From AI Ambition to AI Impact: Trusted, Connected, and Accurate Data

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 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'm the CEO of the Industry 4.0 Club. I have over 40 years of leadership and management experience, most of that at Michelin in manufacturing. I'm joined today by my co-host, Mike Yost. Mike is chairman of the club. Mike, would you introduce yourself further, please?

Absolutely. Excuse me, getting all choked up. My name is Mike Yost. As Mike said, I'm one of his co-founders of the club and also the volunteer chairman of the organization. I've been in industry for almost 40 years myself, pretty much the first half of my career on the technology solution provider side and the last half of my career on the industrial nonprofit side. So Focus has been entirely on the intersection of technology and the value with manufacturers and how we come together and figure out the role and the value and the fit of technologies. So I'm really excited that today we have the pleasure to speak with Damantha Bateju, who's the Chief Product and Technology Officer for Configit. And I'm sorry, Damantha is going to talk to us about moving from AI ambition to AI impact and leveraging the trusted connected and accurate data.

Excuse me.

Uh, so just by way of background quickly, uh, Damantha joined Configit in March of 2023, and he brings almost 30 years of expertise in software development with a focus on enterprise-class products. In his role, he leads product and technology strategy, driving innovation and delivering solutions that meet the complex needs of Configit's Global customers. So welcome, Demantha.

Thank you. Pleasure to be here.

Yeah, thanks. Before we get into the questions, I just want to share that Configit hosted its 11th Configuration Lifecycle Management Summit just recently on May 6th and 7th. The event focused on how manufacturers can master customization complexity by establishing a strong data foundation for future growth. The summit brought together B2B manufacturing leaders. to explore how connected systems, data, and logic enables organizations to manage complexity more effectively. So AI was a major discussion point at the conference, and we're going to speak to Samantha today about some of those key points that were shared at the summit. So anyway, we're glad that you could take time out of your schedule to join us. Maybe we can dive back— dive right into a couple of things. So, you know, I know AI, we talked about, was such a significant conversation, but Many manufacturers have a lot of ambitions around putting AI into practice, but fewer today are seeing measurable impact. So what separates organizations that successfully translate AI vision into operational value from those that stall?

Sure, and I think for me, at least what I've experienced, is it's not the ambition that is lacking. I think AI is everywhere. We're being inundated with the hype, but I think the difference is actually operational readiness. I think the organizations that typically succeed are the ones that do 3 things well. The first is that they connect AI to very specific business outcomes, you know, either faster quoting, fewer engineering changes, improved service quality. So they're not doing it for experimentation alone, but they're taking a look at some measurable operational outcomes and applying it that way. And honestly, I think it's the same thing that we've said about pretty much every technology that we've seen.

Absolutely.

And probably you've probably heard that before with SaaS, with personal computers. You need to start somewhere for sure, and you need to get your feet wet in some way. But it's those organizations that have identified this is a technology that allows me to do something that I couldn't do before, but now I can do that. Now I'm going to use that. So I think it's also selectively choosing what it is that they wanted to—

What you're saying is focused on solving a problem for them.

And I think as a product guy, right, I think that's one of the things that I've learned is to try to avoid being enamored with the technology and be enamored with the problems and, you know, work backwards and say, what problems are out there? How can I change people in— as people's, you know, work, life, whatever, in a meaningful way? What problems are so insurmountable that we couldn't solve it before, but now I can with this thing? And I think it's really if you start there as a— when you're building your products, that's where you have the greatest success, rather than just trying to, uh, what's called, uh, shoehorn some technology into a problem that might not need to be solved that way. And I think that's also a big part of it.

Well, I think that's— we could almost mic drop and stop the conversation right here, right? Like, that's a really powerful point, and I think everybody should take note of that. But obviously you can't stop there, right? So I think having that focus is the right focus, and I always encourage manufacturers that that should be reassuring for them because instead of chasing the tech, you're focused on your business problems and how that technology aligns to support that.

