Bill Canady: Hello everyone, Bill Canady here today with a thousand day CEO. Today we've got Brian McWhorter, and we are excited to have him in here. Brian is the founder and CEO of Bravura-AI, and his company and background spans plant operations, industrial automations, and engineering. His work focuses on making engineering information useful to people and AI. Today we're exploring his idea that AI is creating a new form of capital. And what that can mean for private equity backed CEOs value creation plans. All right, Brian, you have a great value creation plan, a board expecting results, and a limited hold point, Where can AI improve EBITDA and cash flow? And how do you prove the return before committing more capital with it? That's what we're gonna be exploring with you today. Brian McWhorter: Sure, Bill. Well, first, thanks for having me here today. it's it's kind of the origin story of the company. I had been encouraged to pursue a theory that I had for about ten years. And in twenty eighteen I was asked to join a team that was working on a project for the Department of Energy. And the great thing about government projects is there's no real schedule or budget. It's just make sure it's done right. And that allowed me the time and budget that I needed to pursue what I had always theorized would be the best way to do this. So what we were challenged with was to take the configuration of every microprocessor in the plant and turn it into a data set that could then be reported on and audited in order to meet nuclear regulatory commission requirements. Every five years you have to go audit your configurations. And you use paper still. And so it had to be something that could be printable. What that allowed me to do was take these complex configurations that we were having we were using around the plant for everything from a simple valve or a switch all the way to complex PID control loops and and model predictive controllers. And we were able to take the the different configurations that they come in and turn them into a data set that is more or less homogenized. And but and in order to do that, I had to solve a lot of different types of problems. What data structures were how to access those data that data and everything. But once we were to the point where I had that data in place, before we could even meet the challenge of getting it all into a printable report, they started asking me how to solve other problems. Like we need to track the the upgrade of some firmware on some different microprocessors that were operating a set of valves. And then we had to track, okay, these valves are having some problems. What what's common across those valves that's that are having those problems? And what we ended up finding was that it was not a network or it wasn't EMF interference, it wasn't some crazy PhD guy in the plant looking around for, you know, magical interference or whatever. What it turned out to be was there was a certain bit in an integer in a particular firmware version, in the presence of other alarms, we were able to then say that's what was causing the problem. And so I called the manufacturer and said, Hey, I think I found a commonality here. What would happen if if this were if this bit were set to true, all the other ones seemed to be false. They explained it to me, sure enough, they described the behavior that we were observing and I said, Hey, I think we've got the answer here. my my profit margins for my my little engineering boutique firm were going up, spending less time on site, delivering better reports, you know, things just everything got easier and better and more profitable. And I said, Hey, I've got something here. Bill Canady: you really did. Now this is a great example, but this came before your later AI work that you do today. So what can you do today that you couldn't do then? I mean, you you really started Brian McWhorter: Yeah. Bill Canady: seeing things and now you got a whole new set Brian McWhorter: That's right. Bill Canady: of tools. Brian McWhorter: Right. So so what I like to say is we have two ingredients to our secret sauce. But the first ingredient is what I just kind of explained crudely there. We take different T types of file formats. we've we've built transformers to turn those into data sets. And once it's in the what's it's in that data state, now it's somewhat homogenized, and that's ingredient one. Ingredient two is then we take and we use our engineering knowledge from the the team of engineers that I've had with me the whole time, and we put enough information around those individual elements so that even though they're homogenized, the large language models are able to navigate and find where they are. And because they go to the same place in the database every time. Regardless if it came from an offshore oil rig or a pharmaceutical plant or an oil refinery, it does not matter. They always went to the same place every time. They always got that same piece of context of that inf that not that engineering knowledge every time. So now Claude or OpenAI, Codex, whatever large language model you might want to use, can always connect to that s that and find that information that it needs every time. And that's where we've really turned the corner from being just a bunch of data that allows us to do our job a little bit better to now we're able to start solving problems that would otherwise requ require a lot of human effort, a lot of, you know, a lot of dedicated human time to do it. Bill Canady: Absolutely makes sense. So you've spoken about how AI is creating a new form of capital. What exactly did you mean by that? Brian McWhorter: I so I have a theory that I'm pursuing. I've bounced it off of f several different academics and some corporate accounting departments and so far no one has said you're crazy, get out of here. So I'm I'm starting from a consortium to try to pursue this formally. But the theory is basically when you have that knowledge base, let's say, in a state. is organized in a way that it can be used over and over again. Now you've established an asset that is separate and apart from what we used to think of as okay, you have this series you have this group of artifacts, be it designs, be it configurations, procedures, your maintenance database, all these artifacts. And then you have the humans that are using those artifacts. And they're that you know and and that's what we think of as operations. It's it's humans using those artifacts. What AI is allowing us to do, though, is separate the data from the artifacts and the knowledge from the human mind and put them together in a way that has allowed allowing those humans to use that combination repeatedly in a way that's more uniform, it's more valuable, it's removing liabilities. And in doing so, I think you can make an argument that that is a capital asset that improves itself. Okay, so unlike a server. Or a piece of IP that might have been written and then it has to be serviced and maintained, or the the machine or the the server that you buy and it degrades and and you depreciate it over time. We are building these engines in a way for some pretty big I can't I can't share names publicly yet, but some pretty big names have already commissioned us to build these knowledge bases in a way that improve themselves as they're used. Okay. So as the individual maintenance technician, operations manager, process engineer, they interact with this knowledge, that engine improves itself because it says that's how we used it, now we can have somebody else use it that way. It's improving itself, okay? So it's appreciating in value. It's not depreciating. And I I think that we have a argument to say, hey, that is an asset unto itself that we can recognize. And I'll leave it to the accountants on the best way to do that. But I think I think it's worth looking into is certainly to if nothing else turn it into its own profit center. Bill Canady: No, that that makes that makes sense. So is it the data piece that you think the the company can monetize, or is it what you can do with AI looking at that data? Where's the value come from? Brian McWhorter: I I think we have to treat it like a knowledge base. Is it's the corporate knowledge base or the organizational knowledge base that becomes a knowledge retention mechanism that allows for the capture of knowledge from you know, we all know what happens when a senior person either goes on to another job, retires, gets sick, who knows. Okay, and then and then we're left with the best practice hopefully was imparted to the other team members and the younger generation that's coming up to best implement and use those that that best practice along with those artifacts in order to continue the organization operations as best they can. But when you've now separated the those best practices into a a knowledge base and the data Bill Canady: Mm-hmm. Brian McWhorter: artifacts into that knowledge base. Okay, now you have something that's not a human and it's not human capital anymore. It's not intellectual property anymore. It's not a part asset that's going to be depreciated anymore. It's something unto itself. And that's where I believe we can we can start to treat it as an appreciating asset. And it's going to have I mean, you can audit it, you can measure it, you can you can prove ROI on it. I think I think we have a strong argument to make. Bill Canady: You know, I it's it's it's a key point. The the value stays with the company when a key employee leaves. do you find that that you know is it the the folks who are taking over you you're talking about your younger folks, whatever, do they need special training or whatever, or can the can the software really accommodate anything? Brian McWhorter: It's it's very ad it's very easily adopted, I'll say. What what's and I think we're going to we've only seen the beginnings of this. I've been working with for several years now and and I'm so I'm probably a little ahead of everyone else. And even I find myself thinking about it differently than I did even six or twelve months ago. So I think Bill Canady: It's moving fast. Brian McWhorter: I mean we're gonna see a h a tremendous change in the way people think because you're now I I'm now thinking of my computer as a number of assistants. That's another key an another key factor to this is I think we're going to see pretty soon a lot of effort being put into how to optimize those tokens. So in other words, how do we balance the load on what goes to the larger language model that's gonna burn tokens versus what can we pull in and run locally so that we're not burning tokens any more than necessary. And I I think there's gonna be some interesting optimization engineering going on with that. Bill Canady: So if you were and what you do all the time, talking to a a CEO today, private equity or whatnot, where would you suggest they start making their differential investments? Like what should they be doing in this space? Brian McWhorter: Okay, I'm biased, so first Bill Canady: Well we expect you to be, so b go with it. It's perfect. Brian McWhorter: know that you're looking at this through my lens here. I think that the way that people need to start approaching the way to use AI is a little bit more holistically than what we've seen so far. in other words, a lot of what we've seen so far is hey, here's a task or you know, a set of work functions that need to be done. Let's figure out how we can get AI to automate, facilitate, make that task more repetitive and and the outputs more uniform and whatever. I okay, maybe not a that horrible idea. Maybe it shows some results, but I think of that as kind of like trying to measure your crop yield in July. Okay, I'm from Illinois, I grew Bill Canady: Mm-hmm. Brian McWhorter: up in and the what I mean by that is the maturity of the way we think about AI. And the real benefits that I I'm seeing in this, what I'm calling this, this knowledge capital that I think we can start to really leverage at a at a self-improving corporate level. Okay. If you're only looking at it as, well, this job rule got somewhat automated and these tasks got streamlined or whatever, you're you're trying measuring the wrong metric. And we're and we're not nearly inmature enough to see what is the massive value that we're going to be able to capture. Right. And so A more holistic approach to it would be something like All of the designs, all of the configurations, all of the procedurals, be it your maintenance or your your quality procedures, your some of the the operational functions that I'm I'm familiar with from earlier parts of my career, you know, basic your PSM management, things like this. Turning all of those things into a knowledge base that's able to be interacted with, that's where we're going to first start. And from there, then I think the humans are gonna be able to find creative ways to use that knowledge base. And that's where you're really gonna start to see every individual start to put a nickel in the jar and that jar is just gonna keep growing and growing because every person's gonna find a better way, especially with the right training, they're gonna find the better way to do their job given that they're they have access to this well built knowledge base that they're gonna be then enabled or empowered to leverage through making their product you know their their their deliverables more uniform, finding more efficient ways to into you know undertake some of their daily and and weekly tasks. I mean you know you leave it you leave that creativity part to the humans, right? That that's Bill Canady: Mm-hmm. Brian McWhorter: the human mind is still far more powerful than any AI engine. I I don't care what anybody says. I'm finding a lot of interesting to me, like just sort of like as a observer of human nature or whatever, like I do find a lot of inter I you know, I was an automation engineer back in two thousand two or three or whatever, and I was w you know, and and Bill Canady: Mm-hmm. Brian McWhorter: young kid, young engineer worried am I am I taking, you know, the jobs of these guys who have kids that are my age and you know, boy, I that was a weighted on my mind a little bit. And finally I realized I'm I'm not We're not taking their jobs. We're making their jobs a lot easier and we're allowing them to do better and work and and the quality of work they're allowed they're able to do is much better. But I don't Bill Canady: For sure. Brian McWhorter: think of a time in any of my automation career that I actually took anyone that I I feel like somebody's job got eliminated by the things I was working on. And I think AI is gonna do the same thing where it's gonna allow a lot of people to do better work more easily, and it's gonna allow them to do work that's much more engaging and you know. fulfilling in their life to say, hey, I'm not, I don't have to get beaten up by my boss because I I want to do this good work, but between budget and schedule, I'm not allowed to do it, but you know, and I've got to compromise myself. Nobody likes being put in that position. AI is going to allow us to deliver top notch work and do it within budget and schedule. And I that's very interesting. But think about fundamentally what I'm saying here is leave the creativity part to the human. And let the machine do the grunt work that we didn't really like having to do as humans before. Bill Canady: I think what it does is it frees up us to do more high higher value AI for a lot of people really turn out to be like admin work, repetitive work and you know, doing some research and and it and it and it kind of does a nice job with that. how would a CEO today make a decision of where to invest versus where not to? Do you have one concrete example that you've seen? That really makes a big a big difference. Brian McWhorter: Okay, we built AI for industrial manufacturing. And so I I can only speak to it with a lot of expertise from there. But I will say that allowing your design, your, your, you know, a lot of the plant functions and and and you know, just go through the technical personnel that are that you that are part of the organization in that environment, the the industrial manufacturing environment. And start to look at what are the types of things that they do that cause them headaches and what could be done to facilitate and and remove those headaches. What we will find is almost always it's gonna be access to information. And I think that's the answer is the inve the where the place to make the investment is Enabling and and and and streamlining access to information that those those people need. Okay, and I that may seem disconnected a little bit, but in you know, that's really what causes a lot of the