From Insight to Execution: How AI Is Closing Manufacturing's Productivity Gap

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 leadership and management experience Most of that with Michelin. I'm joined today by my co-host, Mike Yost, who's chairman of the club. Mike, would you mind introducing yourself further, please?

Not at all, Mike. Thank you and welcome, everybody. As Mike said, my name is Mike Yost. I'm the volunteer chairman of the organization, one of its co-founders, and Mike's co-host on these regular podcasts. So I've been in industry myself for pushing 40 years now. first half of my career spent with the technology solution providers and the last 16 or so with industrial nonprofits. So really excited to have you here with us today, and we're also excited that we have the pleasure of speaking to John Devagen, who's the CEO of Leading2Lean. And the topic of today's conversation is going to be From Insights to Execution: How AI Closes manufacturing's productivity gap. So just by way of background, John leads Leading to Lean, or L2L, and its mission to eliminate blockers and drive productivity by moving manufacturers from collecting data to actually taking action. Under his leadership, L2L has become the heartbeat of the modern factory, helping global brands turn intelligence into billions and reclaim value by closing the gap between seeing a problem and solving it. For those of you that are joining us live on LinkedIn, please go ahead and if you want to be part of the conversation, go ahead and put your questions or comments or thoughts into the comment section. We will do our best to get them to John and include them in our conversation. So with that, I'd like to welcome you, John.

Thank you, Mike. Thank you, Mike. Appreciate you having me on, having me on this podcast.

Absolutely. Look forward to the conversation.

Yeah, thanks for taking time out of your schedule to do that. Um, I want to jump right in, and, you know, we, we've, we've talked about in the last decade the manufacturers have invested heavily in automation and connectivity, data, but we've seen graphs and, and information that says that productivity has remained really pretty slow over the last decade. Productivity improvements. So, so why do you think that so many organizations are struggling to turn these tools and what data can bring them into real improvement for their bottom-line results?

That's a, that's a really good question, Mike, and I think you said at the beginning of the question there of much of the last decade has been spent investing in connectivity and dashboards, reporting, You know, plants have done a really nice job of getting that visibility, that operational visibility of what's happened. They use sensors and data lakes. They've got sophisticated analytics. But those systems identify where the potential problems are. They show in the past that productivity or production or downtime, what the impact has been. What those technologies don't do is address the gap, the productivity gap. The productivity gap isn't about technology. It's more about these disconnected systems and how they relate to the data, the decisions, the work, what that frontline worker needs to actually action on, on a day-to-day basis. And as you both know, manufacturing is tough. It's fragmented. You have these disconnected systems in many ways, right? Spreadsheets, systems, whiteboards, and all of that. And it's hard to get that shared operational picture.

So it's interesting because we have all the visibility and you mentioned being fragmented, but I begin to think about, is it, do you think that we're just not getting to the root cause and taking the right actions, or are we not actioning enough on the shop floor to really move the needle? What are your thoughts?

I think we spend a lot of time collecting data, and we collect that data every day, every shift. We can collect it from machines, from lines, but where's the context in that data? The machine was down, we had 10 micro stoppages in the last, you know, 15, 20 minutes on a particular line. Why? And a lot of times, uh, that is what steals our productivity. We're not sure where to act. We're too busy chasing that data without context. You know, this is one of the ideas and theses behind connected manufacturing operations, which is where L2L focuses. You know, our objective is not to collect necessarily more data. It's how do we connect the signals from production to the workflows, the escalations, the continuous improvement that's really critical to drive that productivity improvement. The dashboard hanging in the plant telling you that the line is down isn't necessarily going to help. The manufacturing workers already know the line is down, right?

Yeah, I'm wondering, John, you know, you bring up a lot of great points, right? Collecting data and I'm wondering how you position yourself and your company, like, because you have people making dashboards, you have people creating data lakes, you have technology infrastructure, corporate level, plant level, but you're not necessarily coming in to serve them. You're stitching together things, like you said, about productivity and such. And so I'm wondering how, like, who do you talk to typically in an organization? And what's that? How much has the conversation evolved over the years from let's go build dashboards to let's drive productivity?