Yeah, definitely.

You mentioned 3 things maybe.

Yeah, sure. I think the second thing is a strong data foundation, especially with AI. You need to have some context in order for it to work well. Product rules, dependencies, constraints, requirements, you know, pricing logic, whatever. If that data is fragmented or inconsistent, then it's going to be difficult for AI to produce some results. I'm not talking necessarily about machine learning, but having some frame around which the data or which the AI that you work with can reference that is a part of your business. I think that that's super important, having your data foundation in place. And I think the last thing is making sure that you've enabled AI in a workflow So that it's a part of a bigger picture and not just trying to treat it in isolation. I think it needs to be able to support real decisions across an entire, you know, work stream rather than saying, let's solve a point solution. I think many people maybe use it now, like, say, ChatGPT or something similar, to say, okay, I gotta write an email, I gotta write a specific document, and then I'll use ChatGPT to help me formulate that. That's fine, that's, that's okay, but it's not a scalable, uh, thing and you're not really revolutionizing it. You need to again think about an entire value stream and where can we improve that. So I think, I think it's those 3 things, which again I would say it's like probably the same thing as what would apply for any kind of technology, um, not just AI, but it's probably more important here now because there's so much hype.

I think folks need to be careful of that. Well, I know that the summit focused a lot or talked a lot about AI and the point that you're making about the need for the trusted, connected, accurate data as a foundation. Leaders are hearing this and we're not saying it's easy, but we're saying there's a path forward. How do folks assess whether their data is ready to start leveraging AI at scale.

Sure, and I think what I hear sometimes is people confuse this trusted data with clean data, that it's gone through some sort of scrubbing process, and it's more than that. It's about the ability for that data to accurately reflect what can be sold or engineered or serviced, and trusted data is something that you, your organization feels that that they can use without it needing to be constantly revalidated manually. That when you— when you— when it comes to you, it's, it's not like, oh, let me just double-check this 3 different times in 3 different places because I'm not 100% sure about this. So I think a simple, a simple test is that if your sales or engineering organization can answer the same kind of product questions, you know, do they arrive at the same— if they're asked the same question, do they arrive at the same answer? for example, in different parts of the organization. And if that's not the case, then your organization doesn't have trusted data. And another thing I think is important is whether or not this data is very explicit or whether it's kind of tribal. And with tribal, I mean, like, does it live in like individual spreadsheets, in people's heads? There's also this thing about the aging workforce, and there's many companies that are worried about, okay, how much, how much is in how much is in the heads of people that are now leaving that workforce, how much of it is just systems that work sort of organically because we've always done it that way, which could be good because you don't need to put a bunch of systems in place, but it also leaves a bit vulnerable. And I think AI can't rely, it can't scale if you don't have that data out of people's heads and into some sort of system because it's not going to be able to reflect it. It's just going to be like if you just take a random you know, chatbot and put it in front, it's not going to provide a great amount of value because it's just generic answers that it's going to give.

Right. You know, you're— when you describe that, what's in people's heads, and then I almost think about it in 2 different— 2 different— I'm gonna say 2 different buckets, because at least as I hear you talk about this one bucket, which is, you know, the data I'm gathering every day and the system's running, and it's generating feedback and there's a certain range. And then what's in people's heads? Well, yeah, hey, I remember 5 years ago, this thing happened and that's how you deal with it. Is that really the 2 different things we're talking about here?

Yeah. Also, I think just how to systematize it in a better way. It is exactly that thing of like, oh, we know this one particular customer, they have a preference for this. There's also the stuff about interpersonal relationships that you know, can affect the way that you work with customers, I think, as well. But a big part of it is those things that you learn along the way that you never remember to put somewhere.

Definitely.

And I think that, you know, it's not going to work if that's a big part of your way of working. Like, AI is not going to help you if all of your true valuable data or knowledge Is, is, is somewhere else.

