 00:14:05.551 that position themselves well as, as internet was maturing in the nineties and 00:14:05.551 --> 00:14:09.311 early two thousands with kind

of this takeaway of, 00:14:09.731 --> 00:14:13.331 Hey, we're in a constant state of change, but you also can't sit on your hands. 00:14:13.571 --> 00:14:18.311 So like, how would we, is it, is it just data strategy? 00:14:18.491 --> 00:14:24.091 Is that one of the kind of building blocks? I think people overestimate how 00:14:24.091 --> 00:14:29.771 much of a blocker data strategy is right now. Obviously, people should be getting better at it. 00:14:31.071 --> 00:14:35.971 The thing that always comes up is you hear funny numbers like 95% of construction 00:14:35.971 --> 00:14:37.751 data isn't used. Well, what do you mean? 00:14:38.011 --> 00:14:42.391 Do you mean that all my PDFs where I've copied three versions of the same RFI 00:14:42.391 --> 00:14:46.571 or some conversation that isn't relevant anymore because the thing's been put 00:14:46.571 --> 00:14:49.851 in place? It's like, just because you've got a million PDFs doesn't mean you've 00:14:49.851 --> 00:14:52.771 got a million data points. It means that you've got a lot of stuff to go through. 00:14:53.331 --> 00:14:56.271 Yeah, not a lot of points. So talk to anyone that's doing handover and you know 00:14:56.271 --> 00:14:59.651 what I'm talking about. Oh my God, you just gave me a terabyte of irrelevant information. 00:14:59.851 --> 00:15:02.251 Thank you so much for satisfying the contract. 00:15:02.851 --> 00:15:09.351 So I think that first would be, like I said, some fundamental or kind of basic 00:15:09.351 --> 00:15:12.691 capability and using the things yourself. 00:15:12.931 --> 00:15:16.211 Because you're not going to know the limits of it really. I can tell you all 00:15:16.211 --> 00:15:18.991 day long that, oh, it might hallucinate or it might be incomplete or whatever, 00:15:19.091 --> 00:15:21.011 but that doesn't mean anything until you played with it. 00:15:21.271 --> 00:15:26.411 So I think that leaders were using it for their personal life just to say, 00:15:26.531 --> 00:15:29.011 look, this is coming and I'm going to figure out what we're doing this weekend. 00:15:29.011 --> 00:15:30.811 I want to go to a new barbecue place, whatever. 00:15:31.211 --> 00:15:35.471 You can start to get an intuitive sense so that you can apply your judgment 00:15:35.471 --> 00:15:39.051 as a leader or as a PM or as a PE, whatever the level you are. 00:15:39.331 --> 00:15:42.851 So you can apply your own judgment and say, I think this would help me here 00:15:42.851 --> 00:15:46.091 and I don't trusted over there. And no one can tell you that. 00:15:46.331 --> 00:15:51.611 I mean, they can give you guidelines and a starting point, but ultimately... 00:15:52.767 --> 00:15:55.747 People that are in the work are going to know where the blockers are, 00:15:55.867 --> 00:15:59.607 that AI as it currently stands can really help. I'll give you an example. 00:15:59.807 --> 00:16:04.847 So I do some work with Smackna, and one of their members had seen some webinars 00:16:04.847 --> 00:16:07.407 I'd done, and he'd been playing around with doing exactly what I said, 00:16:07.527 --> 00:16:08.307 playing with it himself. 00:16:08.467 --> 00:16:12.507 And he said, you know, my PMs keep driving me nuts. He had 40 PMs, 00:16:12.607 --> 00:16:17.127 and they do their timesheets with Excel, right, like half the industry. 00:16:17.487 --> 00:16:22.287 And that's a big Excel spreadsheet. So his PMs kept driving him nuts saying, 00:16:22.447 --> 00:16:25.947 I don't know what sell to do. And you know what I mean? And that's a tiny thing. 00:16:26.207 --> 00:16:29.307 Are you going to hire somebody to go fix that for you? No, you're just going 00:16:29.307 --> 00:16:31.507 to live with it and, you know, tell them to have a cup of coffee. 00:16:31.847 --> 00:16:35.567 This guy goes in, he asked ChatGPT, how can I fix this? It gives him three options. 00:16:35.687 --> 00:16:38.647 He chooses a surprising one, which was, I'll write some code for you. 00:16:38.847 --> 00:16:42.967 It writes the code, tells him to turn on developer mode, which I didn't even know existed. 00:16:43.719 --> 00:16:47.179 It tells them where to pop it in, and it worked. Now they have this template 00:16:47.179 --> 00:16:50.459 that when you click a cell, the row and the column light up. 00:16:50.779 --> 00:16:56.359 There's a million things like that. And that, to me, is where giving that capability 00:16:56.359 --> 00:17:00.439 to folks at the ground level and at every level, they're going to start to say, 00:17:00.619 --> 00:17:03.459 I'm blowing a lot of time just finding the right cell, and then I've got to 00:17:03.459 --> 00:17:05.039 go check the thing, and so on and so forth. 00:17:05.039 --> 00:17:09.719 And so think about that as a thing that people can do right now is to get smart 00:17:09.719 --> 00:17:13.639 about it at the individual level so that people start saying, 00:17:13.819 --> 00:17:16.139 where might AI make my life better? 00:17:16.239 --> 00:17:21.859 And I think that the skill that is valuable now is – I heard a really good example. 00:17:22.159 --> 00:17:25.079 You guys are familiar with parkour. That's the thing where people run around 00:17:25.079 --> 00:17:28.579 on desk – I knew it. You do parkour. 00:17:28.719 --> 00:17:31.419 I knew it. I've seen you in videos, Hugh. That's right. Totally. 00:17:31.679 --> 00:17:34.339 That's right. Only here in Austin, though. Parkour. Yeah, that's right. 00:17:34.779 --> 00:17:38.559 What happens is people that do that get what they call parkour vision, 00:17:38.739 --> 00:17:42.759 where they look at buildings and they're immediately assessing where they could 00:17:42.759 --> 00:17:43.919 jump up and how they could do it. 00:17:44.179 --> 00:17:48.319 Well, I think there's an equivalent of that here, which is kind of AI parkour 00:17:48.319 --> 00:17:52.519 vision, where you start looking at your work and say, I'm not going to trust 00:17:52.519 --> 00:17:55.619 it over there because of X, Y,

Z, or it won't save me time. 00:17:55.619 --> 00:18:00.419 But I know that if I connect these two steps and have AI sit in the middle and 00:18:00.419 --> 00:18:02.579 I'm there to watch what comes out or however you want to do it, 00:18:02.959 --> 00:18:06.819 that's a real skill that's not going to get less useful. It's going to compound 00:18:06.819 --> 00:18:08.579 over time because you'll get better at it. 00:18:08.959 --> 00:18:11.659 You'll start getting more creative. And at the end of the day, 00:18:11.799 --> 00:18:15.099 you don't hear construction being described as a creative industry. 00:18:15.099 --> 00:18:17.839 It's probably one of the most creative industries because you're constantly 00:18:17.839 --> 00:18:21.039 figuring out and coming up with new ways of making that. 00:18:22.591 --> 00:18:27.771 A circle go into a square hole. We're naming this episode AI Parkour Vision. 00:18:28.231 --> 00:18:32.551 Okay. I mean, that's the best analogy I think I've heard of the way in which 00:18:32.551 --> 00:18:34.011 AI can augment the human brain. 00:18:34.791 --> 00:18:38.211 So I got that from a guy named Nate Jones. So if people want to nerd out, 00:18:38.331 --> 00:18:41.531 he's got a YouTube channel. I got to give him props. Nate B. Jones. 00:18:41.991 --> 00:18:45.751 He's the originator of AI Parkour Vision? That's right. It was a podcast from 00:18:45.751 --> 00:18:49.711 a little while ago. We're reaching the early 2000s for the parkour reference. 00:18:49.971 --> 00:18:52.871 That's right. Oh, good. And if you saw the guy, he probably doesn't do parkour. 00:18:54.651 --> 00:19:02.551 Bless his heart. So absolutely love what you're saying. And even in my own journey 00:19:02.551 --> 00:19:09.071 around, you know, AI usage is exactly what you just said, which is, wait a minute. 