Well, I think at the highest level, operational leaders, COOs, if you will, are focused on profitability. How do they profitably meet demand? But when it makes it to the plant, the plant managers focus on OEE. As we all know. And a lot of times over the last 10 years, that initiative has been, how do we track OEE better? But in reality, what do we do when availability isn't where we think it should be? Or quality? Where are we losing time on the shop floor? Are we losing time because an operator didn't show up for work?

Right.

Or an operator's new and running the line on a slower speed? a mechanic doesn't know how to fix a particular machine and is waiting for the expert to free up to come help them fix that machine. There's a lot of nuances there that steal that productivity, that time. So what our solution does is it tracks those abnormalities, hopefully shortens up the time to action. But if we're capturing the result of the outcome, then we can leverage, you know, certainly technologies like AI, which we'll dig deeper into, to surface that action next time it occurs with better context faster.

Right. Well, and I guess that's a really good point on sort of the historical, like where we've come from, where we're going, and now we've got the AI revolution, or I don't know exactly where we are in all of that, but we certainly have AI that's sort of shifting the possibilities from just doing reporting to actually learning from previous reporting and making recommendations and orchestrating things that could come next. So how do you see, I guess, first of all, how do you see where AI is today in this process? And then how do you see that evolving and changing and informing manufacturing leaders and frontline teams as we move forward into the future?

Well, I think, you know, if we look at it today and we're honest with ourselves, a lot of companies are piloting and testing AI. Sure, they're not really sure how. They know they need to be there because they're reading the news, they're following the news. But, you know, the concept of like, let's go deploy AI, is really not helpful. It's confusing, it's fragmented, it causes unnecessary angst and actually probably steals more productivity. I think in 5 years from now or 4 years from now, and Gartner actually had a stat on this I read recently that said that by 2028, over 15% of day-to-day work decisions will be made by AI.

Wow.

up from effectively zero. So if we think about that 5 years from now, or in that case 2 years from now, it's how do we make those better decisions and what are we doing differently operationally in the plant to make those better decisions? And I think that is where successful AI pilots, if you will, are focused. What's the outcome that I'm trying to solve? Improve mean time to repair, as an example.

Yeah. So that is an interesting thought on this idea of having AI help us make decisions. And you mentioned you guys are starting to get into this thing. I think about the probabilistic approach of AI and then making it deterministic. How do you— what have you seen in terms of balancing that? Because I wouldn't want it to help me make a decision about something that's more— that I want it to be deterministic. So how do you kind of— how do you close that off in, in what you've seen it for a manufacturer?

When you say close it off, do you mean like—

Well, be deterministic. Yeah, how do you, how do you be deterministic versus probabilistic? Because that's a, you know, it's going to give you an answer if you're not— even if it doesn't know the answer.

100%, right? And so I think that establishing the guardrails in the context— remember, humans are still in control, it should be viewed as in control of AI, right? So the use cases that you define need to be impactful and meaningful enough, and we can talk more about what a meaningful use case is. But if I do have that, then I can define, well, what triggers us needing AI And what actions do we want it to take? In many cases, it's identifying the abnormality in manufacturing, identifying that an action needs to take place to fix it, and then who needs to fix it with what. And if I can do that faster, just the identification of that, then I will recover lost time and lost productivity or waste. Right. So let's think in terms of like kind of narrow swim lanes, if you will. Versus broad transformation, which can be paralyzing, as we saw with the digital transformation concept, right? That was painful for many companies, fruitful but painful.

So I think what I hear you saying, which makes a lot of sense, is be specific about the use cases where you want to use it and then go deep on understanding how you want to use it. In order to help you put those guardrails in place so it can actually help you improve, if I can restate what I think I heard you saying.

Yeah, Mike, and if I were to kind of boil it down to maybe 4 points for you, you know, if I was gonna use AI in my plant, I'd probably look at these 4 areas or ask these 4 questions. The 4 questions would be the first one, Is there a problem that meaningfully impacts my throughput, cost, quality? Pick something really meaningful. And is it holding me back? The second question I would ask is, does that decision I need to make every day on impacting that part of the business, does that improvement compound? Do I get better every time I fix or address it? Are the people— the third thing is, are the people or my workforce, are they waiting? Are they searching for resolution? Are they looking for that expert in the plant? Where can I deliver them more context and will they be able to address it faster? The fourth is where I think it alludes to what you're talking about, which is, is there enough history data And is there, is there enough context that I can produce that useful recommendation that the worker will accept? Well, I don't know the answer to that yet. What I do know is if I give that worker the opportunity to accept that recommendation and I learn over time, then I'll get more confident in using AI for other use cases. So once again, narrow. What is it going to do? A good kind of example might be, how do I detect and reduce downtime? And then in what, what context of downtime am I looking to reduce? And then apply AI to that using those 4 questions and just get better, compound and get better over time, build the confidence around that.

I just want to take a moment here to remind everybody that we're talking with John Davajan, the CEO of Leading to Lean, talking about AI and closing manufacturing productivity gaps. So if you have questions, please go ahead and jump in, join the conversation with us through the comments section. One of the things that I'm interested about, John, is, you know, when we talk about things making recommendations, right, it comes back to how much do you trust that, right? I've seen throughout my career big systems completely turned off or worked around because people didn't trust, you know, what the systems were telling them. And plant managers, like, if it's slowing you down, get it out of the way, you know, despite the fact that it came from corporate and cost 7 or 8 figures.

Yeah.

Do you think that we're learning? Because this approach you talk about, as far as us, you know, knowing that we can't really trust it implicitly, so we're being cautious. Right, and to back it up or, you know, prove its trustworthiness. Do you think we've learned as an industry that let's not just throw it away and turn it off, let's grow with this, you know, because of what's known, I guess, so far about AI?

I don't know if we're there yet. I think it's still early, Mike. I think I think there's a trust factor here.

For sure.

And I think there's anxiety. Some people might say, is there the— I guess the big words would be like psychological safety in using AI. Like, is it gonna take my job? There's all of that that's running rampant. It's not gonna take manufacturing's job. It's gonna change their jobs.

For sure.

Their leaders need to look at it by putting in the guardrails that you're alluding to. Without guardrails, it's hard to trust it, right? It's hard to trust the recommendations. You know, the workforce that's receiving those recommendations, they need to— they need the culture of knowing that they can say no to the recommendation.

Yes.

That's not— and the organization needs to applaud that, because the only way that it's going to learn and get better is when people do that. If not, I think you're just going to end up with a lot of failed AI experiments, and it's not going to make you better. You're not going to get all that expertise out of people who have been on your shop floor for 10, 15, 20 years.

Yeah, I guess that's where I was going with the notion that maybe we're learning because AI is so pervasive and we all know that there's a hesitancy to just trust it right out of the gate, that maybe that culture or that receptiveness to your point about pushing back, somebody who looks at a system, gets a recommendation, says, I'm not yet ready to do that, that we might be more mature just societally because we know that that's the way we need to behave. that, you know, maybe that's helping us get to a point where we can evolve and grow with these systems and give them the time that they need.

So it's interesting you bring that up, you know, Mike, because I think it was McKinsey did a study like about a year ago. And what they did is they polled a number of COOs, manufacturing COOs, And they posed a very similar question to what you're posing here, and what they got back was about 50% of those COOs said that like AI technology isn't necessarily holding them back. It's not technology in general. What's holding them back is the culture shift, the organizational culture shift, the trust.

Right.

And that many folks may think that, like, I got to prove AI wrong or else I'm going to lose my job. And that culture impact— and we already know that change management is tough in general.

For sure.

Yeah.

Now you're compounding that change management. And that's what McKinsey had found is like one of the biggest challenges, which I think you're scratching at.