Well, how— so just as I think about that, and I know companies will struggle with this, you know, okay, look, I can go out and measure what's going into it today and say, okay, this is true, I trust this. But for the guy that remembers that this customer did this, and it could be what he thinks he remembers or she remembers, yeah, or how do you validate that? I mean, or, or, you know, I mean, how do you go about validating that sort of tribal knowledge? Have you seen some examples of where that's done effectively?

I think it's actually being explicit about it and maybe being a little bit more systematic about it, whether or not you— whether that's a wiki or some level of case management documentation, but at least it's written down. It's difficult, and I think it requires a lot of discipline. I think that's the biggest thing. It's what— because there's plenty of systems to to do knowledge sharing, right? I think there's tons of corporate systems that allow you to, you know, that have that promise of, you know, we will be the repository of company information. I think it's the organizations that are disciplined in that approach or recognize that there's value in doing that. So, for me, one example is as a software vendor, how much or how little do we record how it went with customer interactions? and whether they were successful or not. And then do we learn based on that? And how much of that do we use to, you know, focus the efforts of our new product development, for example? And if it's just sales conversations that get lost somewhere, then it never comes back to product management. So there needs to be, if you're disciplined enough in terms of saying, okay, well, for the 30 customers that we talked to, 10 of them said this and 20 of them said this and And but that requires human discipline, in my opinion. I don't think there's any way around it.

Yeah, I know humans overall struggle to do that, the conversational piece, and then whatever discipline it requires to enter things into a system. If it slows the process or if it's an extra step, that's a difficult thing to manage. And, well, you know, I grew up on the plant floor In that environment, you have people whose job it is to make stuff and to keep it running and to keep things safe. Anytime you introduce another step in that process, that's really a difficult ask for people, especially if they have to leave what they're doing to go to some system to capture that knowledge.

I completely agree because it's like, I got my real job to do, but now you're also asking me to document this for for something where I can't see the purpose of it. And I think actually, that's a great place for modern AI tools to help with that. So, there's plenty of tools that allow you to record sales conversations, do analysis afterwards. Just meeting transcription tools, I think, is fantastic because then you can— all you need to do is make sure that it recorded the intent correctly. So, I think there's a lot of— I think that's a great place to help actually. to lift the burden of people needing to do that themselves and allow basically secretary, automatic secretary to take care of that for you.

For sure. No, it's so interesting you talk about that. Unfortunately, I can say that I've been part of a couple of knowledge capture initiatives in my career that didn't work because of the The tools weren't very easy to deal with and the discipline wasn't there, but not really the discipline, but they were so hard to do that we didn't have the will to do it, I think, even though we knew it was important.

I think that's, again, it's the same thing as with all technology transformation projects and the core reason why they either fail or succeed is whether or not you have human buy-in to to this and whether or not you're actually working on the change management part rather than whether the technology is good or bad or identifying whether or not this technology is appropriate for this and getting people along with you. And I think it's the same thing that we've seen for the last, you know, probably 1,000 years. I'm sure guys that were introducing chariots probably had the same issue as well, getting their comrades to use chariots rather than horses. I would imagine. Yeah, yeah, yeah.

I've done a little bit of research. I've always— because we're always talking to people and to come back to your earlier point, like, what is— what's the business value of doing this? Right? What's— and, you know, we have a high-level understanding of transformation and digital value and things like that. But then you think, well, what's the value for me in my role across the organization? Right? It gets harder to be that explicit. And I've done a little research on like the industrial revolutions and did everybody just accept it? And there are plenty of stories of people that fought, you know, that kind of change tooth and nail. And so yeah, it's not just a natural thing for people to accept it and, you know, see the value intuitively. So I just wanted to take one second and sort of reset things with everybody and remind you that we are speaking with Damantha Bateju, who's the Chief Product and Technology Officer for ConfigIt. We're talking, following up on the Configuration Lifecycle Management Summit that just wrapped up recently. And Demantha, you know, CLM, Configuration Lifecycle Management, we spent some time with your team and introducing that concept to our audience and stuff. But maybe take just a second, sort of anchor that with folks, because you're talking about working across different departments within a manufacturing business, right? The engineering, the sales, you know, the delivery, the production, the design, that sort of stuff. So when you get into complex products like that, and the focus of the event itself touched a lot on AI and how AI can accelerate, you know, the value for these types of companies, can you talk about really what that means? Like how, you know, what acceleration looks like or where the value you will find in working across those different organizations?

Sure.

So I think it goes back to 2 of my points, data and connectivity. CLM helps create this context that we talked about that, that AI actually needs. For manufacturers, you know, with complex configurable products, it's not just about having good product data, it's also about understanding the logic behind the product. So which combinations are valid, which constraints apply, which choices are are good or bad, and how does it affect lead time and deliveries, etc. And CLM helps capture that logic across the entire lifecycle and also helps create the data foundation. So a lot of it is around, you know, the engineering intent and making sure that the commercial execution matches the engineering intent, i.e., don't try and sell something that you can't build or that was not the purpose of building it. It means that, you know, trying to establish a shared source of truth for how products can be configured and how they can be delivered. We use the term shared source of truth because I think there's a lot of especially software vendors that are, you know, racing to or positioning themselves as the single source of truth. And at least in my opinion, it's difficult because there's so many disparate systems that are competing with one another. And at least from our perspective, it's about trying to establish that there's a— that there could be multiple individual sources of truth. We just need to be able to make sure that they can talk to one another and be aligned with one another rather than saying that it's going to be this one thing that is the ultimate source of truth for your entire organization and just being really realistic about what that means. And I think for CLM, it is really about joining the individual pieces of the organization together and making sure that you have like a good stream of information. And then from an AI perspective, that creates something that's extremely powerful because when you have access to all this kind of data, it just makes the decision-making a lot better. Right, right.

And it does align with promises we've made for a long time. So I'm an old MES guy and we always talk about, you know, matching production with demand. And so now we have better tools to see what that demand is or what it could be and better signals from the market and with suppliers. And you can run the— get better information to the point that you're talking about and then say, all right, it's time to go. And that's where— so I guess I hear it through the lens of good, we're making significant progress here because we're looking for those signals.

Yeah.

Yeah, exactly. And I think it's interesting because it feels like it's many of the things that people have talked about for a long period of time. And it's just, I think every— it's an iterative process. I think, you know, you have some ambitions and then maybe it works, some of it works, we take some steps forward. And I think that those signals, I think especially that ability to react, it's kind of the same thing as I mentioned from a product management perspective, how to gather those signals, that no longer is just in people's heads, but how to systematize it, how to make decisions on it. I think it's also that is, do I have good metrics and a base, a foundation for making decisions? It's not just about collecting the data, right?

Is it—

does it inform me? And do I actually take action on that?

Yeah.

Or is it just a dashboard that appears and I'm just like, oh, okay, so it's gone up, uh, since last week, but I don't care, right? I mean, there's— if there's no feedback loop in there, then, then, then it doesn't make a difference.

You know, you talked about the single source of truth, and I think it's— I know there was a lot of conversation around deterministic and reliable outcomes as well. How do you avoid— how do you avoid— I mean, today we know that AI is going to give us an answer, whether we you know, because it's programmed to do that. But we always— we know that it isn't always for those of us that use ChatGPT and Claude, we know it always isn't necessarily the truth, right? So how do we make sure that the, that the answer is truthful, that we have the single source of truth? And then the other side of it is your point. When, when, how do we make sure we develop our workforce to have the judgment to, to think about the decisions and to maybe challenge what they see a little bit. You know, what are your thoughts on that? Sure.

