From Data To Decisions: Unpacking Industrial AI

[00:00:00] Filip Popov: Hello and welcome to our new episode of Espresso 4.0. Today, we're joined by Rodney Waterman. Rodney, welcome.

[00:00:08] Rodney Waterman: Thank you, Filip.

[00:00:11] Filip Popov: Pleasure to have you in our episode. Thank you for joining me for a virtual coffee or tea or in my case today just the water in my janky bottle which I love it because it's been with me for a while and I don't want to throw it even though my wife tells me that I should. Why don't you start us off for this episode and tell us a bit about yourself, your background, current position, and what you see as your role in Industry 4.0?

[00:00:41] Rodney Waterman: Sure, I guess I'll start out with current position just to introduce. Um, I'm the director of industrial AI at Lityx. And I got into industrial AI mainly because of the growth in my career path. I started out in engineering and I have an industrial engineering degree. Worked with process control as a process engineer, and I absolutely love creating efficiencies within manufacturing and process industries. And it is what led me over the years to work with data, data analytics, which led to predictive and prescriptive analytics powered by machine learning, which we call industrial AI. So that's, that's who I am, where I came from and what I'm doing today.

[00:01:27] Filip Popov: Excellent. Thank you very much. Speaking of industrial AI, what's the biggest misconception companies have about AI and how do you address it?

[00:01:40] Rodney Waterman: I think the biggest misconception is that it's not real yet, or that it's just all hype. And that doesn't, it's not helped any by some of the things that we see promoted, you know, throughout social media and such, but to get over that hurdle is difficult at times even when a prospect can see it for themselves.

It's still, there's a difficulty sometimes in that hurdle of it. It can't be that good, right?

[00:02:15] Filip Popov: Yeah.

[00:02:15] Rodney Waterman: I had a conversation recently, well more than a conversation, a demonstration with a potential customer who gave me some data that they had already analyzed and asked if we could analyze it and see how long it takes us to get to the same answer they got to. We did it in 98 percent the time that they spent in analyzing that data with the same result, and he was a little hesitant about moving forward, and I had to ask him when is the right time to move forward with a technology that reduces the amount of time it takes you to analyze data by 98%. And it, sometimes it's just difficult for people to take that step into something that last week was just completely impossible in their mind.

[00:03:08] Filip Popov: Yeah, well, I mean, I think that's, that's why it's difficult. It was impossible in their mind just last week. And, um...

[00:03:13] Rodney Waterman: That's right.

[00:03:14] Filip Popov: I think that's, uh, that's the root of the problem. Um, having said that many companies struggle to see ROI from AI investment on the other hand, so what's your approach to demonstrating tangible results early in the project?

[00:03:28] Rodney Waterman: Well, it really all begins with the plan and what you're starting with. Too many times executives will hear about successes in the marketplace and send out a directive to incorporate AI with no real direction or how it's going to be used. Right? But when you start with a clear use case and a plan to accomplish a goal that is well aligned with the corporate goals, success is almost inevitable then. And so, it just makes so much more sense to start with a clear use case, how it's going to be used, um, who's involved, and move forward that way rather than just throwing something at it.

[00:04:15] Filip Popov: Yeah. Yeah. It's not about a cool technology. In fact, uh, what needs to be in, in kind of... what needs to be our end result should be a tangible and well defined business value.

[00:04:28] Rodney Waterman: Yes. Yes.

[00:04:29] Filip Popov: And you need to have obviously relevant data as it relates to the scope that you are trying to achieve as opposed to just a fancy tech.

Fair enough. Of course, I think that's a repetitive, um, it's a repetitive, uh, theme and it merits repetition because people seem to start to keep making the same problem. Can you share a specific success story from the industrial division that analytics, including key metrics, like energy savings, increased uptime or revenue growth?

[00:05:01] Rodney Waterman: Yes. Yes. Um, several we could talk about. I think probably one that would resonate the most with people is, a customer of ours who produces wax, they had a desire to increase their yield. It's a petroleum, um, raw material that is processed into wax and it's a pretty large operation. And so, they wanted to reduce their scrap, improve their yield and weren't sure exactly how far they could go.

But as most industrial operations are, some people have, I'll call it a gut feeling just because they know the process well, that it could be better, right?

[00:05:50] Filip Popov: Yeah.

[00:05:51] Rodney Waterman: Excuse me. And, and so we, they'd been working on it with little success, a little bit, but not a lot, they thought it could be more and within just a few weeks, we were able to give them direction on how to improve their yield. Now, our initial models said that we could improve their yield or reduce their scrap by 77%, which was huge, right? And everybody kind of looked at that and said that's not possible. It really, I mean, there's something missing in the model, that's not going to work, and we're not going to argue with that because obviously there are times when you leave things out that something might not have been there, but instead, we all took the positive direction of, well, we can do something here. And so within just a few weeks, about six weeks, we had reduced, uh, actually it's less than that, about four weeks. We had helped them to reduce their scrap by 35%. Then in another couple of months, reduce it a total of 50%. So, just within five or six months had reduced their scrap tremendously. And this was a pretty good size operation. So, it turned into over 10 million dollars a year for them in new product or good product where they didn't have to, uh, add raw materials or add new equipment of any type and built out this better optimized process because of the direction we gave them.

