Manufacturers have spent decades refining continuous improvement and the next evolution isn’t about replacing those proven methods with AI. It’s about making them dramatically more effective. In this sponsored episode of Great Question: A Manufacturing Podcast, IndustryWeek senior editor for technology Dennis Scimeca talks with Fabriq co-founder Octave Lapeyronie about the emergence of Agentic Continuous Improvement: putting AI inside the daily management rituals manufacturers already rely on, while keeping frontline teams firmly in control.

Below is an exceprt from the podcast:

Hi everyone, and welcome to today's episode of Great Question: A Manufacturing Podcast, sponsored by Fabriq. My name is Dennis Scimeca, senior editor for technology at IndustryWeek. Today we'll discuss how AI strengthens the entire continuous improvement loop and draw on real manufacturing workflows to explain why the opportunity isn't autonomous factories. It's real people doing real work supported in a real way. AI reduces the administrative burden, services relevant knowledge faster, and helps teams make better decisions. The result is a shift from episodic, locally driven improvement toward a more connected, continuously learning improvement system.

Joining me for this conversation is Octave Lapeyronie, co-founder of Fabriq. An engineer by training, Octave spent years working alongside manufacturing and continuous improvement leaders to understand how technology can help frontline teams solve problems and improve performance every day. He is passionate about building technology that delivers real value on the shop floor and exploring how AI and agentic systems can usher in a new era of continuous improvement. Welcome, Octave. Thanks for joining us today.

Octave Lapeyronie: Hi, Dennis. Thanks for having me. It's a pleasure to be here.

IW: Octave, manufacturers have been practicing continuous improvement or CI for decades. What has changed with AI that makes you believe we're now entering a new era of CI?

OL: Let's start with the world as it is, the world of CI. The status quo is actually, you go to a shop floor, you're still going to see a ton of pen and paper, whiteboard, people sharing information by spoken voice. That's the reality of it. So before we even think about what AI can bring to the table, it's important to realize that digitization of shop floor rituals is still largely underway. And while that's great, that brings a lot of value to the table because it increases visibility, transparency, and shorter reaction times. It also allows organizations to collaborate in a more complex manner, looping in support functions with the shop floor.

It's also still limited by the amount of information that a human can process, right? And information and opportunities to create value are often still left on the table, even for those manufacturers that have gone through a digitization phase of their CI rituals. What's changing now is, of course, with AI and LLMs, is the ability to do a lot of this brunt work of processing information, finding a needle in a haystack. Everything we know that LLMs are very capable of doing is creating this vast opportunity of value for manufacturers because they're seeing how it can help their CI systems deliver a lot more of the value that it initially promised.

It's also important to realize, and I see this across manufacturing, is there are ticking clocks adding a pressure context to the current environment – the retiring workforce. A lot of the frontline workers in manufacturing today are nearing retirement if they haven't retired already. And the younger workers that they're being replaced with just don't stick around as long. And that's fine, but that also means you need to onboard these new generations very fast and transmit that knowledge. And obviously, the geopolitical context, the economic context in general means that the cost of running a plant inefficiently keeps rising. Energy costs, supply volatility, and all sorts of factors that different manufacturing sectors see across the board that make this cost higher and higher, and make this opportunity to take these CI systems further just that much more attractive.

IW: I want to talk a little bit about agentic CI, which is actually interesting to me because I get into conversations with subject matter experts arguing about what agentic AI means. So we talk about agentic CI. Looking forward to hearing you tell me how is that different from simply adding an AI chatbot or another analytics tool? What does agentic CI look like on the shop floor?

OL: Yeah, I mean, first and foremost, what we mean by agentic is giving an LLM a harness, in a sense, and that's what you'd call it technically, a set of skills, competencies, and goals that it can go after, right? So giving it the ability to perform actions and not just answer questions like a chatbot would, and perhaps do that in a replicable way and do that recurring over time. What agentic CI means and why it's not just a chatbot, it's actually injecting CI where the work already happens, right? What's important about a CI system, it needs to bring actionability to the front line. How are you going to actually write back to your system when a new problem is solved, something is learned? You need to be able to be where the action happens?