So I think it's very much in my mind is using the right tools for the right job. I think large language models as an example of AI, because AI is so broad, you know, there's computer vision, there's so many different types of what you can consider AI. There's neurosymbolic AI, blah blah. Uh, large language models, they're great at pattern recognition. That's how they're built. They're very probabilistic. You never— you're not necessarily guaranteed to get the same answer twice if you ask the same question. Um, but I think there's a good place for that, which is, I think, especially in human-computer interaction and softening the, uh, that human-computer interaction that has been super rigid Up until now. If you think about dropdown boxes or, you know, just even wizards that guide you through, it's very much based on the intent of the creator of the application. They think you should answer these questions in this way and in this order, otherwise it gets messed up.

Right.

And they're definitely not going to be able to answer questions sort of like a little bit off-piste. Right? It needs to be a very, very, very, very— what could you say— sequenced flow of events, structured events. This is a great place for AI to come in and have these natural dialogues, to have a more unstructured dialogue with people.

And—

or sorry, when I say AI, I actually mean large language models, because that's what they're good at, good at simulating human interaction. I think that's a place to use it. But don't use it to calculate partial derivatives or something like that, like partial differential equations. Don't use it for that and expect it to work at speed. Even just simple calculators, if you look at ChatGPT now, and I did this recently, I said, what is 18 plus 35? It kicks it over to an old-school calculator. Um, because it— I think the people behind ChatGPT have recognized, okay, let's just use an old-school calculator when somebody asks us how to do that, and then it'll feed it back to you, which is good because they've understood that their tool is good for this bit, use another tool that is good for this bit. And I think that, that the core of maybe what I'm trying to say is use it, use large language models for these things where soft nonlinear conversations need to happen to gather data, to gather information, but I wouldn't necessarily trust it to be necessarily the ultimate source of truth for a system, especially if it needs to happen at speed and at scale. I mean, it's so probabilistic, you don't know what the response times are going to be.

Yeah.

And if you need to have sort of precision responses within a given period of time, I wouldn't use it for that. So use it for these things, use other tech. And there's, there's a lot of techniques that allow you to marry AI front ends with deterministic back ends. So just use the best of both worlds in the same way that ChatGPT itself does. Yeah, so that would be my take on it.

It's another great point of, you know, language and understanding, you know, truly having your arms around, because you can't even just talk about AI because there's so many It's such a broad term and so many approaches and technologies like you just talked about. I just, I want to come back, you know, to the summit. And, you know, I'm assuming that folks can still access some of the learning from the event. If either folks attended, got energized, or they go now and they want watch some of the content, they get excited about it and they want to move forward. What are the first steps that you recommend people could take, you know, over the next 90 days to move forward and make sure that they have this data foundation in place? Because I love how you started things off, like ambition's not the problem, right? It's the impact. So how would you advise people to take their first couple steps forward?

So I would start by choosing one high-value business process and try and examine whether or not there are new ways of thinking about this. Don't try and do AI everywhere.

Mm-hmm.

I think it's going to be super difficult to roll it out and just say all of a sudden, okay, now we're going to be an AI-enabled business. I think there's an okay amount of business literature that says successful technological rollouts typically start with pilot projects, or it's a good way, at least it's a relatively good way to get your fingers dirty, get some organizational engagement, create some champions around this technology. Also, if it's successful, relatively narrow, and that you can show some business benefits, that it acts as a, what do you call it, like a beacon for others in the organization to say, okay, well, I mean, it worked over here, so let's see whether or not it can work over here. And I think it is really, I think it is kind of change management approach, I think, as well. And a big part of it is you, again, using what large language models, AI, is good for rather than expecting it to be a panacea to do everything. So I think that, that would be, uh, that would be a great place to start. So in 90 days, uh, don't, don't expect you're going to solve all your enterprise problems with it, but at least within 90 days, identify really good solid cases to begin to dive deeper into and also make sure that they are based on some business outcomes that you have wanted to try and achieve but have been unable to do. So I was thinking about, you know, we talked about this earlier, if I asked a horse and carriage manufacturer from the 19th century, I need to get to— I need to go on ski vacation and it takes me 7 hours to get there. Build me a thing that can do that. What they would do is probably take 300 horses and put it on the front of a carriage and say, that's your solution.