00:19:09.111 --> 00:19:12.991 I just opened up this new door, this new capability, solved the problem. 00:19:13.491 --> 00:19:16.831 And now I have 10 more pick your own adventure doors to go in. 00:19:17.691 --> 00:19:21.591 Mind is blown, right? But I want to go back to something. 00:19:22.420 --> 00:19:26.940 And not that I'm not that I'm like shame, shame on that company, 00:19:27.080 --> 00:19:29.920 but it's like they're using Excel for timesheets. 00:19:30.520 --> 00:19:35.140 And it's like that to me is the gap. 00:19:37.080 --> 00:19:40.020 And there's plenty of like, I think, you know, digital apps. 00:19:40.100 --> 00:19:41.820 I've certainly used a number of them just on there. 00:19:41.920 --> 00:19:44.800 You know, people can punch it in, you know, and it's like, no, 00:19:44.860 --> 00:19:48.980 no, no, we're still we're still using Excel for timesheets. 00:19:49.060 --> 00:19:51.960 And you said 40 p.m. in a I'm guessing 00:19:51.960 --> 00:19:55.680 mechanical subcontractor probably 250 500 00:19:55.680 --> 00:20:00.600 plus organization and they're using excel timesheets but but even like I would 00:20:00.600 --> 00:20:05.420 push back

even on that takeaway a little bit because I think if we think about 00:20:05.420 --> 00:20:11.300 the like technology as as a tool that we leverage you know we've we've been 00:20:11.300 --> 00:20:14.980 kind of programmed a little bit to think about technology through the lens of like SaaS. 00:20:15.420 --> 00:20:18.480 Apps, right? You're saying apps? No, apps, solutions. 00:20:18.880 --> 00:20:24.380 Okay. But as Claude and ChatGPT have integrated themselves into Excel, 00:20:25.060 --> 00:20:30.520 it kind of transforms the use case and, you know, the potential value. 00:20:31.540 --> 00:20:40.280 So, go ahead. No, no, it's just this idea around, you know, when I first started 00:20:40.280 --> 00:20:44.200 my first construction business, right, it's lack of structure, right? 00:20:44.280 --> 00:20:49.100 You're sort of starting from nothing. And as an engineer, I desired structure, 00:20:49.280 --> 00:20:50.960 system, processes, workflows, right? 00:20:51.180 --> 00:20:54.640 And Outlook and Excel aren't workflows, aren't necessarily processes, 00:20:54.920 --> 00:20:59.080 right? It's a PDF or it's an attachment, an email with a data cell and a block 00:20:59.080 --> 00:21:01.780 and a number that I can go put in an accounting system. 00:21:02.440 --> 00:21:06.800 But and again, as I'm every morning, I wake up and, you know, 00:21:06.860 --> 00:21:08.580 I get these little AI feeds, AI updates. 00:21:08.700 --> 00:21:11.700 And it's just like my mind is blown with another one, another one, another one. 00:21:11.820 --> 00:21:15.980 And I desire simplification as an entrepreneur in my organization. 00:21:15.980 --> 00:21:22.000 I desire enough structure and enough decentralization and autonomy that the 00:21:22.000 --> 00:21:26.800 operators can go do efficiently what they're extremely good at, right? 00:21:27.040 --> 00:21:30.780 And not overwhelm them with noise. And Stephen, in this case, 00:21:30.980 --> 00:21:35.500 you know, in my mind, what I heard Hugh say was, hey, we're still doing timesheets 00:21:35.500 --> 00:21:37.680 over here on Excel. And I'm like, that's insane. 00:21:37.860 --> 00:21:40.600 But what I heard you just say was like, no, no, no. 00:21:40.600 --> 00:21:46.420 So is that the most simplest and basic way, fastest way to now get the output 00:21:46.420 --> 00:21:50.740 you need as an organization rather than, hey, for $5.99 a month, 00:21:50.880 --> 00:21:55.140 you can have, you know, for every person in your organization can digitally submit a timesheet. 00:21:55.260 --> 00:21:57.760 And, you know, maybe 10 years ago, that was like really slick. But guess what? 00:21:58.240 --> 00:22:01.920 You don't need that anymore. You can orient some of these sheets with integrated 00:22:01.920 --> 00:22:06.060 Python code to those automations that create similar outcomes, right? 00:22:06.320 --> 00:22:12.660 And Claude is in Excel and PowerPoint now. So I think there's another thread 00:22:12.660 --> 00:22:16.140 that this

leads us to, and that is that AI is going to be embedded in everything. 00:22:16.460 --> 00:22:22.780 So if you've got a 45-year-old PM who's been doing things a certain way or whatever 00:22:22.780 --> 00:22:26.860 the role is, and they're really used to the flexibility of Excel –. 00:22:27.750 --> 00:22:31.770 One of the things that AI, even now, and certainly more in the future, 00:22:31.950 --> 00:22:36.250 is good at is taking things the way you've given it, like the way you give it to it. 00:22:36.350 --> 00:22:39.830 You don't need to input it the way it needs it. 00:22:40.050 --> 00:22:43.190 And that's one of the reasons Excel still gets used is because you can just 00:22:43.190 --> 00:22:44.410 drop in there what you want to do. 00:22:44.770 --> 00:22:48.310 Yeah, it's like a piece of paper that you can send people. And I think that 00:22:48.310 --> 00:22:54.890 you'll find that some of the strictness that SaaS requires isn't required by AI. 00:22:54.890 --> 00:22:57.790 Now, you have control issues because now you've got to make sure that it, 00:22:57.910 --> 00:23:00.790 I mean, the value of having things exactly the right way is you can predict 00:23:00.790 --> 00:23:01.630 what's going to come out. 00:23:01.990 --> 00:23:05.510 But you can handle that in other ways with AI where you're telling it what you 00:23:05.510 --> 00:23:07.010 want, maybe giving it an example. 00:23:07.250 --> 00:23:11.250 So there's other ways to get that control. And once you've done it, it'll repeat it. 00:23:11.610 --> 00:23:18.750 So I do think that where you can give people the ability to work the way they're 00:23:18.750 --> 00:23:22.710 comfortable working and then add the value and intelligence on the back end, 00:23:22.710 --> 00:23:25.230 That's one way I think that we can really win. 00:23:25.510 --> 00:23:30.290 And, you know, we've got a lot of folks that are the repository of 20, 30 years of experience. 00:23:30.530 --> 00:23:34.850 How do we make them more valuable? How do we give them more time to mentor, 00:23:35.250 --> 00:23:37.270 to solve harder problems, to talk more? 00:23:37.470 --> 00:23:41.810 Because, you know, you don't know what someone knows until the problem comes 00:23:41.810 --> 00:23:42.530 up that they're solving. 00:23:42.790 --> 00:23:45.390 Like, it's really hard for people to tell you everything. I mean, they can't. 00:23:45.730 --> 00:23:48.690 So the more they're out there helping people and they've got time or, 00:23:48.790 --> 00:23:50.630 you know, they're on the tough job, whatever it might be. 00:23:51.270 --> 00:23:56.810 I think that one of the roles AI absolutely can play is elevating human capability, not replacing it. 00:23:57.430 --> 00:24:00.990 You've heard me say this before. I'm skeptical of AI that doesn't make the person 00:24:00.990 --> 00:24:03.490 better, not the company, not the process, the person. 00:24:03.750 --> 00:24:07.550 That's why Trunk Tools keeps coming up in my conversations in building AI that's 00:24:07.550 --> 00:24:11.350 human-centric, designed to strip away the busy work and let experienced builders 00:24:11.350 --> 00:24:12.850 like you do what they're actually good at. 00:24:12.990 --> 00:24:15.890 And here's the thing that gets me. The average worker takes about five and a 00:24:15.890 --> 00:24:20.210 half hours a week just looking for project data. That's a full day every week 00:24:20.210 --> 00:24:21.950 lost to digging through files and folders. 00:24:22.330 --> 00:24:26.650 So Trunk Tools is tackling that head on. Check the link in the description to learn more. 100%. 