How to Start, Scale & Survey the Success of AI & ML Initiatives to Take Advantage of the Hype!

Welcome to the Industry 4.0 Club's fireside chat. The club's purpose is to engage the power of diverse worldwide talent to accelerate the global evolution to Industry 4.0. We accomplish this by hosting and supporting respectful, thoughtful, and interdisciplinary open conversations and creating quality content on wide-ranging Industry 4.0 topics to expand the reach of our community. My name is Mike Ungar, and I am the CEO of the Industry 4.0 Club. I have over 40 years of experience in leadership, with most of that being in manufacturing at Michelin. Hosting today with me is Mike Yost. Mike is the chairman of the club. Mike, would you care to introduce yourself, please?

Absolutely, appreciate it, Mike. As Mike said, my name is Mike Yost, the volunteer chairman of the Industry 4.0 Club. Welcome you all here today. I've been in industry for about the last 35 years. on the industrial solution provider side at the beginning, and the last 15 years spent on the industrial nonprofit side. So looking forward to our conversations here today.

Thanks, Mike. So we're joined today by Eric Ellwanger. He's the engineering leader for Commscope. Eric will be at the 4th MX.0 Southeast event in Greenville, South Carolina, which will be taking place on April 22nd and 23rd. He can be part of a panel on the 23rd discussing how to start, scale, and survey the success of AI and ML initiatives to take advantage of the hype. That's an interesting title. The MXO Southeast event has become the go-to gathering for manufacturers in the region to connect, collaborate, and learn from real-world examples of successful digital transformation shared straight from the horse's, or shall we say, manufacturer's mouth. Since 2022, the event has empowered local manufacturers to advance their careers and businesses through valuable knowledge sharing and exchange and impactful network opportunities.

And we're glad to have Eric with us here today. He's a data-driven professional with over 20 years of expertise in automation controls and embedded systems. He is skilled at applying machine learning techniques and advanced analytics and statistical methods such as progression, data mining, and predictive analytics on large datasets to reflect and solve real-life problems. Eric has experience with data science languages and deep learning frameworks also. So we welcome Eric here today. And before we get into the conversation, just want to remind everybody that we are live. And so if you have questions for Eric or want to join the conversation, Please feel free to add your comments into the comment box.

Thanks, Mike, and welcome, Eric. We're glad that you could join us.

Yeah, thanks for having me.

So just to get us started, you know, we were talking offline a minute ago, and it's interesting because you've been at Commscope for your entire career thus far, and that's a bit unusual in today's environment. So what's kept you at one company all these years, and how has work changed over that time? From your perspective?

Yeah, you know what, what's kept me here, and we talked a little bit about it, but really, you know, I've, I've had the opportunity to grow and learn with CommScope, and so I've never really had the desire to, to go out and look for that somewhere else. So that's kind of why I've been here. And, you know, as, as we talk about how I've grown and learned in my job. You know, when I first started with CommScope, I did a very short stint in IT where, you know, I learned basically how server administration and network architecture and, and those kinds of things. And that's changed quite a bit in the 20-plus years that I've, I've been here. But shortly after that, we got into a growth period and I kind of got moved into industrial controls and automation, which I'm so glad that I did. I loved it. When I got there though, you know, we were in a growth phase, so we were putting in a lot of equipment, drinking from a fire hose kind of thing. But I really learned a lot about what the equipment is capable of doing. And so that really is something that I continue to take with me in different roles that I do. And then I've moved into process engineering where I learned about what the What are the key process variables that go into making the products that we make and how to control those and improve those. And then I got into product development where I learned what goes into making a product, the materials, the mechanical design of it, and using my other experience, how to bring that into a, you know, design for manufacturability kind of thing. And then I, you know, I got the opportunity to shadow operations people, so understanding the things that are going on on the floor, scheduling, planning, and like I say, all the financials that go into it.

Yes.

Now my current role is leading smart manufacturing initiatives, so being able to take everything that I've learned and being able to develop initiatives that are going to improve the cost, quality, and actual throughput of our products. So that's kind of how things have changed for me over the years.

And I think it's really valuable input right off the start for folks that, you know, as a company, they give you room to grow. And then from a smart manufacturing perspective, I was jotting down the different disciplines you've been a part of. And so having that cross-functional understanding of, you know, IT, automation, process, product, operations, That's really important when you're getting into the conversations about smart manufacturing.

Absolutely.

Excellent. Well, we know you're going to be at the MX.O Southeast event here coming up quickly, and your title is— you're talking about starting, scaling, and surveying success of AI and ML initiatives to take advantage of the hype. What do you hope folks will get out of your event? What do you— how can you plant a hook here to get folks to come and listen to you?