Yeah. It's interesting that you— we speak about that. I was reflecting the other day about my own experiences with when we have data available. We were talking to the folks on the floor saying, what does the data tell us? What does the data tell us? Because in the past they would just tweak and they'd use the system, they would just based on feel and it'd be— it would always be, what does the data say? In fact, we would normally say, well, just don't tweak it. I want you to tell me that you did that based upon the data. almost like the data can't lie. Now with AI, the question is, what does AI say? But now it's a little bit more open to doing something that AI doesn't say to do because it's not quite as absolute as what I— at least I think of it this way— as what I would say about what does the data say. It's very interesting because it's almost a different— you almost have to be more willing to allow the— maybe then at least I was with respect to data, the frontline worker to say, no, I think we need to go this way. And that's a very different leadership and management opportunity for the frontline leaders, I think.

So one aspect of it is that, right, Mike, is like the machine runs a certain way. I capture that data every second, what's happening. I can then predict downtime, predict maintenance and all that. Like that's been around for a little bit. But one aspect I want to kind of spin on you here is it doesn't necessarily have to be all around the machine data. And like another use case for us to think about is schedules. I might have 20, 30, 40, 50 manufacturing plants around the world. I need to schedule the product orders as they come in. Maybe I need to— one of the inputs needs to be supply chain and logistics around, you know, where the raw materials are.

Yeah.

Another needs to be around how the machines are performing or where certain expertise relies across my plants and how do I balance that? You know, AI could speed that up and therefore make it a competitive advantage for you. Another one is I've got my plant, I show up every day. Yeah, sometimes people don't show up for work. Sometimes people go home sick. How do I reschedule for the next shift or for the second half of the shift that I'm in, in real time? Instead of gut feel, AI can help assist and give you options, ways to look at this so much faster, save you hours.

In fact, we have Brad listening in and he shares an article from Harvard Business Review about complexity. Your point is exactly right. Now things are much more complex, and this is interesting. It goes with what you're saying. Now I can have many more options and many more decisions because AI, and it makes it much easier to know how to move forward. Thanks, Brad, for that reference.

Well, I think also, John, I think you're hitting at a really important point. Early on in the conversation, you talked about the complexity of manufacturing, and there's so many different— it's always a fight against loss, right? So no matter what we're doing, we want to go make something, and that could come from your process or your machine, but it could also come from the things like you just articulated of supply chain issues or worker shortages or any number of different things. So I think it's really exciting time to think about how maybe not looking just in isolation at, you know, an area or a cell or a, you know, part of the plant, but up to higher levels and across departmental boundaries and, you know, different areas of the business where, you know, the plant needs to be able to go and deliver when it's told to do so. But what are the implications You know, on things that you're talking about from a scheduling perspective or disruption in materials availability or any number of things. So I think it's really, really an interesting point.

I mean, like, the job of the plant manager is really hard. It's been really hard. Yeah, I think it's getting harder because of the technology. Yeah, we could talk about the transformation of the workforce. and the aging workforce and the skill sets. Yeah, that compounds it, but it's also really hard. Margins are tighter. It's a, it's a really hard job.

Right.

So when you deploy technologies in a way that's going to steal their— further steal their time to figure out how do I deploy this technology, how do I use it, they're not going to use it. It's you need to give them something ready-made that's going to add value immediately, that's narrow enough that they can embrace. And I think that's where there's this IT/OT, there's always been the struggle there. That's really the kind of the— we're back to that again from 10, 15 years ago where it was, no, you're not putting a sensor on a machine, or no, you're not installing that software. We're kind of back there a little bit too.

Yeah, yeah, I agree with that. It's funny, I mentioned my nonprofit experience. I used to, back in 2008, I was on the MESA International Marketing Committee, and the team came up with the idea that the annual theme for the conference, or the theme for the annual conference 2008, should be convergence of OT/IT. And we didn't exactly call it that then, but they have this convergence idea and the board shot it down because they said this OT/IT convergence topic, it's overplayed. Like nobody wants to hear about that anymore, right? We've been dealing with it for years. And so now here we are in 2026 and I agree with you, we're back to a lot of the same issues.

So John, as we think about as manufacturing leaders, you talked about The plant managers, as you're thinking about their situation as you described it, what advice would you give an executive who's really maybe still at the beginning of their AI journey today? And what mindset or leadership principle do you think would have the greatest impact on closing this productivity gap that we've talked about and really building a more resilient and high-performing organization?

I, you know, I— well, there's a couple things that I would say to that. The first is leverage AI as a way to strengthen your organization's problem-solving capabilities. A lot of what we do on the shop floor and across manufacturing is problem-solving. And so I would say that first, but a quick follow to that is design it to the people that need to solve the problems, right? It needs to be consumable, delivered to them in context in the moment, right, that they can understand and act on. And then the third piece, I'd go back to what we were talking about earlier, which is the employees need to be empowered to know when they should not take the action or when they should take the action. And if they're struggling with making that decision, then probably it's too broad of a use case, right? You know, it should be pretty binary for that worker that's on the floor in front of the machine. They have significantly more context. The AI is delivering them faster, more comprehensive, the how-to, but they're not looking at the environmental factors that might be surrounding that. whether it be a machine fix or what might be occurring on the line. So earlier I made the comment psychological safety, I'd probably broaden that and say safety in general, right? If we, you know, the workforce still needs to be aware of policies and procedures, you know, AI is still learning.

Yeah. Yeah, that's for sure. I've heard a colleague of ours talk about how safety became ingrained as a pillar of manufacturing many years ago. And argument is that it's time for digital to become a similar pillar in manufacturing, right? So what does that mean for each of us in all of our different roles? And I think that's part of the evolution that you've been poking at.

Well, so we're kind of reaching the end of our discussion. We've been talking with John Duvagin for the last 30 minutes or so. And so as we kind of wrap up here, Mike, John, I'm interested in maybe a last thought that we could leave our audience with. So Mike, you want to start us off and then John, we'll go to you?

Yeah, sure. First of all, John, it's been an absolute pleasure talking to you and learning from you. I think a couple things that stood out to me is you made a comment about chasing technology can actually steal productivity. I think that's a hidden gem here for anybody listening. And I also think, you know, we talk Industry 4.0 and oftentimes people think that that's a technology discussion and there is technology involved, obviously, But, you know, John's focus on selling to a COO audience, right, to the operations audience, I think is really powerful. And so if you're an IT leader and you don't have a partnership with your ops team or vice versa, I think it's good to understand how Industry 4.0 is spanning across all those different functions. And John's a perfect example of somebody that engages with that operations audience. So appreciate everything that you're sharing with us, John.

Thank you, Mike. Thank you, Mike. I'd say this, the opportunity ahead is really transformative for manufacturing. If we connect the shop floor to the corner office, technologies like this don't come along very often. And it's really simple when you look at it at the plant floor. If we're able to recognize those problems, those abnormalities with the right expertise, and then focus on the actions, the priority actions that are gonna drive productivity, we actually shorten up that time. We recover that time. And if you're in the corner office, that translates to profitability.

Right.

Right? There's a clear connection there. And I think the next 4 or 5 years are gonna be really exciting in manufacturing. That's where we're focused. That's what we're excited about over here at L2L. So I really appreciate you having me on today.

Yeah, we appreciate it.

Well, thanks, John. So I want to just thank you. I know you've got a lot going on and really appreciate you taking time out of your schedule to kind of share your insights about, you know, in general about manufacturing, but AI can help improve productivity on the plant floor. So thank you for that. So the Industry 4.0 Club would like to encourage you to take action. Obviously, you've been listening to us. That's one way to learn something new. Read an article, participate in a workshop, even initiate a smart manufacturing roadmap for your facility. We'd love to hear from you and hear what you're doing to improve your— to improve and use Industry 4.0 to help you solve the problems you've got. Drop us a note on our LinkedIn or at our website at www.industry40club.com. Share with us what you're doing. It's important that we move forward in this so that we can advance and become more productive and more effective. We also would like to offer you the opportunity to join us on one of our future broadcasts. Again, if you'd like to join us and want to share something with the audience that perhaps is an area of focus for you, again, drop us a note on our LinkedIn page or our website. With that, again, John, thank you. Mike, thank you for your time. And we wish everybody a great rest of the day.

Take care, everybody.

Thank you.