00:24:27.516 --> 00:24:31.056 Well, and let's talk a little bit about SaaS and the state of SaaS, 00:24:31.216 --> 00:24:32.236 because I think, you know, 00:24:32.476 --> 00:24:38.236 part of an underlying assumption in kind of the software space is that each 00:24:38.236 --> 00:24:43.356 of these companies have a moat around their tech stack, around the time to build. 00:24:43.696 --> 00:24:47.836 And all of that's being challenged by a variety of different kind of, 00:24:48.336 --> 00:24:51.996 well, whether it be, you know, the kind of citizen development happening or 00:24:51.996 --> 00:24:55.696 just for engineers themselves to kind of spin up new apps. 00:24:55.696 --> 00:25:02.956 So when you look at some of those challenges, I guess, what do you see as the 00:25:02.956 --> 00:25:05.116 future, Cliff Sass? How does that evolve? 00:25:05.796 --> 00:25:12.836 Yeah, I think no one knows. So, right, part of anchoring us in the early part 00:25:12.836 --> 00:25:17.476 of a decade-long process or potentially longer is you just don't know. 00:25:17.676 --> 00:25:21.596 What you probably do know, though, and I quote Satya Nadella, 00:25:21.716 --> 00:25:27.336 right, the CEO of Microsoft, is that one way of looking at apps is business 00:25:27.336 --> 00:25:28.956 logic on top of a database. 00:25:29.196 --> 00:25:33.356 And agents can sit on top of the database and replace the business logic that 00:25:33.356 --> 00:25:35.536 was in SaaS. Now, fair enough. 00:25:35.776 --> 00:25:39.156 The moats that they have, especially larger companies, but really anyone, 00:25:39.336 --> 00:25:41.636 isn't just the tech. It isn't just the software. 00:25:41.856 --> 00:25:45.236 It's the relationships they have. It's the trust they have. 00:25:45.476 --> 00:25:49.976 It's the kind of built out familiarity with how to do things. 00:25:50.116 --> 00:25:53.456 So the moat's probably a little bit stronger than people might think. 00:25:54.172 --> 00:25:58.652 What really has started to hurt is if you sell on a per-seat basis, 00:25:59.272 --> 00:26:02.952 are you going to be able to grow the number of seats per account? 00:26:03.392 --> 00:26:06.092 And probably not. And in fact, people are going to say, well, 00:26:06.192 --> 00:26:11.212 if I can string up an agent to do this for me, maybe I don't need 10 licenses. Maybe I just need two. 00:26:11.492 -->

00:26:15.492 It's definitely why ServiceNow, Salesforce, and some of those other folks had 00:26:15.492 --> 00:26:17.172 a rough week a couple weeks ago. 00:26:17.472 --> 00:26:23.212 So I think the economics of AI, sorry, of SaaS are up for grabs a little bit. 00:26:23.212 --> 00:26:28.932 You're hearing things like, let's do a new business model where we charge per outcome. 00:26:29.592 --> 00:26:32.972 That's a really tough one. I get why people are saying that, 00:26:33.132 --> 00:26:37.312 because agents are supposed to do stuff as opposed to just be there for you to use. 00:26:37.532 --> 00:26:39.552 The problem is nobody can budget for that. 00:26:41.472 --> 00:26:47.052 You'll hear developers and startup founders talk about $10,000 AWS bills because 00:26:47.052 --> 00:26:49.792 someone wrote something wrong or they couldn't deal with it. 00:26:49.792 --> 00:26:54.452 And I've heard contractors say the same thing when they went on to AWS or Azure. 00:26:54.632 --> 00:26:58.972 They were really surprised because they thought they were going to spend X and they spent more. 00:26:59.192 --> 00:27:01.512 And more to the point, they just don't know how much they're going to spend. 00:27:01.672 --> 00:27:07.552 So this per-sip, this per-use idea, I think they're going to be industries where that makes sense. 00:27:07.712 --> 00:27:10.832 And like customer support is where it's seeing some success. 00:27:11.332 --> 00:27:14.912 I think that beyond that, I think people want to know what they're paying. 00:27:15.072 --> 00:27:18.232 And they'd actually rather pay a little more than have surprising months. 00:27:18.792 --> 00:27:22.292 So yeah, I think I just gave you a word salad to say, we don't know. 00:27:22.772 --> 00:27:26.632 It's 2026 is a tough year to make a prediction like that. 00:27:26.732 --> 00:27:30.132 And obviously people are confidently doing so because they think it's their 00:27:30.132 --> 00:27:31.792 job, but I think no one knows. 00:27:32.872 --> 00:27:36.092 I think one thing we do know is that I think it's going to take... 00:27:37.014 --> 00:27:42.994 Some real kind of advances in, in various use cases or like just business model 00:27:42.994 --> 00:27:48.834 creativity to validate some of the valuations that had historically been in the sector. 00:27:49.034 --> 00:27:54.594 Right. And so, I mean, I've said to you, and I appreciate you're not prompting 00:27:54.594 --> 00:27:57.994 me, but I'll go out and say it. I don't know how you allocate capital right now. 00:27:58.374 --> 00:28:02.974 Yeah. I don't know how you, unless you're investing in someone in Silicon Valley, 00:28:03.114 --> 00:28:04.474 who's doing some frontier research. 00:28:04.754 --> 00:28:09.734 I don't know how a VC does what they do. I mean, they're paid the big bucks to figure that out. 00:28:09.894 --> 00:28:13.714 So God bless them. But right now, you don't know what the cost basis will be. 00:28:13.894 --> 00:28:16.314 You don't know what your growth is going to be. 00:28:16.614 --> 00:28:21.094 Everyone says they're going to have a billion dollar business because VCs require that. Okay. 00:28:21.714 --> 00:28:27.654 You can't have $50 billion businesses in construction. I think you get one, maybe two. 00:28:28.514 --> 00:28:32.394 So long-winded way of saying, I think the 00:28:32.394 --> 00:28:35.414 financing of software companies is going 00:28:35.414 --> 00:28:38.154 to have a rough year for the same reason that i 00:28:38.154 --> 00:28:42.114 just said is you don't really know well i 00:28:42.114 --> 00:28:44.974 think it does it changes some of the dynamics of of the 00:28:44.974 --> 00:28:48.574 the build versus buy decisions and equations in 00:28:48.574 --> 00:28:53.174 a lot of these businesses right like you know i came from an opportunity where 00:28:53.174 --> 00:28:58.834 in in a gcc we were analyzing that kind of build versus buy and historically 00:28:58.834 --> 00:29:03.274 have been fairly anti building internally because of the lack of kind of product 00:29:03.274 --> 00:29:06.474 management understanding in some of these contractors. 00:29:07.394 --> 00:29:11.094 And, but I think as long, and you're seeing this, right, you're seeing Suffolk 00:29:11.094 --> 00:29:15.694 hire product managers and folks with AI titles and data analysts. 00:29:15.854 --> 00:29:18.394 And so as you bring some of that talent on board. 00:29:19.928 --> 00:29:23.048 Build concept looks more attainable. How do you look at that? 00:29:23.448 --> 00:29:29.068 I think you're 100% right. I think that the CIO role needs to finally be strategic 00:29:29.068 --> 00:29:31.428 in GCs as opposed to being functional. 00:29:31.768 --> 00:29:35.428 I think it needs to be literally C-suite instead of reporting to the CFO. 00:29:35.848 --> 00:29:37.868 I mean, look, every company is going to make up their own mind, 00:29:37.868 --> 00:29:42.068 but to do what you just described, which I think is absolutely true. 00:29:42.248 --> 00:29:46.928 And the reality is that other companies in other industries have had strong 00:29:46.928 --> 00:29:52.488 CIOs for decades and they build their own mini apps on top of big platforms. 00:29:52.628 --> 00:29:57.448 This has been going on forever, but they have a group that is dedicated to that, 00:29:57.528 --> 00:30:00.828 that's good at it, that has the right rules, has the right incentives and all that. 00:30:00.908 --> 00:30:04.648 If you're trying to run that kind of an activity with people that came off the 00:30:04.648 --> 00:30:10.388 field and are thinking with the same risk profile and the same sensibility of 00:30:10.388 --> 00:30:11.968 putting a building, it's just not the same. 00:30:11.968 --> 00:30:15.948 So I think that AI is going to, just like other really big changes, 00:30:16.148 --> 00:30:21.268 you're going to need to think about who am I staffing and what do I think is strategic for us? 00:30:21.568 --> 00:30:26.988 And I think that having a CIO

who can look at the business and say, 00:30:27.128 --> 00:30:29.948 here's where we should be and here's where we should invest. 00:30:30.508 --> 00:30:34.568 Obviously at the discretion of the CEO and presumably the board or something 00:30:34.568 --> 00:30:39.528 like that, but I think that's going to become a much more important thing than it has been in the past. 00:30:39.528 --> 00:30:42.608 I don't disagree with you on that. 00:30:42.968 --> 00:30:49.168 What percentage of companies in this construction blob are. 00:30:50.103 --> 00:30:53.683 Actually have a cio yeah i don't know more 00:30:53.683 --> 00:30:56.723 than i thought more than i thought but the question isn't only 00:30:56.723 --> 00:30:59.563 whether someone has a business card with cio on it the question 00:30:59.563 --> 00:31:02.483 is who do they report to so if they report to the cfo 00:31:02.483 --> 00:31:05.743 then they're not really i mean i don't want to cast any aspersions on people 00:31:05.743 --> 00:31:08.963 with that title but it means that they're structured in a way where winning 00:31:08.963 --> 00:31:13.783 is harder than it should be no they have to like yeah they're just pushing uphill 00:31:13.783 --> 00:31:17.263 every day right we're pushing against someone who's looking at cost and risk 00:31:17.263 --> 00:31:19.843 which is great but ultimately innovation and speed. 00:31:20.083 --> 00:31:25.123 Yeah. And what's the upside? A great example, if I may, one of the things that 00:31:25.123 --> 00:31:28.383 I think you're going to see people, and I've heard glimmerings of this, 00:31:28.863 --> 00:31:33.843 knowledge can be a strategic asset, knowing how to do things and capturing that 00:31:33.843 --> 00:31:34.823 and doing things with it. 00:31:34.863 --> 00:31:37.523 It's not so strategic when you're just showing it to other people, 00:31:37.523 --> 00:31:41.823 but if you're codifying how we do things, because AI can do that for you and 00:31:41.823 --> 00:31:45.983 can serve it up and can see what someone's working on and real knowledge management, 00:31:45.983 --> 00:31:49.603 as opposed to an Excel spreadsheet with less lessons learned on it. 00:31:49.803 --> 00:31:55.583 That starts to mean that you're not just covering risk, which is the first step, 00:31:55.723 --> 00:31:58.903 but you're actually creating new opportunities. Yeah. 00:32:00.503 --> 00:32:06.203 The challenge I have with nothing you described necessarily, 00:32:06.203 --> 00:32:11.803 it's I wonder what percentage of companies, 00:32:11.803 --> 00:32:14.763 the top 100 ENR, you know 00:32:14.763 --> 00:32:18.403 versus the rest of the the industry 00:32:18.403 --> 00:32:25.643 from a revenue versus versus total companies and or you call it top 1000 right 00:32:25.643 --> 00:32:30.283 and what i'm getting at is that i believe there is a massive chunk of this industry 00:32:30.283 --> 00:32:37.583 that many many many owners actually have to deal with that will never ever budget for a

cio. 00:32:38.455 --> 00:32:45.215 And that because they don't have the time, they don't have the vision, 00:32:45.315 --> 00:32:49.195 they don't have the awareness, they see it as a drag, a distraction. 00:32:49.695 --> 00:32:55.535 And so it's almost like the separation gap just continues to occur. 00:32:55.735 --> 00:33:00.095 I don't know, but that's that small business, large business, sophisticated. 00:33:00.655 --> 00:33:03.815 Now, I believe there's transition occurring, right? All of a sudden, 00:33:03.975 --> 00:33:08.795 that young, hungry entrepreneur who now is experimenting with Claude Bott or 00:33:08.795 --> 00:33:12.915 Pickett, all of a sudden has the most badass estimating spreadsheet because 00:33:12.915 --> 00:33:18.055 they were able to just put all this intelligence in there. That's fun. That's a game changer. 00:33:18.375 --> 00:33:22.255 Don't you think this is one of those moments where it's worth remembering how 00:33:22.255 --> 00:33:27.275 many different project types there are and how many different GC focuses there are? 00:33:27.475 --> 00:33:31.775 So this is where the talking about construction generally kind of breaks down 00:33:31.775 --> 00:33:36.215 a little bit. But if you're building the same type of box in the same area over 00:33:36.215 --> 00:33:37.255 and over again, you're right. 00:33:37.495 --> 00:33:41.255 You kind of know it already. As opposed to building really big, 00:33:41.335 --> 00:33:44.095 complex buildings, God forbid it's a data center. 00:33:44.435 --> 00:33:47.535 So I do think that you're going to find a real difference. 00:33:47.755 --> 00:33:50.735 And it might – look, at the end of the day, the company has to make up their 00:33:50.735 --> 00:33:53.515 mind about how they're competing and who they're competing against. 00:33:53.815 --> 00:33:57.675 Yeah, what the market sector is. Yeah. I mean, and it matters more now because 00:33:57.675 --> 00:34:01.195 there's opportunities to be that much better, but it's still going to be an expense. 00:34:01.555 --> 00:34:04.475 Yeah. And if that expense really doesn't buy you very much, then they're right 00:34:04.475 --> 00:34:05.155 to do what they're doing. 00:34:05.335 --> 00:34:09.815 Or more importantly, if they're saying within the next year or two years, 00:34:09.975 --> 00:34:14.615 I don't trust that our adoption isn't going to create more risk and expense than it's going to save. 00:34:14.755 --> 00:34:18.155 Then maybe they say, we're going to hang on and just do what we do and take 00:34:18.155 --> 00:34:19.115 more people out to dinner. 00:34:19.811 --> 00:34:24.051 But I think that's where, I mean, we are in such a period of disruption that 00:34:24.051 --> 00:34:25.851 I think we're going to have some real separation. 00:34:26.271 --> 00:34:28.751 You know, I think we've gone through a period of, I don't know. 00:34:29.471 --> 00:34:33.211 34 plus years of, I don't know why I picked 34. 00:34:33.411 --> 00:34:36.511 It's a very specific number. but you know 30 plus years 00:34:36.511 --> 00:34:39.851 where you know the the big boys 00:34:39.851 --> 00:34:43.691 have gotten bigger the regionals have everyone's gotten bigger and obviously 00:34:43.691 --> 00:34:49.031 but there's a bit of parity in terms of end product but i do think that's coming 00:34:49.031 --> 00:34:53.391 to an end because i i could see and i've made this prediction before obviously 00:34:53.391 --> 00:34:58.411 aecom you know acquiring consigli kind of took the the lid off that box but. 00:34:59.043 --> 00:35:04.123 If a contractor is fairly, you know, forward thinking and maybe doesn't have 00:35:04.123 --> 00:35:06.443 that CIO talent, they go buy it. 00:35:06.643 --> 00:35:12.643 They go buy a business on the tech side that is going to kind of fundamentally 00:35:12.643 --> 00:35:16.183 enable a differentiation in their end product, 00:35:16.183 --> 00:35:21.503 but also close the gaps in terms of internal capabilities to think through strategically 00:35:21.503 --> 00:35:22.863 how to leverage these technologies. 00:35:22.863 --> 00:35:27.023 Is they're going to be in a different place than a contractor who maybe historically 00:35:27.023 --> 00:35:31.863 has a CFO, has an uncle, has a renovation business, like that historical model. 00:35:32.123 --> 00:35:34.803 So I do think we're kind of coming upon that time. 00:35:35.103 --> 00:35:37.723 I think you're right. I just think not for every segment. 00:35:38.043 --> 00:35:40.963 I think there's going to be some segments where they are. I agree, construction's big. 00:35:41.303 --> 00:35:44.803 That said, you know, what's funny is we've just worked on a project with an 00:35:44.803 --> 00:35:47.643 aggregates company, and they're everything you just described. 00:35:47.823 --> 00:35:52.663 Like somebody founded it, and it's guys with big trucks moving lots of rocks. 00:35:52.663 --> 00:35:57.383 I mean, I don't want to oversimplify it, but the thing is, if one of those breaks 00:35:57.383 --> 00:36:00.403 down, they can tell you how much money they lose per hour. 00:36:00.763 --> 00:36:04.143 So we built a machine learning model that took their data, which they thought 00:36:04.143 --> 00:36:07.343 was really bad, and they were right, it was less than they wanted it to be. 00:36:07.483 --> 00:36:11.463 But we were able to say, I can't tell you what broke, but I can tell you it's 00:36:11.463 --> 00:36:12.383 one of these five things. 00:36:12.623 --> 00:36:16.583 Which means that instead of having to drive 50 miles out to a job site, 00:36:16.763 --> 00:36:20.143 identify what it was, drive back, get what they need, and then basically take 00:36:20.143 --> 00:36:21.543 two days to go fix something. 00:36:22.083 --> 00:36:24.623 You know, at every hour, thousands of dollars have lost money. 00:36:24.943 --> 00:36:27.203 They're able to go just drive straight out there and say, okay, 00:36:27.283 --> 00:36:28.323 I've gotten the truck what I

need. 00:36:28.895 --> 00:36:31.375 I think there's a lot of opportunity for that kind of thing. 00:36:31.475 --> 00:36:36.795 And we use AI right now to pretty narrowly mean generative AI and chatbots, 00:36:36.915 --> 00:36:38.455 really. But it's a big tent. 00:36:38.715 --> 00:36:42.915 There's machine learning in there that you can, if you don't think I need this 00:36:42.915 --> 00:36:47.115 to be like software, where I press a button and it gives me the exact answer. 00:36:47.235 --> 00:36:51.595 If you say, if you can make my decisions better, this isn't quite the parkour 00:36:51.595 --> 00:36:52.995 vision thing, but it's sort of like that. 00:36:52.995 --> 00:36:58.875 If you can give me the ability to reduce decision time or more efficiently make 00:36:58.875 --> 00:37:02.955 decisions, I'm still making the decision, but it's faster and we've saved money. 00:37:03.075 --> 00:37:05.015 So I think there's an opportunity there too. 00:37:05.235 --> 00:37:08.295 It's just the right person has to meet the right person to tell them that because 00:37:08.295 --> 00:37:12.495 very few people running an aggregates company are saying, you know what I think we need? 00:37:12.515 --> 00:37:15.475 We need some ML in here, right alongside the old cowbell. 00:37:15.475 --> 00:37:18.455 So I think that the other thing that there's an opportunity for, 00:37:18.555 --> 00:37:19.875 I'm obviously biased here, 00:37:20.075 --> 00:37:25.435 but just like contractors bring in consultants for engineering they don't know, 00:37:25.635 --> 00:37:29.155 whether it's MEP or something like that, there's an opportunity here to do what 00:37:29.155 --> 00:37:33.235 the rest of the economy does and say, I don't know AI that well. 00:37:33.335 --> 00:37:35.595 Let's bring in somebody to help us out. 00:37:36.855 --> 00:37:40.715 Yeah. Yeah, that's interesting. So let me ask this question. 00:37:40.855 --> 00:37:43.715 You are building a SaaS product, the link. 00:37:43.715 --> 00:37:47.075 It helps provide kind of clarity in the spec space and 00:37:47.075 --> 00:37:50.615 you know from you know our kind of history together 00:37:50.615 --> 00:37:56.355 i think you know specs is is a thing that hasn't gotten a lot of sas love it's 00:37:56.355 --> 00:38:00.495 maybe getting a bit more these days but kind of in the light of the previous 00:38:00.495 --> 00:38:05.795 conversation we've had around the challenges in the sas category how are you 00:38:05.795 --> 00:38:08.615 thinking about building a sas product. 00:38:09.835 --> 00:38:14.115 And you know you're kind of layering in some some kind of strategic services 00:38:14.115 --> 00:38:18.295 around ai consulting you actually you know the link got a shout out from a previous 00:38:18.295 --> 00:38:21.855 guest the steve little from fresh market was was just mentioning some of the 00:38:21.855 --> 00:38:24.915 tools that he really likes interacting with and mentioned the link is one of them. 00:38:25.879 --> 00:38:31.739 So I guess walk me through how you, the marriage between your kind of industry 00:38:31.739 --> 00:38:35.739 view of AI and how you're baking that into what you're building with the link. 00:38:36.299 --> 00:38:40.579 Yeah, I think this is everything I just said about this being an interesting 00:38:40.579 --> 00:38:43.179 year for SaaS applies here. 00:38:43.379 --> 00:38:46.659 So it's, you know, if you've got a product that's, especially if you don't have 00:38:46.659 --> 00:38:51.759 a ton of distribution, like a Procore or a Kahua, now is the time to say, 00:38:51.879 --> 00:38:55.979 how is this enhancing what I have already and maybe build a little less than you might. 00:38:56.499 --> 00:39:01.299 Like before you go build an entirely new category of product type, 00:39:01.539 --> 00:39:07.419 maybe hang on a little bit and see where the world ends at the end of this year or early next year. 00:39:07.499 --> 00:39:11.239 I think that we don't quite know what an agent's going to be able to do. 00:39:11.459 --> 00:39:14.879 The other thing that we've spent a lot of time on is aggregating data. 00:39:15.219 --> 00:39:21.439 So a database of usually publicly available specs that we're able to get from 00:39:21.439 --> 00:39:27.919 various places that we've categorized and really broken apart so that as AI 00:39:27.919 --> 00:39:29.639 gets more powerful and more useful, 00:39:29.979 --> 00:39:33.439 you have a, whether you call it a mode or just a differentiator, 00:39:33.599 --> 00:39:38.419 the ability to assess a spec based on 10,000 specs that, you know, 00:39:38.499 --> 00:39:42.279 that you're looking at, that you've got in your database. I think that's one opportunity. 00:39:42.659 --> 00:39:47.279 If you asked me that question in February of 2027, I promise you I'll have a 00:39:47.279 --> 00:39:50.719 different answer because it's just, we just don't know. Like it's. 00:39:51.374 --> 00:39:57.614 The fact that in December of last year, I couldn't have told you that co-work 00:39:57.614 --> 00:39:59.094 and open claw and those other things 00:39:59.094 --> 00:40:02.554 would come out and would be as big a difference as I think they are. 00:40:02.794 --> 00:40:06.894 And I think this year is going to be, this is the year of the agent. 00:40:07.034 --> 00:40:10.834 I don't know what that means, but the agent tooling, like making it so that 00:40:10.834 --> 00:40:15.194 people don't have to know anything about AI really to be able to just tell it 00:40:15.194 --> 00:40:18.914 what they want, go through some work to make sure it gets them what they want. 00:40:19.014 --> 00:40:23.154 But all that's using normal language. All that's just talking to it patiently, 00:40:23.154 --> 00:40:26.314 I have to admit, but nevertheless, just talking to it. 00:40:27.894 --> 00:40:32.694 How many OpenClaw instances do you have running right now? None. None. 00:40:33.954 --> 00:40:38.174 Lars? I didn't want to go through the, that's a great

question. OpenClaw, 00:40:38.374 --> 00:40:41.694 it means that you've decided to somehow either create a virtual machine, 00:40:41.694 --> 00:40:45.934 which is on its own, a little bit of a risk, or go and buy a Mac mini, 00:40:46.074 --> 00:40:48.314 which have been running out because so many people are doing it. 00:40:48.454 --> 00:40:50.014 I didn't see the value in that. 00:40:50.534 --> 00:40:54.434 What I've done instead is really dig into co-work and build a bunch of skills 00:40:54.434 --> 00:40:59.854 and plugins and so on and so forth, because I think OpenClaw is going to become very co-work-like. 00:41:00.034 --> 00:41:03.914 They'll probably do some unique things. They can do images, which Claude can't. 00:41:04.194 --> 00:41:08.714 But nevertheless, it's going to be in the same vein, and I expect it probably by the end of March. 00:41:09.693 --> 00:41:13.013 Yeah, I have one instance of open claw on a virtual machine. 00:41:13.493 --> 00:41:18.033 And, you know, I've kind of used it as purely an experiment to kind of see what 00:41:18.033 --> 00:41:20.873 it's possible of doing and are capable of doing. 00:41:21.273 --> 00:41:24.873 And, you know, mainly from like a podcast production standpoint, 00:41:24.873 --> 00:41:27.913 it's pretty helpful in doing research and those things. 00:41:28.113 --> 00:41:32.493 But, you know, it's not cheap because it, you know, it chews up tokens. 00:41:32.493 --> 00:41:35.393 Yeah, because the big thing is memory, right? 00:41:35.593 --> 00:41:40.293 And so that's a big kind of database that it's kind of maintaining from a memory standpoint. 00:41:40.613 --> 00:41:44.333 Well, you talk about, that's a really good thing, if you don't mind, I'll just dive into. 00:41:44.553 --> 00:41:47.453 And it's my point a moment ago about we don't know where we're going to be. 00:41:47.633 --> 00:41:51.933 One of the things that makes any AI model a lot smarter is if it can think longer. 00:41:52.153 --> 00:41:55.193 And the reasons they can't think longer is because that costs tokens. 00:41:55.673 --> 00:41:59.673 Every time a model comes out, and honestly, every time the chips come out, 00:41:59.673 --> 00:42:03.133 the cost per token goes down by like an order of magnitude. 00:42:03.593 --> 00:42:07.993 So I think what you're going to find is towards the end of this year into next 00:42:07.993 --> 00:42:11.853 year, as these big data centers come online, and then more of them next year, 00:42:11.953 --> 00:42:12.733 and then more of them the year. 00:42:12.833 --> 00:42:15.733 I mean, they're spending, the hyperscalers, the companies we know, 00:42:15.833 --> 00:42:21.133 the metas and so on, are spending something like $650 billion, which is wild. 00:42:21.453 --> 00:42:25.513 The 2008 bailout was $850, $900, something like that. 00:42:25.613 --> 00:42:28.333 I mean, We're almost spending as much money on data centers, 00:42:28.573 --> 00:42:31.093 just those companies, as the U.S. 00:42:31.253 --> 00:42:33.713 Government did to not quite get us out of the 2008. 00:42:34.393 --> 00:42:42.153 Like, if only, if only. But my point is the magnitude of spend is utterly unprecedented. 00:42:42.453 --> 00:42:46.093 Even inflation adjusted, the highways didn't cost this much at all. 00:42:46.293 --> 00:42:50.193 Forget about in one year. The railroads didn't cost it. It's unprecedented how 00:42:50.193 --> 00:42:55.213 much money is being spent. So I think that one of the unlocks over time will 00:42:55.213 --> 00:42:58.913 be having agents that can run for a long time and not break the bank. 00:42:59.033 --> 00:43:02.533 Right now, you're right. If you were to get open claw and just let it run. 00:43:03.329 --> 00:43:06.909 I hope you host your own model because it's going to be a mortgage payment. 00:43:07.209 --> 00:43:11.029 Will that overcome the quantity of available compute? 00:43:11.189 --> 00:43:17.749 Will that overcome the actual cost of compute with interacting with it? 00:43:17.789 --> 00:43:25.369 I was reading a thread and this gentleman calculated every time you ask Chachapiti 00:43:25.369 --> 00:43:30.749 or Claude a question, and if you look at the actual compute cost of that question. 00:43:31.641 --> 00:43:36.741 And if you were to actually be charged the actual cost, it's much higher than your $20 a month, right? 00:43:36.961 --> 00:43:41.361 And right now, everyone's in this, like, market penetration phase attempting 00:43:41.361 --> 00:43:43.101 to kind of gobble up users. 00:43:43.421 --> 00:43:46.321 But at some point, is that kind of the calculus they're making? 00:43:46.461 --> 00:43:50.721 Is like, if the quantity of compute is enough, then that cost comes down over time? 00:43:51.241 --> 00:43:55.141 And they've depreciated all of their spend and so on and so on. 00:43:55.261 --> 00:44:00.321 I think the idea here is just like AWS was loss-making until it wasn't, 00:44:00.321 --> 00:44:04.421 And just like Uber was loss-making until it wasn't, their thinking is, 00:44:04.601 --> 00:44:10.021 we want to get the whole world on one of the platforms, preferably the one who's spending the money. 00:44:10.221 --> 00:44:15.341 And at that point, you know, again, every time NVIDIA comes out with a new chip, 00:44:15.601 --> 00:44:21.601 the most recent one, Vera Rubin, is 40x faster and more effective than what 00:44:21.601 --> 00:44:24.221 it's replacing. And everyone thought Blackwell was the second coming. 00:44:24.421 --> 00:44:28.981 But also, beyond NVIDIA, there's companies that are making chips that are specific 00:44:28.981 --> 00:44:32.101 to inference. And the difference being training is when you do the training 00:44:32.101 --> 00:44:33.201 and then you publish a model. 00:44:33.501 --> 00:44:36.441 And that used to be all the expense. But now that everybody's trying to use 00:44:36.441 --> 00:44:39.461 all these tokens, it's actually on the usage, on the inference side. 00:44:39.841 --> 00:44:40.901 That's what

this, I don't know 00:44:40.901 --> 00:44:43.781 if you're familiar with the Cerebrus thing that OpenAI just announced. 00:44:44.101 --> 00:44:46.821 But they have their first data center with these. And those are. 00:44:47.381 --> 00:44:50.081 The entire data center is made just for inference. 00:44:50.301 --> 00:44:54.401 Well, the implication is that the cost of using them is going to continue to 00:44:54.401 --> 00:44:55.481 just drop through the floor. 00:44:55.861 --> 00:44:58.601 At some point, everyone's like, you know, it'll be too cheap to meter. 00:44:58.601 --> 00:45:01.441 They've said that before about other things, and I'm not sure I buy that. 00:45:01.661 --> 00:45:06.061 But I do think it'll get so cheap that it won't be the major consideration. 00:45:07.481 --> 00:45:11.821 So I know you've got a hard stop, and I want to focus maybe the last few minutes 00:45:11.821 --> 00:45:18.481 we have on, you know, there's a portion of our audience that is on the field ops side. 00:45:18.621 --> 00:45:23.181 And I think, at least looking at the data, we don't have a ground-up challenge 00:45:23.181 --> 00:45:26.661 in getting kind of users to adopt some of these technologies. 00:45:26.661 --> 00:45:31.721 But we do have like a senior leader and C-suite, which we have some listeners in that seat. 00:45:31.861 --> 00:45:34.421 So let's talk to them specifically. And we did a little bit before. 00:45:35.918 --> 00:45:39.638 They listen to this day after. What do you think a VP of Ops, 00:45:39.818 --> 00:45:44.378 a COO, CFO, what do they, I mean, you mentioned it, just start using some of this stuff. 00:45:44.718 --> 00:45:49.438 But from an organizational strategy standpoint, what are the practical steps 00:45:49.438 --> 00:45:54.578 to thinking through how they need to interact with where we're at in this curve? 00:45:55.198 --> 00:45:59.978 Yeah. And the tough part about a question like that is people really are all over the place. 00:46:00.178 --> 00:46:03.418 I mean, there are companies that where the leadership has already done a policy 00:46:03.418 --> 00:46:06.438 and they've done training and they're thinking about it and they've got internal 00:46:06.438 --> 00:46:11.698 people building custom GPTs and you've got other people where they haven't and 00:46:11.698 --> 00:46:16.498 they maybe put a policy out that was pretty simple and was very much about security. 00:46:17.218 --> 00:46:20.278 Let me re-contextualize that because I think you're right. 00:46:20.618 --> 00:46:25.678 So I think the scenario is this. I think it's a large, we'll say GC in this 00:46:25.678 --> 00:46:31.078 instance, who is, you know, their AI strategy is mainly from a risk management standpoint. 00:46:31.158 --> 00:46:35.298 So they're really focusing on governance. They're focusing on what not to do. 00:46:35.938 --> 00:46:40.158 How would you respond to that company? Yeah, I think that there's a few things. 00:46:40.338 --> 00:46:44.918 One of them is, like you said, get using it, but also get out there and talk 00:46:44.918 --> 00:46:48.658 to people and bring in people, whether it's a consultant, whether it's somebody. 00:46:48.778 --> 00:46:51.238 I mean, often I'll just talk to people, give them an hour and say, 00:46:51.318 --> 00:46:52.418 hey, let's talk about what you're doing. 00:46:52.758 --> 00:46:56.878 I've got a number of GCs that I just once a month, once every other month, 00:46:57.098 --> 00:47:00.218 someone in there and I'll just walk them through what I'm seeing, 00:47:00.518 --> 00:47:03.938 right? And right now with co-work, obviously, I've got a little bit of a bee in my bonnet about it. 00:47:04.018 --> 00:47:07.338 So I've been doing a fair amount of that type of conversation to say, 00:47:07.458 --> 00:47:09.638 this is early, but it's going to mean something. 00:47:09.678 --> 00:47:12.898 And it's actually going to be useful in a way that a chatbot is not. 00:47:13.278 --> 00:47:14.798 But I do think that this is... 00:47:15.825 --> 00:47:18.925 Start thinking about what real 00:47:18.925 --> 00:47:23.785 change of process might, what value real change of process might mean. 00:47:24.105 --> 00:47:30.445 And for me, one of the big ones that seems obvious is the chain of handoffs 00:47:30.445 --> 00:47:32.705 between design intent and handover, 00:47:32.725 --> 00:47:36.705 where you've got this information that gets transformed and people lose a thing 00:47:36.705 --> 00:47:39.225 or they do a thing and they say a thing and they were supposed to do this, 00:47:39.345 --> 00:47:40.145 but they didn't quite do that. 00:47:40.445 --> 00:47:45.285 All that stuff, the amount of risk that's still in that transmission of information, 00:47:45.605 --> 00:47:47.225 I think is unreasonable. 00:47:47.945 --> 00:47:51.685 And yeah, there's contracts and people are responsible for things and they have 00:47:51.685 --> 00:47:55.005 economic incentives for doing as little as possible at certain steps. 00:47:55.245 --> 00:47:58.985 I don't know who that might be, but maybe the people that wrote the contract. 00:47:59.445 --> 00:48:02.985 But the point is, fair enough that that's true, but nevertheless, 00:48:02.985 --> 00:48:07.005 this stuff should be a lot more controllable and visible than it is. 00:48:07.085 --> 00:48:08.745 And I think that AI can help with that. 00:48:08.985 --> 00:48:13.265 And a lot of our processes when it comes to moving documents and data back and 00:48:13.265 --> 00:48:18.425 forth look indistinguishable from when you were couriering prints to each other. That's correct. 00:48:18.685 --> 00:48:21.885 It's a PDF and you're emailing it, but okay, if you swap that, 00:48:22.005 --> 00:48:24.145 the transmission out, it's the same process. 00:48:24.385 --> 00:48:28.665 It's the same. And I'm not sure that needs to be true. You can absolutely audit 00:48:28.665 --> 00:48:32.165 custody and all the things that a printed paper might help you, 00:48:32.245 -->

00:48:33.345 in fact, way better than that. 00:48:33.725 --> 00:48:37.505 So I think that asking where's our risk and where are we losing money and where 00:48:37.505 --> 00:48:42.065 are our claims coming from and really saying, is AI able to help us tighten 00:48:42.065 --> 00:48:46.765 that up and lose less money from claims or from delays or from rework or from, 00:48:46.825 --> 00:48:48.285 everybody talks about reworking it. 00:48:49.169 --> 00:48:51.769 It's really harder than people think it is. But nevertheless, 00:48:51.789 --> 00:48:54.309 that's a good thing to start thinking about. 00:48:54.569 --> 00:48:58.429 And back to my point about the more you use it yourself, you're going to have 00:48:58.429 --> 00:49:02.129 an intuition to say, okay, maybe I don't know AI that well yet, 00:49:02.249 --> 00:49:03.629 but I really know that problem. 00:49:03.789 --> 00:49:07.789 And if I can get an intuition for the tool, I can start seeing where, 00:49:08.289 --> 00:49:10.369 all right, you know what, we're not going to retool everything today, 00:49:10.429 --> 00:49:15.609 but I want to inject it in these steps or I want this person to spearhead doing that on a project. 00:49:15.609 --> 00:49:20.489 There's always the case that there's that PM or that superintendent who absolutely 00:49:20.489 --> 00:49:24.089 loves technology and is the first, you know, he's like Mikey with the cereal. 00:49:25.069 --> 00:49:28.489 Finding those people and really leveraging them. One of the things that I'll 00:49:28.489 --> 00:49:35.189 also say, I do a lot of training for contractors, but also for associations and so on. 00:49:36.029 --> 00:49:40.869 The degree to which you can get 20, 50, maybe a hundred's a little bit much 00:49:40.869 --> 00:49:44.409 and get them talking about what they've done in a kind of moderated session 00:49:44.409 --> 00:49:46.769 where you say, all right, everyone, let's write a prompt and you do it on your 00:49:46.769 --> 00:49:49.049 phone. All right. What did you do? What did you do? What did you do? 00:49:49.899 --> 00:49:52.839 A couple of things happen. One is you realize more people are doing stuff than 00:49:52.839 --> 00:49:55.099 you thought, or maybe the other, you're like, what are you people been doing? 00:49:55.399 --> 00:49:58.759 But also they start to get ideas from each other because again, 00:49:58.899 --> 00:49:59.939 this is a creative group. 00:50:00.059 --> 00:50:04.719 These are people that anytime there's a problem, they want to find a tool and go fix it. 00:50:05.239 --> 00:50:08.799 So they're probably said, oh, if they got a little comfortable with it, 00:50:08.859 --> 00:50:12.399 they probably found some weird thing that they're using it for or some not so weird thing. 00:50:12.599 --> 00:50:15.559 So I think to the degree, you can start thinking about 00:50:15.559 --> 00:50:19.039 leveraging the creativity of your own people to 00:50:19.039 --> 00:50:21.859 to kind of inch each other up i think

that's super powerful and 00:50:21.859 --> 00:50:24.979 you've seen contractors do that with like hackathons or with meetings or 00:50:24.979 --> 00:50:27.699 whatever it might be it doesn't cost a lot you can do it 00:50:27.699 --> 00:50:30.559 usually on on you know in person is best 00:50:30.559 --> 00:50:33.479 but you do it on zoom if you need to i really 00:50:33.479 --> 00:50:36.399 like that it's almost it's like creating that spark of creativity 00:50:36.399 --> 00:50:39.199 and then kind of seeing where that goes and i've seen that even with 00:50:39.199 --> 00:50:42.459 my wife like she runs a small little co-schooling 00:50:42.459 --> 00:50:45.799 here in the richmond area and you 00:50:45.799 --> 00:50:48.639 know back and i started my kind of ai during early last 00:50:48.639 --> 00:50:51.979 year we i kind of walked through some of the use cases and 00:50:51.979 --> 00:50:54.939 now like we're trading back and forth it's like oh man you did this oh 00:50:54.939 --> 00:50:58.539 man i was able to do this and then running a home 00:50:58.539 --> 00:51:01.819 school this has to be transformational for her because 00:51:01.819 --> 00:51:04.579 like you're not stuck with wait wait who did what 00:51:04.579 --> 00:51:08.819 when it's like hang on a minute you know that's education 00:51:08.819 --> 00:51:11.819 i think there's a lot of opportunity yeah and she's 00:51:11.819 --> 00:51:14.699 you know the little the school has grown from like we 00:51:14.699 --> 00:51:18.099 had like 30 kids last year we have 48 kids 00:51:18.099 --> 00:51:21.339 this year but that's amazing so some of the parents are teaching themselves 00:51:21.339 --> 00:51:25.099 or teaching as well and part of the you know thing is like how do you manage 00:51:25.099 --> 00:51:29.819 like content and yeah in a very specific kind of methodology and and this is 00:51:29.819 --> 00:51:33.559 is enabling her to scale herself and in a way that she wasn't previously able 00:51:33.559 --> 00:51:37.139 to well this has been really helpful lars any final thoughts. 00:51:38.178 --> 00:51:41.818 No, you awesome. Thank you, bud. I'm pro core. 00:51:41.938 --> 00:51:46.798 I'm going to AI pro core or parkour myself here, parkour myself here. 00:51:46.998 --> 00:51:53.158 No, just, just love the perspective, love the sharing, love the experience. 00:51:53.158 --> 00:51:58.618 And I just, we are at this just very interesting time. 00:51:59.258 --> 00:52:04.638 Things are getting incredibly easier, but at the same time, it's like, 00:52:04.798 --> 00:52:06.058 it's also getting foggier. 00:52:06.058 --> 00:52:11.938 Yeah where is it going what what what what can it really be like and it's like 00:52:11.938 --> 00:52:15.618 just hold on let's watch let's see well and let's coat let's create it together 00:52:15.618 --> 00:52:19.598 together i mean at the end of the day this isn't this isn't the weather like 00:52:19.598 --> 00:52:23.218 we get to we get we get a say in how it gets used yes 100, 00:52:24.218 --> 00:52:25.958 well and here's the last question. 00:52:27.096 --> 00:52:30.556 How big of a prepper are you? Are you a doomer on this? 00:52:30.636 --> 00:52:36.176 Or do you see a hopeful future where we all are just skating on an amazing API? 00:52:36.556 --> 00:52:39.556 I think none of the above. I think it's going to be like everything else. 00:52:39.776 --> 00:52:41.036 It's going to make changes. 00:52:41.876 --> 00:52:43.296 Society is not going to collapse. 00:52:44.156 --> 00:52:47.536 We're going to find work for people. We absolutely will. This idea that, 00:52:47.776 --> 00:52:52.456 and there's an assumption that is in there. I'm really glad you asked that. 00:52:52.776 --> 00:52:55.756 People have talked about AI as if it's going to be general. 00:52:56.476 --> 00:53:00.936 I'm sorry, I know the lab guys have to say that, the open AIs and so on. 00:53:01.076 --> 00:53:07.196 I think they're radically underestimating how good the human mind is at dealing 00:53:07.196 --> 00:53:09.716 with new things and being literally general. 00:53:09.956 --> 00:53:14.096 I think we're going to have superhuman AI just like we've had superhuman calculators. 00:53:14.236 --> 00:53:18.376 And it's going to be amazing at some things and remarkably dumb at others for 00:53:18.376 --> 00:53:25.996 as long as I can think of. I think their use of the term general intelligence is just really narrow. 00:53:26.196 --> 00:53:31.356 And I think instruction is such a good illustration of how that's not true. 00:53:31.516 --> 00:53:36.336 Think of all of the implications that a PM, like something happens and they're 00:53:36.336 --> 00:53:38.676 immediately saying, well, that's going to happen and that's going to happen 00:53:38.676 --> 00:53:41.216 and I got to call that guy and got to make sure they have whatever. 00:53:41.216 --> 00:53:44.476 That sort of complexity there's it's 00:53:44.476 --> 00:53:47.496 it's it's not just knowing what might happen but it's knowing what's 00:53:47.496 --> 00:53:51.356 important and why it's important and the personality of the person you got to 00:53:51.356 --> 00:53:55.016 call first because god forbid you know what i mean like there's just knowing 00:53:55.016 --> 00:53:59.276 that that's the reality of of the economy whether it's in construction or somewhere 00:53:59.276 --> 00:54:03.936 else is it's a lot of people making decisions under uncertainty based on what 00:54:03.936 --> 00:54:07.156 they had for breakfast and only a human can really manage that. 00:54:07.336 --> 00:54:10.816 And I think AI can absolutely empower it. It can automate certain steps. 00:54:10.996 --> 00:54:13.676 And the final thing I'll add that we didn't talk about is. 00:54:14.358 --> 00:54:18.478 Being in the work is part of getting the work done. So this implication that 00:54:18.478 --> 00:54:23.458 you can automate everything implies that humans were just approving stuff or 00:54:23.458 --> 00:54:27.738 just running through steps, but they didn't think along the way. 00:54:28.398 --> 00:54:32.498 Doing things makes you're being creative, even if it's in small ways. 00:54:32.798 --> 00:54:36.578 Being in the work is the only way that you can approve certain types of work. 00:54:36.778 --> 00:54:40.618 So just having the human in the loop be an approver at the end, 00:54:40.758 --> 00:54:44.218 there's going to be processes where that's appropriate. I think people overestimate 00:54:44.218 --> 00:54:45.218 how many of them there are. 00:54:45.458 --> 00:54:49.058 So having AI serve things up to you, explain things to you, maybe give you a 00:54:49.058 --> 00:54:51.498 different view of something, maybe it's a graph, it's a whatever. 00:54:51.898 --> 00:54:55.038 I think that that's going to be way more powerful than people think. 00:54:55.338 --> 00:55:00.778 I think giving a decision maker unique or new ways of seeing information is 00:55:00.778 --> 00:55:03.978 going to be super powerful because we're way better at being visual than we 00:55:03.978 --> 00:55:05.558 are being numeric or textual. 00:55:06.158 --> 00:55:08.838 But automating humans out of it is is a 00:55:08.838 --> 00:55:11.938 mistake that i think a lot of fortunes are going to be lost trying 00:55:11.938 --> 00:55:14.858 to do because who's going to hang their hat or i would 00:55:14.858 --> 00:55:17.798 you approve something you didn't weren't part of making well 00:55:17.798 --> 00:55:21.198 it depends on what it is but there's a lot of things where if i wasn't in it 00:55:21.198 --> 00:55:24.678 i don't know like a spec writer forget it there's no way they're going to say 00:55:24.678 --> 00:55:28.618 it looks right those thousand pages read perfectly to who's risk who take who 00:55:28.618 --> 00:55:32.638 absorbs that risk that's right i love that i mean it's like we still need the 00:55:32.638 --> 00:55:37.418 man in the arena and we're just arming the man's arena with arming the man. 00:55:38.218 --> 00:55:42.318 So i'll end by saying i love this stuff so if any of your listeners just want 00:55:42.318 --> 00:55:46.318 a half an hour i i you know i love talking about it and i love i love where 00:55:46.318 --> 00:55:50.038 people are versus where i might be you know i just learn a lot from talking 00:55:50.038 --> 00:55:52.538 to folks that's great where do they contact you. 00:55:53.578 --> 00:55:58.518 Linkedin linkedin's great okay i'll yeah and i'll link the link the link link 00:55:58.518 --> 00:56:02.738 as well in the show notes i always commit to show notes and then i forget but 00:56:02.738 --> 00:56:06.278 um i'm not really gonna try this time to do that. Thank you so much. 00:56:06.598 --> 00:56:08.918 Hey guys, thanks for having me on. Great job.