Right.

Or they would say, it can't be done because I need to put 300 horses in front of my carriage to get you there in 7 hours to do that at a reasonable speed. I think my point is that now we have this technology that can do some stuff that you couldn't do before. You could solve problems now that you thought were unsolvable. So I think a big part of it is to reset your thinking and say, okay, how can, how can I get from Copenhagen to Switzerland in, in 7 hours?

Right.

Right.

Yeah, yeah, I use a similar analogy, trying to have a similar point around like wearable medical devices like sleep rings and things like that. You can do things today inexpensively, you know, that you couldn't a decade ago, right?

Yeah, exactly.

And end up in a, you know, a special facility hooked up to all sorts of sensors and things like that versus wearing a ring today. I think it reinforces a point you mentioned before, is that the problems we're looking to solve aren't necessarily dramatically different from years ago, but the options that we have available to us in order to attack them are changing and are more powerful. And so it goes right back to your beginning point of what are you trying to solve? And now we have these new technologies to help.

And I think I don't know if you guys— you guys probably remember Clippy in the '90s with Word, and I think that's the classic example people use. The intent was there. I think it's the same thing that it tried to do as what Copilot is doing now.

Right.

It just wasn't super great, but the problem was there. The problem was that people don't like to use menu items and dropdown boxes and I can never find out how to align all objects in my PowerPoint. It's like I always need to— I can't find that even though I've tried it like 100 times. Maybe that's just me, but you want to be able to have like this natural conversation with computers. And so I think we just need to take a look at what can we do now and what can we offer our customers, I think, as well.

Right.

When it comes to guided selling, I think it's a great example of using AI. It's to say, okay, we'll actually gather the intent rather than coming in and coming with these dropdown boxes. I mean, again, this thing with— I recently bought an electric car. Getting into that whole world of what does it mean with this kind of consumption of number of kilowatts per kilometer and things like that. I have no idea. I want somebody to tell me whether or not I can take my skis or not in this car. What's the best way? What's the best car to do it so that I can fit all my family and my luggage and things like that in there? So I think there's— I just feel it's going to get better and better. We just need to rethink what's possible.

That's excellent. You know, you've shared some really important points today, and it's really been great to To, to sort of hear your reflection on this, this issue about the importance of data and, and then AI, and then how to get started on that journey, and really things that, that you talked about during the summit. So I really appreciate the time that you spent with us. So I want to thank you, Damanta, for joining us to, to discuss the importance of accurate data in making AI successful for manufacturers and sharing your insights from the CLM Summit. you know, that took place just, just a couple weeks ago. With that, I'd like to just share that the Industry 4.0 Club would like to encourage you to take action to advance the deployment of Industry 4.0. So read an article, attend a workshop, develop a smart manufacturing roadmap for your plant or company, or challenge your organization to, to look into AI, as we've talked about, and look at your data and identify the problems that they want to solve and understand the tools that are available today to do that. But take some action. We host conversations like this routinely. We'd love to hear from you, hear the things you're doing to advance the deployment of Industry 4.0. And if you would like to be on one of our future broadcasts, we'd also love to hear from you as well. Please visit our website at www.industry4oclub.com and let us know what you're doing, or drop a note on our LinkedIn page. Again, we'd love to hear from you. With that, Samantha, thank you so much for the time today. Yeah, it's been great, Mike. Thank you as well.

Of course.

And, yep, take care. We look forward to hearing about the future of configuration lifecycle management and the impact it has on, on business. So thank you.

Yes, thanks, Samantha.

Thank you.