Now, I will add this was about 4 years ago and about 2 weeks ago, we received information from them, an email of a screenshot of a chart that they run showing that over a weekend, they actually performed at 70 percent reduction of scrap. So, while that initial, um, effort, um, didn't produce that 77%, it gave them the information and the belief that they could achieve that 77%. And now today it's becoming something that may be a regular occurrence. So it's a, it's a fascinating story, um, on all sides.

[00:08:05] Filip Popov: Fascinating indeed. I actually, I want to dissect a little bit if you, if you don't mind. So, this is running a, um, recommendation set for recommendation, uh, or is it running in closed loop?

[00:08:17] Rodney Waterman: It is not closed loop. So, our offering is more of a decision support, um, offering. So, the data is collected, analyzed, information provided back, specific actionable directions on how to achieve what you're trying to accomplish. So, it is not closed loop, at least not today. There are some companies out there that offer that.

We do not.

[00:08:41] Filip Popov: Yeah. Yeah. And nonetheless, this is a, you've moved the company and basically a cultural change in terms of the way they operate from experience based and hard coded based possibly, um, a way of, to manage their production to a data driven production management to which, as you said, it yielded some pretty, even if you told me, honestly, 70%, I would say I would call bold, but, uh, but it's very, but it's, uh, it's very good to hear and it's, and it feels really good.

It feels warm inside when a client sends you a screenshot saying, we've actually reached your initial.

[00:09:18] Rodney Waterman: It is, yeah.

[00:09:19] Filip Popov: Your initial estimation.

[00:09:21] Rodney Waterman: Right. That customer has been so thrilled with the outcome that they have spoken with us at conferences and, um, offered, um, various support in other ways to help other customers understand what's real and what can really be done.

Excellent.

[00:09:39] Filip Popov: As far as scaling and the future of AI in industry is concerned, so scaling AI from a successful pilot to full production is a common challenge, right? A lot of companies end up with that pilot purgatory or a lot of projects that are right up there. What are the key factors that determine whether an AI project will scale successfully?

[00:10:01] Rodney Waterman: I think there are a couple of things that that are key, um, one is we talked about it a little bit earlier, when we first help identify a project a use case and making sure that not only does that fit with the corporate goals, but that everyone in the, I'll say the line of stakeholders, is aligned with that need that is part of the goals and what we're trying to accomplish. When you have that, you have a cooperation and, um, an interest and buy in from so many different people that, uh, it drastically improves your chance of success and then scaling after that first success of a pilot, um, scaling is much easier because you have, again, the support across the organization.

So, I'm going to first say the people in the alignment. Um, you mentioned the word culture a moment ago and in culture, there's often a culture shift that has to occur because, in that example, I gave a moment ago that IGI wax, like most operations, you have operators that are used to adjusting the dials, if you will, and tweaking the operation. Um, second shift comes on, tweaks it the way they want to run it. Third shift comes on, tweaks it the way they want to run it. And so it's not consistent. Um, but by believing in the results or the direction that AI provides, and everyone surrounding that and supporting it, you have a culture shift of not just an operator being completely relied upon, but the directions that are given and all of us working together to make sure that those work with the system that's in place. Long answer, but alignment of those people in the culture. The second thing I'd mention is making sure that the right tools are in place that you've got tools that are in place that are easy to use, um, accessible to everyone and, uh, are there ready to, uh, to move, help things move along quickly. So, that's important as well.

[00:12:24] Filip Popov: Yeah, fair enough. Actually, yeah, the answer is quite, quite layered, right? Starting from the culture to architecture to the right tools and on, on, on and all things need to be in harmony in order for that to actually work.

[00:12:37] Rodney Waterman: Yeah. Which is another reason why you want to start with something small, a pilot.

[00:12:42] Filip Popov: Correct. Yeah.

[00:12:43] Rodney Waterman: You don't want to typically, you know, try to, um, throw AI across the whole enterprise and say, hey, let's just do AI tomorrow, that's not going to work.

[00:12:55] Filip Popov: There's a dichotomy there because you need to think big. You need to think solving for future problems, but you got to start small. So, I guess the saying or model there is think big, start small, and then reiteratively kind of build on top of that. Absolutely.

[00:13:11] Rodney Waterman: Yeah. Yeah.

[00:13:13] Filip Popov: How do you see the role of AI evolving in industrial decision making? Will it remain a support tool, or do you see a shift towards more autonomous systems? That's kind of going off my earlier question on whether it's running in.

[00:13:26] Rodney Waterman: Yeah. Um, well, I'm not sure that we'll ever get, you know, all the way into autonomous systems that are... I guess what I mean is as we grow and AI takes on more and becomes an enabler for autonomous systems, there will be other decision support realms that we're not even thinking about today. So, um, so I think that, yes, there will be more and more applications of creating autonomous systems as we grow, but there will be other more and more applications for decision support. So, I'm not sure that it will go one way or the other. I think both will always be around, we'll just see both of them grow and in areas that we're not even expecting right now. Same concept as we had earlier about our belief and what we're not even expecting next week.