headaches is huge fractions of time get spent just going to find the right design or finding the right alarm value or finding who knows or the right serial number, model number to replace the thing that just broke or whatever it is. That is what AI is going to take and That's grunt work in my opinion. That's technical grunt work. AI makes that stuff easy if you have it put into the right place. And that's what I think that's going to be one of the first places we see lots of improv you know, improvement. And so the investment needs to be made in, at least my my perspective is the investment needs to be made in using AI to make the information more available in the right form and with the right organization so that the the humans are then allowed to do their jobs and, you know, like we said, engage in engage in the activities that are more value added, more engaging, you know, allow them to do the work that they that they they feel comfortable and and allows them to perform at their highest levels. Bill Canady: Where where would you tell a CEO, the leader of the business not to invest in AI? Yeah, an example maybe of where you found it maybe it just wasn't ready yet or maybe it's not mature enough. Where w if you're sitting there in front of today, where are you going to say Brian McWhorter: Yeah. Yeah. Bill Canady: don't go this direction? Brian McWhorter: Well, there's sorta there's there's sort of the anecdotal approach to it, right? Is like tr trying to find trying to say, hey, just go figure out how to use it. you know, that's that's kinda like telling you know, one of my students, go into the library, the books are over there, I'm gonna write you a test, and if for every every question you get wrong, I'm gonna smack ya and then go find the right answer. And like that's that's the totally wrong. We don't teach humans to do things that way. And that but unfortunately I've seen some some organizations are trying to do things like that where it's like, yeah, just go figure out how to use it. And then they try to they try and then they come back and they say, didn't you know, it didn't work very well. And then, well, it was a failure. We're not we're not happy, blah, blah. You it's kind of the equivalent of getting smacked. And that's that's not that's not the way to do it. The other one though, I will say, and this is this is a a serious undertaking that the whole I mean I mean not not just segments, sectors are gonna have to deal with this. And we've run into it several times already ourselves. Is the security that we're going to face, okay, the security challenges are are real. And you know, we've seen this a little bit. We heard about the water plants that got hacked or whatever. you know, a couple of years ago there was the Liberty Pipeline, I think that I think it was what it was called. There was the solar winds, here's the the problem that we're we're facing is that most of the PLCs that run these plants and pipelines and power generation plants and what they're built on relatively old computing process technology. Okay. And so there's lots of free memory space in there that can be exploited. And that creates a problem because then if there's any exposure at all, once the intruder's in the door, he's got a lot of lot of free, you know, f free memory, which is like the the the booty to them, or the the you know, the gold. and that's a problem that we're going to have to figure out how to deal with quickly. making the the wrong investment first is probably worth considering and making sure that you've made the investment in the security so that you don't run into the wall of, you know, you say, hey, invest in AI, and then you invest in AI and find out you can't go any further because this there's too much security threat to it. Now. Bill Canady: security one is a really important one. It was actually what I was gonna we were gonna go to next is this scares a lot of companies and and and it should. How Brian McWhorter: Sure. Bill Canady: do they control it? What do they do it? You mentioned a couple of examples and or a couple of things. Can you give a maybe a a real example or or or an illustrative example? Brian McWhorter: Well there's a couple of simple ways to think about it. Basically think about security as an umbrella and you have a couple of different approaches. You can either buy another umbrella and try to get be get under or you can move the other person or the in this case the LLM under your umbrella. That's a simple Bill Canady: Mm-hmm. Brian McWhorter: way to what it is. There Dell is offering an approach to say, at least they called me the other day and said, But basically what they were saying is we can provide you the compute resource to move under an umbrella and put it on site. Which honestly, back two or three years ago when I actually started thinking hard about how would we do this, that's the way I thought that they would want it done. And so I had kind of thought about that already. Microsoft has another approach though. They're saying, Let's just build a bigger umbrella. Let's build a or another umbrella and let's let's make it monolithic and we'll build a whole tarp or a canopy so that everybody can you know, and we'll build it all the way out to the edge where Where those PLCs are operating. And so that's another approach to it. So there's th it's not like it's not being addressed, but I do think it's important that, you know, people that are making investments in this technology make sure they make the investments in the right order, right? So that first you're secure and then that way the full value of the AI can can start to take hold. I in the meantime, there's plenty of value to be had in saying, Hey, let's take what we have and we can put it into a secure place and use it there right now. it's probably in the millions of man hours that can be saved by just putting it in a s moving the information we have into the right place and and then letting the engineers and in particular in oil and gas and pharma where I've where I've done most of my spent most of my career, I know exactly where you would start to save the millions of man hours in those two seg segments. Bill Canady: Where's the bigger risk today? Is it your data, the company, your data getting out and people getting access to that? Or is it creating, like you were saying, you know, opportunity for them to, you know, kind of worm into your system? Because like many of the companies, they'll have good firewalls, but once you breach that, you have free reign of everything, right? So where Brian McWhorter: This is Bill Canady: where would you be focusing? Brian McWhorter: I well the the main the major threat is that in that second one. Okay. And the the reason for that is Like you said, once they once they've penetrated, now you have exposure to a Bill Canady: They can take all your data it shut your whole system down. Brian McWhorter: right. They can lock you up, both the Home Depot and the Target hack that happened years ago where they got a hold of a lot of credit card numbers and stuff, that was they came in through the the build the the security system, the the building management, the HVAC system, right? Like because once you're through like there's a lot of good re computing resources available there to be u to be exploited. you know, so that's the kind of the kind of thing where there's the security has to be built. Like I said, there there's I i I'm not a security expert, so I can't really say which way I think is gonna prevail. I will say that I'm surprised at the low cost, and at least there's there's there's gonna be some interesting dynamics here. I think it's feasible to say that the sort of what I called the Dell approach, just because they called me the other day and asked about it. But basically building a server rack that has enough compute capacity to bring the large language model into your office or into your server room and be able to work with it from there, given that, you know, we're not we're not trying to write Shakespearean sonnets here. We're we're working with a relatively known set of information that we know how to work with as Bill Canady: Mm-hmm. Brian McWhorter: a that does not require a massive, you know, petabytes of of of data and model training. So, you know, it is feasible for something like a hundred thousand dollars to move the server into the server room and with the prop with and that will have plenty of storage and compute capacity to run that model right on site. So, you know, if you think about $100,000, well, that's less than most engineer salaries. And if you know, if you if you're saving more than 10% of the time of every engineer and you got more than 10 engineers on staff. Bill Canady: Adds up quick. Brian McWhorter: that's a relatively low cost for that kind of a you know, improvement in across the team. now, that said, we've one of my close advisors is a a a high ranking and and some retirees from Microsoft, and they keep me up to date on the way that Microsoft is approaching it. And I think the difference is gonna be earlier on, it's the kinds of things that I've I've described doing. The Dell approach is going to be fine because what we need to do will work just fine for that. Bill Canady: Mm-hmm. Brian McWhorter: If we want to get to the point where we have what we what I've heard called the dark factory, which is basically, you know, you the factory that runs so much on its own you can shut the lights off and there's no humans in there anyway, and so the thing can just run on its own. What you have to have in order to do that is what although sometimes gets called like gen general AI, Now you've got autonomous decision making, measurements coming in, logical processes going on, commands going out, and judgment in between in between. Okay. That's gonna require that monolithic security approach because you've got the you've got the the processors on the edge, the machines that are actually doing the work, you've got the the the judgment or you know the large language model that's being able to perform the reasoning and then you've got the logical stuff in between. You're gonna have to have a Bill Canady: Mm-hmm. Brian McWhorter: broader security platform in order to do that. And it's not gonna be like just, hey, let's bring stuff under our umbrella. It's gonna have to be, we need something bigger than an umbrella. We need an entire canopy and that's the Microsoft approach. And so I think, you know, near term, probably if it if people were gonna invest that way, they'd probably say the Dell approach would be better. If the long term, I think Microsoft has that long term vision and of course it's Microsoft. They've got plenty of time to and Bill Canady: They they th I think they both got plenty of money. So maybe maybe we'll be be better for it. So all right, so final question. So you're a CEO, you've heard this and and maybe you've got an untapped asset here. What do you do? Like what what would you diagnose over the next thirty days so you can figure out where you're gonna go? what are you gonna prove then by day one hundred so you feel good about what you're doing? And then and then what do you want to be good at at a thousand days in three years, right? Because it takes time to get this stuff. So and it's changing so fast. So, so so if you're that CEO, thirty days, a hundred days, a thousand days, what are you gonna do? Brian McWhorter: Yeah. So I okay, I'm gonna recall recall upon my career in process technology development. Okay, I I spent a lot of time in biofuels and we were we were doing cool stuff, turning sugar into gasoline, turning diapers into ethanol and stuff like that. And one of the lessons I learned from the researchers that were doing that was keep it small and you eliminate as many unknowns as possible and you don't worry about economics because we will figure out how to deal with the economics later. First we need to prove the technology works. So within thirty days, I would pick out a couple of tasks that we we know we can grab the knowns that we have, put the logic process to it that we need, and start to prove, okay, it can it can be done. Can it be done? And we're actually working on a project at Stanford right now doing that. Same for for Well, we'll just say it's standard. Okay. And we're doing and and that was almost exactly the same request was hey, you got until the end of the year, prove to us you can do this. And basically what they want to be able to do is ask questions in natural language and get responses back that are something that a maintenance mechanic or a it's a from their f it's a facilities management group, you know, the maintenance mechanic, a security person, you know, some they want to be able to ask questions and get answers from the live data that's happening. Okay, that's like you said, maybe 30 to 100 days. And the way we're doing it is we're eliminating a bunch of unknowns, putting it into a safe spot where we know we can we can wreck it and nothing's gonna really, there's no consequence to to wreckage, but we're gonna be able to prove that it works. Within a hundred days, I would then expand that to, okay, let's let's now start to do some things that can reason on their own. So we can say not only just ask a question and get an answer, but we can ask. a question that requires two or three different inputs and start to have it give us back, you know, enriching answers or or more robust answers. That will then prove that okay, there's merit to it and it's it's it's going to be able to do things that otherwise would require human thought and and I won't say judgment, but it will help humans do its job because it took a lot of grunt work or you know non value added work off of the human and allowed the human to do more valuable things. Bill Canady: Mm-hmm. Brian McWhorter: I think that's achievable within a hundred days. But boy, within a thousand days, I think what's important is every every person on the team that's going to be using it needs to be trained on how to think about it properly. Okay, because very important from from our experience on our team. Making sure that the people using it can think about it properly is key to actually showing its value. Because if you just say, here guys, start using it. I I I could tell you hard knocks. We we did this. Some guys said, great, and they just start using it and then okay, we got this stuff. But then we had a couple of, you know, real deep thinkers and one guy's getting his PhD. They were able to really organize their thoughts around here's the best way to do this, and here we're gonna divide this up into subtasks and and we're gonna apply the right j we're gonna allow the machine to do things and then and then we're gonna build another agent to back check the thing that just did and what that That is going to be very important and I think that's what is going to be the difference between the hundred and a thousand days is the teaching the team at scale how to think properly about the way to implement the technology and the way to use it best. And that's when we're going to start to see every individual put a nickel in the jar every day and value of knowledge value for the company is gonna start to grow exponentially, right? Because it's Bill Canady: Kind of compound on itself. Brian McWhorter: it's like gonna be six cents and seven cents and it yeah, it's it's we're gonna really see the an explosion there. Bill Canady: Brian, you've given us a lot to think about, particularly the idea that if you're running a company today, you may have a an asset, a hidden asset potentially that you haven't figured out to monetize. AI today potentially gives us a way of doing that. And it's happening in real time. So for the CEOs listening, I hope you can take one idea from today's conversation, and define the result that you expect. We'll include Brian's links in the How would they contact you? Brian McWhorter: Sure. So I'm the you know, founder of Brevera AI and we've developed a system that takes all of that that different type of information that the the users around the industrial facility have at their disposal now and puts it into a place that allows them to start using Claude or OpenAI to work with it and It's called Plant Unity. We have a YouTube channel. We have you know a website. We are about to release a really interesting agent on the website where you can chat with it and it will tell you, hey, if you're in this job role, we can help you do this. And so you can contact me through that. Bill Canady: Fantastic. We'll include all those links in the show. So again, thank you so much, Brian, for being here with us today. I've certainly learned a lot. Take care, everyone. Brian McWhorter: Thank you, Bill.