The problem with chatbots, first of all, they tend to be expensive. They sit on the side. They require someone to think, hey, why don't I ask a chatbot about this and perhaps pull in information from system one into the chatbot, maybe get an accurate answer, maybe not, and bring that information all the way back to an existing system. It's just not practical. And really what agentic CI means is taking the same system, the same principles, but making it run a lot more smoothly and a lot more efficiently, surfacing the right level of information at the right moment, empowering the frontline to actually make much better decisions. And a good example of that is at the moment of a problem, right? There's a ton of data out there that sit in manufacturers' databases. If they have digitized and not erased a lot of that information off a whiteboard is how do you find that needle in a haystack when inevitably that problem is going to occur again, same plant, same line, or actually maybe halfway across the world, how do you make it so that you're facing this problem and you're fed the right context and the right information? Not because you thought of, oh, maybe I should go ask a chatbot separately and give it the right context. It just surfaces the information at the right moment in the right place.

IW: Octave, can you walk us through the continuous improvement loop and tell us where AI can have the greatest impact in helping teams detect problems, act on them, and then learn from what worked?

OL: I like to think of the CI loop as a four-step process: detect, act, learn, and ensure.

If you start with detect, the concept behind it is what do you focus on? What problems do your teams need to solve? And the answer to that is usually a function of a lot of different factors, but a lot of it is like, what happened on your production line in the past 24 hours, in the past seven days. What happened on the night shift? Maybe you weren't there at that time, right? So the value that AI brings here, it surfaces what the naked eye perhaps cannot see. And this becomes even more true for if you're someone who's in charge of multiple production lines or a broader scope of plants, where it would be virtually impossible to gather up all that information and surface what's important. So in the detect part of the cycle, deciding what we need to focus on, AI can play a great role in gathering that information, cross-referencing it with what's important for the organization, what's important for the plant, what are my goals this week, what are my goals this month, and what do I need to focus on to help drive these goals? And agents play a great role in goal seeking and helping teams align behind these goals.

Step 2 – act. It's about solving problems, the problems you've identified that you wanted to solve. Where AI brings value here, and I was mentioning it earlier, it's what has my organization done previously to solve a similar identical issue. A common pattern in manufacturing is that most issues are known, perhaps not documented, but someone's run into that issue before at your organization. Very few problems are actually new at the end of the day. And so the ability to give those teams access to the entirety of your organization's know-how on a specific problem, how it's been solved, what worked, what didn't, is again a great capability that AI can bring to the table in this second step of the CI loop.

The next step is learn, and that's how do I capitalize on what I am learning. Some of these problems are new and you have great teams out there coming up with innovative solutions and how to solve them. The trouble there is how do you translate this local win or improvement into a global improvement that the entire organization can capitalize on, right? AI also here brings value to the table because it can actually detect these improvement opportunities from a problem resolution that happens somewhere in the shop floor and perhaps suggest the creation of a new standard or update an existing standard with this added capability.

And last but not least, ensure – this is the role of orchestrating these entire systems, these CI systems that run at scale. It's making sure that the entire organization is aligned towards the same goal and that the system, the CI system, is actually delivering the business impact that it promised, right? How does the quality and the maturity of my process across my plant network drive business results? Today what happens at a typical manufacturing organization is that you have these maturity audits where the maturity of your CI system is audited maybe once, twice per year across the plant network. It often requires traveling, being on location. It's something that's slow, not a very fast feedback loop. What AI can do is it can actually evaluate in real time the maturity and the quality of your CI system, and perhaps give you the opportunity to act on it much faster. And because those are leading indicators of performance, the ability to better predict and to be more deliberate in driving performance.

IW: Octave, there's a lot of concern about AI, generally speaking. Do we think it's going to take over the world? Probably not. Wipe out humans? Probably not. But the concern that AI will replace human expertise, that actually sounds understandable to me. So what do you think should remain fundamentally human in CI, and where should AI take on more of the work?