[00:14:32] Filip Popov: Yeah.

[00:14:32] Rodney Waterman: You know, anyone who's keeping up with dates of this particular recording, last week, this DeepSeek release that hit, no one was expecting that, right? That's huge. And those kinds of things are coming more and more, uh, faster and faster. And so, there's no telling what will be happening this time next year.

[00:14:55] Filip Popov: Yeah, faster and faster indeed. Yesterday Alibaba announced they have another model that's even stronger and more powerful than DeepSeek. Um, so, it's an arms race at this point.

[00:15:08] Rodney Waterman: It is. The same thing as yesterday, yes.

[00:15:15] Filip Popov: Oh, okay. And, um, last but not least, um, the adoption of AI in industrial setting is growing, um, I think you'd agree, and you just said so as well. What do you see as the next frontier or untapped potential for AI in this space?

[00:15:36] Rodney Waterman: Yeah, I guess that last discussion is really tightly related to this because from the plant floor perspective, more and more autonomous systems will grow out of this and what's coming, right? From a decision support perspective, some things that aren't being considered today at a business level will be.

For instance, if a corporation is trying to decide if they want to add robotics to their product mix. How do you decide that? Well, today you have people who are researching what's out there now. What's the market look like? What is the potential sale value? What are the costs? What are our potential margins? Um, you know, what's the supply chain look like? All of those things. But one day decision support, as we know it may play a big part of that, where you might ask your AI, let's just simplify it and say, AI, it be advantageous to our company to add robotics to our product line? And then in a few seconds, AI spits out a hundred page report that gives all the advantages and disadvantages with a recommendation at the end, and even how you might need to change your business model in order to be support, uh, order to be, uh, successful in that. Right? So, again, decision support will never go away. And there are ways that we will use it in the future that we're not even thinking are possible today.

Yeah. Yes. Yes. So sort of, uh, I guess that would have been an example of some sort of an adjacent AI that's trained on the real life simulation, real life, uh, real world. Um, indeed. Okay. Interesting. Thank you for sharing that. That kind of concludes my questions regarding, um, regarding industry and, you know, your know how and typically what Espresso 4.0 is about. Thank you very much. Um, typically I end these with more personal questions for the audience to get to know Rodney Waterman a little bit better. So, to start with, what's the coolest thing in tech you're nerding out about right now? Um, not necessarily in Industry 4.0.

 Well, uh, I would say that, um, the dynamics between releases like DeepSeek and um, quantum computing, you know, because quantum computing requires a lot of power because of the cooling systems, but DeepSeek is showing us that it doesn't require near as much power as, um, as all of the, you know, systems that we have seen up until, up until then. And so this, there was this, there's been this huge surge towards supporting that power infrastructure that's going to be required for the systems of tomorrow, but then all of a sudden we're seeing that maybe there's a path where there's not so much power required. And so, it's been interesting to look into those things and see how those play out. Um, because that will have impacts around the world.

[00:19:09] Filip Popov: Yeah, it would be a punch in the gut if in about a month China comes out with a quantum computer stronger than Google, faster with less, with less, with a lot more efficient, with less energy requirements and all this and that, that would be. That would be something.

[00:19:25] Rodney Waterman: It would. It sure would. You're right. We'll see if you were the, uh, the prediction, you provided the prediction that occurred.

[00:19:37] Filip Popov: I'll keep an eye for the newspapers for sure. And last but not least, if I called you on Sunday at 3PM, what would I be interrupting you in?

[00:19:50] Rodney Waterman: Probably hanging out with family.

[00:19:53] Filip Popov: Hang out with family. Nice. Like at home, or do you guys have a hangout spot?

[00:20:01] Rodney Waterman: Probably at home. Um, yeah, it's probably home.

[00:20:06] Filip Popov: Nice. Recharging.

[00:20:07] Rodney Waterman: Not really exciting there.

[00:20:11] Filip Popov: Well, I would disagree. I don't know, exciting, perhaps not. But certainly, certainly not as much as quantum computers, not as, exactly, certainly not as much as quantum computing, but, uh, but certainly enjoyable and a great opportunity to recharge.

[00:20:25] Rodney Waterman: What about you? What do you do on Sunday afternoon?

[00:20:29] Filip Popov: Well, listen, as someone who lives far away from his family, I'm very, uh, I'm very envious of your Sundays, 3PM. So, in fact, what you would find me doing at 3PM is calling my family on a Sunday.

[00:20:44] Rodney Waterman: There you go. That's good.

[00:20:44] Filip Popov: I'm not much different than you. Excellent. Thank you very much. Rodney. Uh, it was a pleasure talking to you. And, um, I hope that sometime in the future we'll get a chance to grab a coffee again. Maybe once the new quantum computer is released, maybe before.Um, but, uh, nonetheless, um, take care and good luck.

[00:21:05] Rodney Waterman: Thank you, Filip. You too.