Yeah, you know, I guess there's 2 main points. Well, maybe 3 main points. One is you can do this, right? It's not rocket science, and it's something you can do. The other thing that I'll say, and it's a borrowed phrase, I think I got it from MXO maybe 2 years ago, but it's and even last year, but it's start small but think big. It's about taking an idea for something you might be able to do AI with, and you've got to come up with that through data-driven decisions. If you're already doing things with Lean or Six Sigma, those are how you're going to get these ideas on where AI might be applied. But that's— so basically taking a line or a cell or something where you can apply AI, but then also thinking about, okay, how can I scale that to another line or another plant? And so being able to scale that. And then the third thing is, is that I hope people take away is AI is only as good as the data that you put into it.

Yeah.

So you really got to come up with a data strategy As well on that. So when you're thinking about how you're going to deploy AI, thinking about how you're going to get the data, making sure that it's clean, it's reliable, even contextualized, you know, are you dealing with Fahrenheit or Celsius or PSI or our bars or things like that? So I guess those are the kind of the 3 things that I would hope that they take away. One is you can do it. 2 is Start small but think big, and 3 is data strategy.

That's excellent. And I heard you kind of talked about this for a minute, but I want to dive a bit deeper. You know, there's a— AI is in almost every article every other day, it seems like, where someone's talking about AI. And if I'm a smaller or mid-sized manufacturer, It's a bit daunting to figure out where to go to maybe to see if it would apply in my factory, let's say. What advice do you have for a plant leader or manufacturing leader who's trying to figure that out? How would they go about maybe thinking about applying AI?

AI can be applied in a lot of different areas. You hear a lot about predictive maintenance, quality control, trying to minimize defects, maybe even energy management, and then also in automating the processes that you have. So there's a lot of ways that you can apply it. Now, how an individual company may want to apply that, that's where they've got to take kind of a data-driven approach on where are their pain points, where are the key issues, what are the key issues that they're having and go after the ones that are going to have the biggest impact. It's hard to tell them exactly where to go to look, but I guess that's the way that I would take it. If you've got lean initiatives or Six Sigma projects, AI can be used in conjunction with that. You're going to come up with the ideas and then You can figure out where, where AI can be applied. And if you don't have the knowledge or the know-how on how to implement or put into place these AI models, you know, there's a lot of low or no-code things with all of the big-name platforms— AWS, Azure, GCP. So there's, there's ways to get in there. into AI not knowing a lot of coding. But if you get a little bit more involved, you probably want to bring in a vendor that might have some expertise in that, or maybe a contractor or something like that. But it's all something that can be done fairly easily.

You bring up— I certainly resonate with what you just said, or I'm on the same page when you talk about pain points and impact and connection with continuous improvement, lean, those sorts of things. that is certainly a lesson learned that I think is very valuable for people to take away. Do you get the chance to talk— we get to talk with a lot of people, a lot of different spaces, right? So we hear people and their concerns. We hear their concerns about AI replacing people, taking jobs, those sorts of things. What's your insight on that or your feeling as to whether the— the bots are going to take over or they're going to be more helpful in supporting our workers than replacing them?

Yeah, that's a really good question because it is a question that comes up on the shop floor. Because if you listen to the news or if you listen to any of these AI leaders talk, right, artificial general intelligence is right around the corner and 80% of us are going to be out of jobs. And the reality is a lot different than that. You know, the bots and to be effective on the plant floor, they've got to have access to a lot of data, and it's got to be clean and contextualized and all that. And there's, you know, we're decades away from probably seeing that on the plant floor, but it is a concern. And, you know, you've got to acknowledge that. But what I try to do is tell people what AI is, right? And AI You know, you're hearing a lot about large language models now, the ChatGPTs and—

Yes.

I guess DeepSeek recently.

Right.

That's just a small segment of what AI is. You know, when you look at the bubble of what AI is, AI can be rules-based models where if X happens, do Y. You know, then you've got your kind of statistical machine learning models like linear regression or logistic regression, decision trees, random forests, SVMs, those kinds of things that are based on statistics. And then you have— you go down a little bit deeper and you've got these neural networks or these deep neural networks that are a little bit more difficult to maybe interpret what inputs cause the output, but they, you know, are very good at learning patterns in data. And, and so when you talk about that and then you get down, okay, even the, the ChatGPTs and the Geminis, those are even a smaller segment, and they definitely have a place on the plant floor. I could see where they're going to be good for training or troubleshooting or or something like that. But, you know, in terms of doing someone's job, I don't think that's going to happen. The other thing I like to do when somebody asks about that is I highlight something where we've used AI. And, you know, I can take them over to this line and say, hey, here's something where we implemented this and we were able to, you know, increase the throughput of this line, but we didn't get rid of anybody, right? They're still here.

Right.

We're just getting more output with the same person. So, you know, certainly if you listen to the news, it's something that you want to answer that, but you want to assuage their fears by giving them some factual information that's really happening out on the plant floor.

I want to— not to keep the mic, Mike, but you started off— Mike started the conversation talking about Your career at this company. You've been there a long time. I'm wondering if people watching this are thinking you're a unicorn, right? And that because you know the business, because you've built a reputation within the company, they trust you, that you got the smarts to understand, you know, these new technologies that like, all right, we have an answer because Eric knows what to do. And so I'm wondering if you think that that's commonplace, that folks have in-house talent that are looking to understand the technology versus they're looking for outside resources to come and help them.

Yeah, so, you know, I've talked to a lot of people in different companies, and there's a lot of people that, you know, have taken initiative to learn about AI, machine learning in particular, And so I think a lot of companies probably have that talent. They may be hidden talent, but I think a lot of people have that skill set inside. They just need to figure out who it is and, and go there. Um, you know, some of them are probably, you know, itching at the bit to do these kinds of things. And it's not, you know, it's not difficult to, to learn in, in today's environment. You know, I, I took— I've taken a lot of courses on Coursera. Now I've, I've went on and got my master's degree that specialized in data science. But, you know, I used a lot of Coursera courses to learn about this and continue to because it is changing at a very rapid pace, especially in the large language model realm, right? And so, you know, that's kind of new. So maybe you don't have that particular skill set, but I'm thinking people are learning that as well. So I think that skill set is in a lot of companies, but for ones that are very lean, that, you know, we've only got a process engineer, we've got— and he's wearing a product engineering hat and the process engineering hat and the automation hat, maybe that person doesn't have— they've certainly got the knowledge about the processes, but about putting a machine learning model in place, maybe they don't have that. But I think most companies have that talent within their ranks.

I bet that's refreshing to hear.

Good. Yeah, that's a great comment. I know from my experience when we started down that road when I worked at Michelin, rather than in the organization tell people, what the structure should look like, we went out and said, who in the plant, just like you alluded to, who in the plant is interested in Industry 4.0 or smart manufacturing? And, you know, we had some folks were industrial engineers, some were in production, some were in line production, some were in quality. I mean, engineering, and the volunteers came forward and, you know, they, you know, they either did it part-time or we created a role and they moved to new roles. So, I think your point's exactly right. When you ask, that interest level in a company can— you'll get the folks that are willing to be leading in that area and have skills and interest for that.

Right. I mean, there's a lot of people out there that are interested in, you know, Industry 4.0, digitalizing information. I mean, even in my company, we still have places where people are jotting down a test result for something that we don't have a computer at, right? And so when you're trying to figure out, okay, did this process parameter have an effect on that, you don't have that data, right? And so it's— but you've got people that are interested in doing that. And so I think we can drive that with the people that are already in place.

That's really neat. Can— to kind of I know part of this conversation around the, the, your presentations, the word hype is in there. Okay. And so I, I found that kind of interesting because, um, what is the hype that manufacturers maybe should not pay too much attention to today?

Yeah, you know, that's a— and we kind of alluded to it a little bit with the bots taking over and things like that. So this fully autonomous factory, right? That's— it sounds great, and it's, you know, maybe one day we'll be there, but, you know, that's really not going to happen, and you could spend a lot of money trying to get there and fail. So I think that's one hype, is just this fully automated factory is not a reality right now. The The technology is not mature enough for that right now. I think the other thing, when you start talking to vendors, you'll get vendors that come in and they promise you the world, right? We're going to increase your OEE by 15%. And they haven't even asked me, well, what is your OEE right now? What data do you have that can offer that. So, you know, sometimes you got to be wary of when vendors come in and are kind of overpromising what they can do. And the other thing I, I think is that this immediacy of return on investment— I alluded to it a little bit before, but AI is only as good as the data that you put into it. And number one, you got to have the data. And a lot of times the data that you had for maybe BI isn't the same data that you need for AI. And so there's going to have to be this investment in data, and it doesn't have to be huge, but you're going to have that. And so it kind of lengthens the amount of time that it's going to take to implement a project. So I guess that's the other thing is just these— if you hear how AI is going to give you immediate return on investment.

Yeah.

is probably— it's, you know, going to take a little bit longer to get that investment return because of the underlying data that you'll have to collect.

I just want to again remind folks, if you have any questions or comments, please go ahead and put them into the comment field here for us. I just wanted to sort of follow up on that point about working with the vendors because Yeah, I believe that you still— vendors may have toolsets that are proven to be capable and powerful, but they don't know your processes, right? And so, you know, I've seen and heard about instances where there's an initial return found or problem uncovered, but then once that initial problem's solved, I've talked to companies where the vendor's on one side saying, all right, what's next? And the manufacturer's on the next side, on the other side saying, all right, what's next? And there really has to be a blend, you know, on the process side as well, not just the tool itself.

Yeah, you've really got to put together kind of a statement of work of what you want with this vendor going into it. And like you say, okay, after we get this and it shows it's good, okay, what's the next thing that we want to do? And can you help us with that, or will you be able to help us figure out what the next thing is that we can do? You know, for us, we're a big enough company where we're doing continuous improvement things all the time, so we, we kind of identify where our pain points are and where we want to go after. where we think AI can help us. And so we've kind of done that work, but there are small and medium-sized companies that may not have those resources to do that, and so they rely on vendors to help them with that. And so, you know, having a close relationship with a vendor that has done this type of work can help you.

Well, we've been speaking with Eric Ellinger for the last 20 minutes or so, He's going to be at the MXO Southeast event in Greenville on April 23rd. The event runs from 22nd through the 24th. As we get ready to wrap up our conversation, Eric, I don't— we have one question from— that maybe I'd like to put up there as kind of a final one and see if you can address it. So Timothy asks, when you discuss AI in manufacturing, are many companies using AI that actually learns and makes decisions that were not programmed? Or is this a case of using more sophisticated software in combination with big data?

I guess when you're talking about AI that learns and makes decisions, I call that reinforcement learning. That takes a little bit more programming. But it is capable of doing that. When I talk about AI in manufacturing, I haven't— we haven't gotten to that point yet. Now, we do have AI that is controlling the process, but it's, you know, it's doing that based on a model that it did learn, but it's not continually learning the process. So I don't know if that answers your question or— Yeah, it does. But yeah, so I mean, we use AI to control our processes through a process. You know, when you train AI, it's basically learning a model. But there's also something called reinforcement learning that kind of continually is looking at your processes and then adjusting. And so if things change, it can learn that. Now, it's not a quick process, it's a slow process. We haven't—

Yeah.

We haven't experimented with that.

Yeah, it's pretty sophisticated. Well, do any— just so we can wrap up, Mike, any final thoughts for our audience today?

I would—

I'm always looking for people I can learn from, and so I would encourage people to get to the event and to meet Eric, pick his brain, because I think this is a great example of the body of knowledge that exists out there in our ecosystem, and I think it's really, really valuable. Stuff very well grounded. I'm still not convinced Eric's not a unicorn, but that, that's okay. We can figure that out or argue about that later. But certainly appreciate all of the insights you have. And yeah, just encourage folks to get to the event.

Eric, how about yourself? Any final thoughts before we finish up today?

Yeah, I guess I'll wrap up with a couple. One is AI is just a tool. It's not going to replace anything. If you're already doing lean initiatives and Six Sigma projects, AI doesn't replace that. It's just used to help with those. The other thing is, is AI models and data can drift. You've always got to have— if you're implementing an AI pipeline, you want to make sure that you're looking at that data distribution. You want to make sure that the predictions are matching with what you think they should be. And then the last thing I'll say is AI models aren't 100%. So, you know, even in— they can be, but, you know, sometimes they may have 99% accuracy, right? So there's going to be some false positives and false negatives, things like that. So those are my final thoughts.

Thanks, Eric. I really appreciate that. Yeah, that's excellent. Well, so I want to thank you, um, for joining us to give us a preview of your participation at MXO Southeast And you'll be talking on April 23rd in Greenville, South Carolina. And of course, we hope you decide to join Eric and the Industry 4.0 Club at the event. Here's the website address. Please go there. Please register. If you're a manufacturer, this is an outstanding opportunity to learn and network. So again, encourage you to join us in Greenville. Just to conclude, the Industry 4.0 Club would like to encourage you to take some action. Whether you attend the event is one way, read an article, develop a roadmap for your plant, or encourage your leadership team to develop a roadmap. Regardless, we would like you to try to move forward and integrate Industry 4.0 into your factory. We'd also like to learn about what you're deciding to do, so please drop us a note on our LinkedIn page or our website. We'd love to hear from you. The Industry 4.0 Club's website is www.industry4.0club.com. We host conversations like this routinely on LinkedIn and YouTube. If you'd like to be a part of a conversation in the future, please also let us know on our website or on our LinkedIn page. With that, I want to thank you guys for your time and wish everybody a good day and a good rest of the week. Excellent.

Thank you, Mike. Thank you, Eric.

Thanks, Eric.

Thanks, Mike.