OL: I'm of the opinion that we should always leave the human in the driver's seat and keep them in the loop. For us, critical reasoning, decision-making needs to sit at the end of the day in the humans' hands on the shop floor. AI is much more useful not to replace these decisions or these jobs that are being done on the shop floor today. It's much more useful in doing the work that nobody's doing today or that teams aren't able to do enough of. And it's the work that falls through the cracks. It's what I was describing, right, is taking notes, connecting the dots across a plant network, having the discipline to follow up and see if a problem's actually been solved and push the teams to go document, did this solution, et cetera, et cetera. So it's not about, replacing the human in the high value added critical reasoning tasks. It's about doing all the busy work around that so that the system works as a whole.

I'd add that, you know, at the end of the day, agents and AI models make mistakes. They hallucinate. They don't always get everything right. And I wouldn't advise anyone to put an agent in the driver's seat of highly valuable production lines with the kind of the high value products that come off of it. It's interesting actually, philosophically, when we speak to manufacturing leaders, they agree with this. I always ask them, take the example of problem solving, your team sitting in front of a blank page, root cause analysis, they've got to fill out their fishbone diagram, then do a Five Whys analysis. Do you want those teams to be able to, at the press of a button, ask an agent to analyze the problem, use your existing knowledge, and suggest or write a first draft? Or would you rather have the AI actually proofread, suggest alternate paths, make corrections or suggestions to what could be improved about this problem resolution.

And they all agree, they much prefer this vision of the world where the human is left in the driver's seat and AI is there to empower them more so than they are to do the work instead of them.

IW: So, Octave, if agentic CI works as you envision it, what changes not only for operational performance but for the people doing the work? How can AI help manufacturers capture expertise, develop skills, spread know-how across sites, and ultimately create more productive and competitive manufacturing operations?

OL: I think the biggest problem that this solves is the circulation of knowledge. Today, knowledge is often trapped. It sits somewhere in the organization in someone's head, or it's constrained to a specific production line or plant. And the way agentic CI helps us is actually creates this self-improving system that puts this knowledge in a network, right? And specifically for the frontline workers, the people that are operating the system, what it means is better recognition for your knowledge. So the opportunity to contribute to that knowledge base proactively and be rewarded and recognized when you do that. Another aspect of it is instant access to the existing knowledge, right, for you to be trained if you have a question. It's really all about empowerment, right? You're giving, and this is the key to Lean, is autonomy on the front line. So if you're giving everyone on the front line the opportunity to train, to learn, and to have access to all the necessary information that they can solve their own problem right then and there, truly you're delivering on the promise of Lean. And this is what agentic CI addresses at the end of the day. It's driving performance through frontline autonomy.

IW: All right, Octave, let's zoom way out here. What does getting continuous improvement right matter beyond the four walls of the factory? What role could more productive, capable manufacturing operations play in strengthening industrial jobs and manufacturing communities?

OL: I mentioned it earlier. The cost of running a plant inefficiently is rising. Of course, energy costs, supply chain volatility, regulatory pressure, global competition, what have you. And often the efficiency of a plant, its productivity is a deciding factor in whether this plant gets offshored or not, or whether it can be reshored. This is what we're looking at. And of course, a strong local manufacturing sector has tremendous benefits on a society and economy. It also is, as we're talking about empowerment and of the front line, it drives actually not fewer jobs, but better jobs, higher value-added jobs because you're relying on people's decision-making ability rather than, you know, the ability to perform a repetitive task. And so there's a tremendous amount of value in running a plant efficiently that goes beyond the bottom line of the organization that owns this plant. I'd say last but not least, CI is not just about economic efficiency. It's also about safety and quality, right? Safety of the people making the products that come out of these plants, but also safety of their consumers. Vast sectors of aerospace, pharma, medical, that if things go wrong in production can have a tremendous impact on the on the consumers of those goods and services. And of course, it goes beyond the economic impact of productivity.

IW: Unfortunately, that's all the time we have for today. I'd like to thank our guest, Octave Lapeyronie and our sponsor, Fabriq, for today's discussion. Octave, where can listeners go to learn more?

OL: So you can go to www.fabriq.tech. We'll have a blog post on agentic continuous improvement that your listeners can give a read. And of course, if you want to find us on LinkedIn, if this has sparked some thought, feel free to tag me or Fabriq and share your comments.

Sponsored by: