The most valuable data is not in your database

Yes. Mikolas, it's great great to meet you.

Yeah, no, it's great. I I grew up a technologist. most of my early career was both building technology and with McKinsey advising companies across different industries on how to, you know, kind of grow. So a lot of things around growth strategy, go to market, new product strategy. And over time I became more and more fascinated with industrial companies because we I think there's tremendous potential within industrial companies that have, let's say, large workforces, large service teams, lots of machines, just incredible amounts of data coming off of the machines, right? And and I felt like there's enormous value creation potential and that's what drew to drew me to it. And so over the over the last decade and a half I've I've primarily focused on that and just, you know, had a had a had a really terrific time helping helping a large number of companies. So

Sure, absolutely. So in in you know, a lot of industrial companies, like I said, have large number of machines, large number of technicians, service workers. They a lot of them have ERP systems, CADA systems, and literally millions and millions of data points that they generate all you know, both historically and in in real time. What's interesting is that because of the scale of these businesses, even a small improvement right in one or two areas could mean, you know, millions and millions of dollars of, you know, kind of new value, if you will, not new you know, Ibita if you so what I what I'm fascinated by is not necessarily technology per se, but how technology could be applied, right? In a very smart, targeted way to help these companies become much more profitable, much more successful. And so and yeah Let me let me pause there.

Sure. especially today, I think there's a lot of pressure on companies to do something in AI or with AI, right? but where I like to start is to try to understand what are the leverage points. We call them value levers, what are the things where there might be opportunities for net new revenue, cost savings, operational efficiency, right? Right. So if you take a large service company, for instance. It could be how efficiently they dispatch, right? When they go on service calls. Or it could be what equipment configurations to put into a particular site, for instance, right? These are all of these small decisions. And historically they're made using only a small amount of data that they have, right? And and lot of the knowledge, the real value and the intelligence is not necessarily in the in their databases, however. Larger data set they have. A lot of it is embedded in people's heads, all of these operators that may have been working 25, 30 years, right? Have a lot of inside intuition where they go and make real-time decisions. So real literally hundreds and thousands of decisions get made every day. But number one, they're not standardized. There may not necessarily be the best decisions, right? And it's there's a lot of tribal knowledge that isn't rolled into the organization in terms of how to improve. The board. So I think the real opportunity with AI is to combine that human knowledge, right? To try to capture it all of these. What are the four situations in which you see vibration patterns that lead to failure or quality issues, right? Combine that type of human knowledge with all of the data so over time you have very intelligent decision-making models, right? But again, it's not the outcome is not the model. The outcome is can you make better decisions on a day-to-day, minute to minute basis, right? That drives profitability, for instance.

Yeah, I I think there's there's a there's a there's a number of them, which is, you know, if you if you look at any any large company, there's usually about twenty to thirty leverage points where you can apply AI, right? So a lot of companies start to invest early in a large portfolio of AI opportunities, right? That's probably the wrong place and the approach, right? The right approach is is to very carefully but quickly do estimates, right, of where the real leverage points are. What are the five, six things that could create the most value? Right. And then the next thing is: do you have the operational data? Do you have the the people that you can gather and do pilots? So what we like to call operational validation, right? Don't go and build something and scale it up expensively, but do small pilots to see if you can make better decisions, combining the data that you have. the data may you that may you have that may not be clean that you can clean or add additional data you can get, combine it with people that are making those decisions. How do you combine them and start to make better decisions? So in a in in you know generally in a three, four months or you know relatively short period, you can validate that you can actually capture the value. And that's when you want to start to, you know, kind of invest in the the right data infrastructure and the technology and to scale that up. And and generally you know You know, when when you do that cost estimation, you might like the 30 opportunities might mean, you know, there could be a couple of hundred million dollars of potential value, but a lot of those are very difficult and not very feasible. And so, but if you focus on the five, six things that really matter, you could you could do very well, you know, make twenty, thirty million dollars of new revenue, for instance, or new new operational.

Yeah. No, absolutely. So if you if you think about modern software or hardware companies, they've they've they're relatively new and they've had a clean sheet ability to build their organizations from scratch. So there's not a leg lot of legacy issues and systems that drag them down. Right. So you can you can, you know, very easily take a you know team of software developers and move them to agile and do cloud based development. You know, the we we've seen f you know, eight, nine generations now over the thirty years of companies, completely new type of companies starting up, right? When you excuse me for a second, let me get my so we may have to edit this part out. Yeah. So yeah, yeah, yeah. So in in contrast

If you think about a lot of manufacturing or B2B services, so kind of traditional, you know, blue-collar industrial services, there are a lot of legacy companies that may have decades and decades of old systems, right? Even in a factory, you may have very modern machines with modern sensors next to very old machines that can't be instrumented, right? Each factory may have different variations, right? And when they were built. Okay. Also, you have a very uneven workforce, right? You may have very senior people, right? that are the people everybody go to, but then you may very have very young people running night shifts that are making different kinds of decisions, right? So they're very messy, complex things with lots of moving parts, lots of uneven data, right? and lots of, you know, fragmented knowledge or you know, tribal knowledge is like. So there's a huge opportunity to kind of normalize that, right, if you will, meaning bring it all together and to be able to make smart decisions that that that create value. So that's you know, that's it's it's a much more challenging, you know, environment and and and and I I personally enjoy that that complexity.

Yeah, yeah, exactly. So so for instance, defects in quality that go to product waste, you know, you know, scrap is is is one of them. a lot of times you'll see downtime, right? you'll see a lot of inefficiency and when they you know reset a line or change production, right? Or you have to re-refit a line for a different variation of a particular product. There's a lot of inefficiency there. also there's a big gap between people who do the production planning and then people who execute it in terms of how efficiently that's done. Right. there's a lot of issues with inventory and supply chain and and trying to understand what you know what parts are needed to start production right away. So that's there's the there's always that those type of issues. Then there's a lot of personnel issues, like I said, right? You have peep different people with different pockets of knowledge, different experience levels, making different decisions at different times. I mean, a lot of factories we see, you know, a night shift, the output and the quality might be very different than the two other shifts. And it's simply because of the people, right? Even though all of the data is the same and all of the, you know, the operating procedures are the same, it's very different. So those are the value levers. When you now get into service type companies where you have a lot of machines, let's say installed and you have a lot of service technicians, right? It's about you know, are you how are you prioritizing how you dispatch, how you roll a truck, right? Do you have the right because you might roll a truck and they might drive three hours to fix something, but it may be the wrong repair person, right? Or they may not have the right part with them, right? So being able to be smart about it, being able to get signals from the field and optimize all of it. Right? When to go, what to do. So so yeah, those are those are the typical things. And then you get into things like pricing, right? Are you are you pricing it? Are you are you you know replenishing it? Are you managing downtime, customer satisfaction? So yeah, literally, I mean there's there's much more than 20, but but those some of the things that I mentioned are are are very common. Then of course you have things, I think I think everybody talks about predictive and prevented maintenance, right? There's a lot of work on that. There's a lot of new sensors that have been put in, right? To a lot of machines. some of it works, some of it doesn't. I think it's so far it's it's under delivered, right? But that's typically a value lever that we we look at, which is can you look at different temperature vibration signals, right? And and and understand that that machine, you know, if given those four parameters and the pattern that we see, there's a likelihood of that failing in the next 72 hours. for instance, right? That's typical. Or let's say that you're you're try you're starting to you know manufacture you know something let's say wood or paper or something like that right just the temperature right or the quality of material right can have significant impact so are you are you now adjusting all of your production parameters right to to take that into account right so yeah the the things like that one other one other thing I think is that that I'm seeing more and more more clients looking at is a lot of times there's a big delay between when something, let's say you're you're you're you're you're doing you know, explastic tube in extrusion, right? So an extruder, and then all of a sudden you have different variables that got go off kilter, a lot of times they don't turn that off for hours, right? And and only then do you find out, hey, last four hours all the material is subpar or defect or it's it's you know it's not for the the very valuable market. Is there a way to what we call, you know, do real time closed loop edge detection, right? So how do you bring AI models into the edge, right? And make a decision on, you know what, the the parameters are off rather than run it for two hours, right? And and waste all of this stuff. Can what can I turn this off now and fix it? Right. So those are the those are the kind of the more emerging high end things that that we're starting to see.

Yeah, I think I think you know, you know, we're we're using the term AI, but for for a long time we've looked at large amounts of data coming off machines. So if you think about 15 years ago, we started talking about IoT, then advanced analytics, we built these data warehouses, we can do machine learning where we can look at forty, fifty uncorrelated variables and try to understand either, you know, you know. cor correlation or ca causation between different you know variables and outcomes. So we've been doing versions of this for a long time. I I you know we we're calling all of it kind of more or less AI now, right? what I think is going to happen is first of all, you know outside of kind of what everybody's excited about, which is kind of the large language models, right? you're going to start to see each company take what's their proprietary knowledge and build models around Right. So every but because every company has, you know, unique distinctive intellectual property, right? In the form of data, the the the what's in their in their employees' heads, right? So how do you build models and enhance those models? And then how do you borrow things from the outside world that kind of enhances that? Right. So you'll start to see that more and more. I mean the the 10-year trend is if I I like to think that as everything goes into AI. Right, whether it's small stactical things or strategic things, people are gonna stop referring to it, right? Because, you know, if if you think about car breaks, right? Like in the nineteen eighties when disc brakes first came, everybody thought said, I have disc brakes. Nobody talks about it because that's expected, right? So I think we'll get to a point where it is understood that AI is is helpful and everything and you put it in the right way and you don't care about it and you don't worry about it. It's just a you know, it's just a Another advanced technology we embed into every little thing, right? And and so yeah.

Hmm. I'm not I'm not sure I would I would necessarily do anything because I you know because we are advising companies, right? And and working alongside them as as operating partners and bring you know leaders to to help them do things they couldn't necessarily do by themselves, right? so even in our twelve year journey now, we've we've grown with our clients and with technology. Right. So it's it's not like I have a a built product that I'm like I had to re-architect, right? if you think about it, a lot of things that we used to do manually, right? Like we used to build large complex models that estimate value, right? that but we've been able to use across a large number of clients. So those type of things, there's much more advanced technology we're applying to, including AI, right? to to make that easier and faster and and and more robust today. But yeah, I I'm not I'm not sure you know, I would do anything differently. I think yeah.

And and one one one other one other thing, I and you can I don't know how you want to intro integrate this into the conversation. One of the things I think is also really interesting, and in addition to our work in in advising and and and let's cut that. Okay. in addition to our work in working alongside clients to really solve complex problems that create a lot of value for them, right? With With advanced data, machine learning, AI. what we do is we actually are very hands-on in helping them build those solutions. And so more and more we're taking on opportunities where we're not just advising and helping them build solutions, but then looking for partnerships with those clients to bring that to broader market. Because when you build something for a particular client, this is valuable to them, but it could be valuable to the broader market. So we're also in the in you know, looking at opportunities where we could carve something out, you know, and and take it to broader market and and have participation with all of the people that contribute.

Yeah. Yeah, absolutely. So every industrial B2B services company has a large umbrella AI opportunities. But the goal is to be very deliberate and methodical in understanding what are the four or five things that create the most value for you. Right. Rigorously quantify them, understand how much value it is, understand what the feasibility of getting to it is. And do some small scale pilots that actually show what you can get to. Right. And then get to get to real results by by by implementing and scaling those up. So that's that's the biggest takeaway. Right. Don't you know, don't be overwhelmed because this only about you know the the challenge is finding the three or four things that are gonna make the biggest difference for you, rather than do everything else.

Sure. you know, I'm on LinkedIn, I'm easy to find. If you type my first name you'll more than likely you'll you'll find me. Also you can go to our website, thalamuslabs dot com, which has a lot of a lot of description of what we do and how we do things and what we believe and and and lots of examples and then you can you know you can easily reach out to us there.

Thank you, Nicholas. I really appreciate it.