The Physical AI Revolution Starts With a Reality Check

0:00Shane Kehoe: AI is moving on to the factory floor with the promise of machines that can understand what's happening around them and decide what to do next. For manufacturers, that can mean less downtime, better quality, and more productive operations, but getting their means changing systems that companies have spent decades making reliable. So how do you introduce AI when stopping to experiment carries a cost, and getting it wrong could carry an even bigger cost? On today's show, Alan Hinchman, the Chief Revenue Officer of the industrial technology and consulting company Gray Matter, joins Shane Kehoe, WWT's global head of its manufacturing and distribution practice, on where physical AI is starting to deliver, what stands in its way, and what it takes to trust a machine with more responsibility.

0:45Shane Kehoe: This is the AI Proving Ground podcast from Worldwide Technology. Let's jump in.

0:57Shane Kehoe: Well, Shane, Alan, welcome to the AI Proving Ground podcast. How are the two of you today?

1:02Alan Hinchman: Excellent. Thank you. Yeah, looking forward to it.

1:05Shane Kehoe: Yeah, wonderful. Thanks for having me on. Yeah, appreciate you both taking the time out of a busy schedule. So we'll get right into it. I mean, compute is getting cheaper. Models are becoming more powerful and efficient. And that's certainly helping unlock any type of intelligence that we can look to squeeze out of or in the physical world, so to speak. I mean, you could argue, and I think maybe, Alan, you have at some point, that we could be on the cusp of another industrial revolution here. So, you know, Alan, just to level set here, you know, what is becoming practical inside the factory floor or the plants or any type of operations?

1:40Shane Kehoe: What's practical today that we couldn't accomplish yesterday or a few years ago? You know, I like to look at the industry in kind of eras, and I'm old enough to have lived through a few of them. And so there was an era where we digitized the physical world, and it really ended in kind of the late 80s. And that was a wonderful time until the PLC, which is programmable logic controller with the factory floor uses to control things, was kind of king. And that went into the mid-90s where PLCs then needed a visualization world, and the SCADA world was built. I was working with a large environmental consulting company in the mid-90s, and we had just completed a multi-water plant upgrade to this new world of SCADA.

2:31Shane Kehoe: And the operators asked us, is it cheaper to start this pump over here or to increase the speed on another pump? And frankly, we couldn't tell them that answer. And it was embarrassing because they had just spent, at that time, millions of dollars on this system. But the system was a display and it was a control, but it didn't have intelligence. And that really kicked off into the next era where we began to be decision support. And that's kind of where the where we began to see digital twins, where we could use that to make better decisions on the factory floor to if then. And what controls controlled that, and there were many that tried and failed to make that a global kind of world worldview of what's happening was compute power.

3:20Shane Kehoe: And just like you said, now the compute power is changing. What we're seeing is AI begin to drive into, well, I'm not going to just support the decision, but potentially I'm going to help offer up solutions to that. So in these four eras from the physical world to the visualization to decision support to now AI, we're beginning to see people looking at things that they would have never dreamt of controlling before, wanting information from. They're wanting that because now controlling that with large language models, understanding and being able to put that into a way that they can drive better decision support in a place that could never stop running is really the era.

4:01Shane Kehoe: And that's where it's at. So folks want that data, and we've captured that for the years. They want to contextualize that data, and that has been done, and now it's they want some intelligence from that data. And that's kind of where the factory floor is. Yeah. Alan, I mean, missed opportunity. You say you're old enough to consider these in eras. You could have just said you're hip enough to consider them in Taylor Swift-type eras. Shane, maybe get a little bit more specific here. I mean, what is it actually unlocking within the plant floor from a use case perspective? You know, what are we able to do now?

4:33Shane Kehoe: What are we able to do in the near future? And what are maybe some lofty visions for what we can expect?

4:38Alan Hinchman: So, yeah, I think, you know, it's interesting because I've been in manufacturing for 30 years. And, you know, now we're starting to see certainly like a lot of the language around manufacturing and, you know, a lot of this language around physical AI. really the market is changing what Alan was just saying. And it's changing very, very quickly. And a lot of that is due to, obviously, a lot of the language and a lot of the press and a lot of the marketing that's coming out around AI. And a lot of the language around physical AI. And a lot of these companies are really starting to pay attention.

5:15Alan Hinchman: And it's certainly the biggest shift that I've seen in the 30 years that I've been in manufacturing easily. And there's a lot of money that's now going into manufacturing. If you think of manufacturing from kind of a bigger perspective, certainly, like I said, over the last 30 years that I've been in it, manufacturing has always been a cost center. It's never really been an area of value creation. And that's the way manufacturers like it. It's the classic, if it ain't broke, don't fix it attitude. But what's starting to happen now is manufacturers are seeing manufacturing differently. And they're seeing manufacturing from a perspective of how do we really drive value?

5:59Alan Hinchman: How do we drive value out of the plants? How do we get more value out of the plants? And a lot of this has been driven by that vision, what you just said, the vision of physical AI. What does that mean? How do we get AI into the plant floors and how can we drive value? But the big challenge is, like I said, manufacturing has been a cost center for decades. So one of the biggest challenges that these manufacturers have to deal with is, you know, legacy systems, aging workforce, massive data silos, limited security. So how do they get past that shift to be able to then drive, you know, these high value use cases?

6:42Alan Hinchman: is where we are seeing success today, I would say, and Alan touched on this, is like a lot of the solutions that you hear around physical AI, things like vision AI, digital twins, humanoids gets brought up a lot, fully autonomous factories gets brought up a lot. But remember, vision systems have been around since the 1980s. Digital twins have been around for 25 years, I'm gonna say. So these aren't new technologies. It's just what's new is the ability to leverage AI with these technologies. So if we take vision systems as an example, vision systems have traditionally been hard programmed. Okay.

7:32Alan Hinchman: But now what can you do with a vision system when you've got AI, AI patterns, AI modeling? I mean, you know, the flexibility of those vision systems becomes way, way, way, way, way more. So you can start to drive these types of use cases very, very quickly. But a lot of this scenario comes back to, like I said before, how do you go from the manufacturer's reality today to the vision of physical AI that they're hearing about in the press, in the media, in the investment community? Alan, what are your thoughts on that?

8:08Shane Kehoe: Well, I think what a lot of folks may not get, if you haven't spent a lot of time on the factory floor, it's actually a very direct, local-driven thing. If you think about the car that many of us drove to work today, it's a single casting meets a single injection-molded car, which meets a single piece of leather, which meets a single piece of material. and ultimately all of those have to work together in unison. So when you think about why does local AI matter? Well, it's because there is no global auto industry. There's a global assembler of global parts that all have to intermingle in the right way.

8:54Shane Kehoe: And that plant can absolutely never go down. And when you do, you're looking at tens of thousands of dollars of loss every single minute that that plant's not running. And within six minutes of an auto plan going down, the tier one suppliers begin to shut down. Within a few more minutes past that, the tier two. So you're talking about a stacking of data and information. So when you think about where the industry is, is the seriousness of how is AI going to make a decision that is informed enough across everything that we do? Because it's very easy, especially with digital twins. And this was the, I would call it the conundrum that happened with digital twins.

9:37Shane Kehoe: It's really easy to say, we're going to make this part to the absolute cheapest possible way that we can. We're going to save a little bit here. We're going to save a little bit there. We're going to model that out. And then that part not fit with by color, not fit by texture, not fit with another part that it's supposed to. So having those interconnections is also important. So it's the scale of the challenge is local. It's intimate with that individual piece. But then it also has to be informed by finance models and PLM models and maintenance models and all of the other pieces. Then into a control system, then into a supply chain control tower, then into a whole bunch of other things.

10:23Shane Kehoe: And so that's what the power unlocks, is the ability to understand things that are beyond what can be ran through in a very quick way using the present systems. Yeah, Alan, I want to maybe double click on what Shane said about how, you know, manufacturers might be thinking about manufacturing differently. And by that, he was talking about, you know, how can we leverage AI to create more value? You know, that's where you and Gray Matter, you know, seem to live on a daily basis now. What is the playbook for going about creating value? Is it thinking big, shooting for the stars? Is it starting small and developing incremental value?

11:01Shane Kehoe: I was reading an article just the other day about GXO expanding its humanoid deployment after using it for a really specific material handing ask. And it struck me because it wasn't like a giant use case that was like, blow me away by the headline type of thing, but it was certainly providing value. Is that the inroad right here is to just start small and work your way up? Well, yes. And what we've seen in our business over the past 12 months has changed from how manufacturers are beginning to look at this problem differently. And so what Gray Matter did earlier this year is we acquired a factory floor consulting business.

11:41Shane Kehoe: And what they do is lead manufacturing and quality, because what's being asked is how do we not just how do we go make 10 ,000 of this individual part? But how does that fit into our business framework? That is the number one driver change that we've seen is the accounting and the IT worlds coming in and wanting to understand how does this contextualize out with the rest of the system. And a modern partner is not just a control system integrator as we were for many years. It's going to be a true business partner that understands, here's how we're making this. This is how we're moving this materials through.

12:27Shane Kehoe: And then it comes down to what's interesting is a lot of the things that were the kind of the idea of modern manufacturers, which is organized systems. And so you would go in and you would see a warehouse and everything would be neatly stacked. We were with a customer a few weeks ago and he went in. It was just absolute chaos in their warehouse. And what you find out is it wasn't chaos. What it was, it was an algorithm that was running that was determining what is the fastest way for the robots to drop off a load and prepare it. Because they have a different level of understanding. There's a different language that happens there because they don't have to have it in bin 23 row six.

13:14Shane Kehoe: They know where it is because of the DPS. So they're repositioning things. And it looks from from our outside view. You look at it, you know, like, man, there's just stuff everywhere. What you find out is, no, it's exactly the shortest path that it was going to need to be between the handoff of that pilot to the next piece of the line. And those changes are happening rapidly. And that's what we're going to see. We're going to see a new factory floor emerge out of this where algorithms are going to challenge the systems that have been in place. And that's what we're seeing already in the early innings of this game is folks are asking us, how do we reposition things?

13:54Shane Kehoe: How do we think about this differently? And I think, Shane, you've seen a bunch of that, too.

13:58Alan Hinchman: Yeah. Yeah. You know, I think like it I was reading this very interesting statistic just earlier this week. And it was 88 % of AI pilots in a manufacturing environment don't make it to production. That's a crazy statistic, 88%. So, you know, you got to ask, you got to ask why. And I think as you peel back the layers of the onion on that, it's, you know, there's a lot of factors. One, you know, are they targeting the high value use cases? Like, what is it that they're trying to do? Two, the other piece to this that I think is really, really important is culture, change management, people. If you don't have the right people on board across the entire organization, it will fail.

14:51Alan Hinchman: And a lot of these use cases, you know, that we're seeing, yes, it's a manufacturing use case. Yes, it's a productivity use case. Or yes, it's targeting productivity. It's on the plant floor. But if you don't have the right people engaged, if you don't have IT sitting in the room with OT, with the security folks, with the executive sponsor, if you don't have buy-in across the entire org and have everyone committed to what is the success of that actual use case, it is going to fail. And that is what we're seeing in those numbers and those statistics. So I think it does go back to what we are seeing as we go out and talk to a lot of these manufacturing companies.

15:40Alan Hinchman: It's about identification of a high-value use case. It's about getting the right people on board, getting the right people in the room, getting the right people to commit to what is that use case, and getting them to all understand who's responsible and who's accountable for the success of that. And then it's about being able to prove out that use case. Okay. Which, you know, we're not talking about, you know, we're going to, we're going to autonomize the entire plant or the entire facility or whatever. We're talking about a very specific use case. And then it's very much that crawl, walk, run type model, show the success, prove the success, and then it's scalable.

16:20Alan Hinchman: And then you can start to scale, repeat, scale, repeat. And that's how we're seeing how the leaders in manufacturing are being successful.

16:30Shane Kehoe: Well, the reason I think that you're seeing the high failure rate is a lot of folks are living in the art of the possible and not in the art of the practical. And it's everything that you said, Shane. When we're working with a lot of manufacturers, it's very easy. And we have one customer that famously launched 12 ,000 AI use cases inside of a year. They got to the 14-month mark and almost shut down the entire program because all of those use cases were just kind of these ideas of, well, we could do this. And no one asked, well, should we do this? Is it going to be impactful? Is it tied to a business case?

17:11Shane Kehoe: And change management, Shane hit the nail on the head there. You cannot have an AI opportunity that turns into an AI pattern recognition, which turns into something you're going to do with it without change management all the way across that. Because there's so many pieces that play in, especially on the factory floor, where there's so many pieces of the physical world that have to be a part of this, that those have to all be accounted for. Well, not only so many pieces that have to be accounted for, but so many pieces to your opening line of, you know, it could be represented from any of those past eras, whether it's 80s, 90s, whatever.

17:52Shane Kehoe: And even the people operating those pieces also might be from the 80s, 90s, whatever. And they have institutional knowledge that is directly related to that. So, I mean, that is a tall order from a change management, maybe even one of the more challenging use cases I've seen across any industry. How do you balance that, Alan? Well, the meantime between failure of a Rockwell PLC, which is the kind of the king of the world here, is over 20 years. So the reality is they're going to live past the eras. And especially when you're in a hyper change environment, you're going back to it. You're trying to pull information from things.

18:32Shane Kehoe: It's very common now in most industries. You're going to be working off of things that were installed in the 90s and early 2000s. And so, yeah, they're not ready for that. And you're 100 percent right. The practices, and it's kind of what I went back to before, you don't want to begin challenging how do we re-digitize this. It's, is the process even going to exist going forward? And should it exist in this world? And that's I think I think a lot of times folks are I want to do AI because someone on on the board said we need to have AI for the for Wall Street and we need to have a message around that.

19:12Shane Kehoe: But they're not thinking in the holistic view of, well, does that how does that process really change? How does how do we move into the next the next era? And I think that that's where it really happens because you're always you're always kind of fighting with the tools that you had in the last war. And so you're always thinking about it from this from this view. And now they're being challenged very, very quickly through autonomous vehicles, through autonomous thinking, through filling your favorite thing behind autonomous. And that's going to change a lot of that factory floor in ways that people haven't even dropped yet.

19:49Shane Kehoe: This episode is supported by NetScout. NetScout's AI-driven network performance management and observability platform provides end-to-end visibility across cloud and on-premises environments to ensure digital service reliability and detect cybersecurity anomalies. Protect and optimize your network with NetScout's intelligent assurance solutions. Well, Shane, I do want to ask an architecture question here because whether we're aiming for the art of the possible or for the art of the practical, this is going to require modernization on a certain level. And as we just went over, this technology, this infrastructure can be decades old.

20:27Shane Kehoe: And oh, by the way, nobody's giving you permission to allow for any downtime on the floor. So what types of architectural decisions do teams need to be making now? does leadership need to be making to start to enable? Maybe let's just start with art of the practical, knowing that art of the possible will come later.

20:46Alan Hinchman: So I'll even take one step just further back just to kind of really hone in on the scale of the challenge that manufacturers have. So if you're a global manufacturer, you could be sitting on, I don't know, say 100 different manufacturing facilities all around the world. and you know there's a high likelihood that a lot of those plants might have been acquired they're they're all you know they could have all different age profiles they could all be running different platforms there's typically no standard is standardization across manufacturing facilities from manufacturing platform perspective and if it is it's usually very regional and then to make it even more challenging the way that manufacturers have really always been aligned is the plant manager owns the keys to the kingdom.

21:40So, you know, that plant manager in charge of that plant

21:44Alan Hinchman: or those series of plants calls all the decisions that goes on in those plants. So if you have an initiative that's coming from corporate, if the plant manager decides he's not going to do it, he's not going to do it. And then, you know, what, what is it that, that manufacturers care about? What is it that those plant managers get paid on? They get paid on meeting production goals. They get paid on meeting quality goals. They get paid on, you know, reduction of downtime, reduction of waste and, you know, energy spent. And at the end of the day, that's, that's really all they care about. Oh, and sorry, worker safety is obviously very important.

22:21Alan Hinchman: That's at the end of the day, all they care about. So when you, when you start coming in with these kind of you know the vision of physical ai the power of ai what ai can drive across a global manufacturer it's it's you know obviously the board is paying attention what what alan was previously saying the board is paying attention the c-suite's paying attention and you know they're they're they're listening to all the marketing and they're they know that there's tremendous value in the plants. So what they're saying is, hey, how can we take these advancements in technology and in IT and AI and take that and put that into the plants and drive high value?

23:04Alan Hinchman: But then you've got that fundamental architecture challenge that I just laid out as to why it's so difficult. So really what we're seeing where people can start and Again, an interesting statistic is there's a tremendous amount of money going into manufacturing right now. But the vast majority of that money, I think the last time I looked, I think it's over 70 % of the spend that's going into manufacturing right now is going into infrastructure and security. It's not going into actual AI. It's going into the infrastructure. It's going into the security aspect. because, you know, what a lot of these companies are saying is, hey, we've been ignoring manufacturing for essentially decades.

23:53Alan Hinchman: So what we need to do is first and foremost, and where we're seeing a lot of that use case number one, security. How do we secure the plants? Okay. How do we prevent the plants from getting attacked? And manufacturers are really, really nervous about, you know, a lot of these frontier AI models and cyber attacks, all you have to do is turn on the news pretty much any week and you're going to see another manufacturer being attacked. Manufacturing is the biggest, OT and manufacturers is the biggest point of cyber attacks across any industry. Manufacturers are terrified about that. So a big spend and big use case number one is around OT security.

24:40Alan Hinchman: And then that bleeds into what does that look like from an infrastructure perspective? What does it look like from a foundational perspective? What does it look like from a compute perspective, networking perspective, storage perspective? So that's where we're seeing a lot of that initial spend going is around that infrastructure. Okay. Then what we're going to start to see is, okay, now once we get the security, the infrastructure, then it comes down to the data. And the data is really the, you know, that's where this whole conversation starts to come together is around the data. I'll just take a second and then I'll pull you into this conversation, Alan, because this is obviously very important.

25:24In a manufacturing environment, think about on a plant floor, you could have, you've got all of these different manufacturing platforms and systems.

25:36Alan Hinchman: You've got the PLCs, the programmable logic controllers, or the DCS systems. You've got the SCADA systems. You've got the historians, the MES, the maintenance management systems. You know, you've got all of these different systems. And then you start getting into energy management and building management and ERPs. Now, the challenge is that when you've got a plant that could be, you know, decades old, a lot of these platforms and systems have been put in over those 20, 30, 40 year period. So you could have PLCs that are 20 years old, SCADA systems that are 15 years old, MES systems that are 10 years old, maintenance management systems that are 15 years old, and they've all been put in by different integrators.

26:25Alan Hinchman: Okay. And again, it comes back to this classic manufacturing mentality of if it ain't broke, don't fix it. So if we're meeting our production goals and we've got all of these systems, then traditionally they're not going to be touched, which is A, a big security potential hazard, but also B, you've got to think about this now from getting into this AI and data perspective. So if you have a motor on a production line that has multiple different tag names and IO points and data points, all of those points across all of those platforms and all of those systems are going to be referenced differently.

27:09Alan Hinchman: OK, so how do you come in and put AI on top of all of these massive data silos, legacy data silos that all have different ways of referencing these data points? Because the AI isn't going to know. So the fundamental challenge for any manufacturer in delivering these larger scale AI solutions and take away the security, take away the culture and the change management is the data. you've got massive massive massive data silos that reside across all of these manufacturing facilities so what you have to be able to do is you have to be able to normalize that data contextualize that data put that data into a format into a fabric into an architecture that you can then apply ai onto that's the fundamental challenges yeah we're seeing Yeah, Alan, I'd love to get your thoughts on that, Alan.

28:15Shane Kehoe: To go back to something you said at the very beginning, which is around cybersecurity, because that's not a de minimis process here. This is it's a very critical one. If you think about the industry, it began with the idea that electricians would begin to move from relay logic into the digital world. And it's a really important thing because everything was based off of that. The code that's written ladder logic is actually written to reflect relay logic. When you look at how connectivity works, it's designed that everything connects to everything, that there's easy paths of information that flows.

28:57Shane Kehoe: So the entire factory floor IT infrastructure is zero security, easy to change, interchangeable, and varying levels of speed. And so when you ask why is it so easy to have a cyber attack, folks often think of a black guy living in his mom's basement with a hoodie. That's not really all of this. That's not the threat vectors. And it's not just some national situation of cybersecurity or cyberterrorism. It's as easy as kind of three threat vectors. One is making a change to an undocumented change to a program. And that is a massive issue that someone comes in in the middle of the night, makes a change, and the next day it has the code was dated.

29:55Shane Kehoe: It came from a different spot. It didn't have all the other changes. And so being able to manage that in volume is a real challenge. The second threat vector then is you have, in most cases, the most important pieces of your organization, the IP to make. If you're making, we often say you're handling the formula for Coke. One of the most closely guarded secrets in the world lives on these machines. So being able to access that information and being able to have that is a very keen interest. And you look across many multinationals, whether they were dealing in national security, whether they're dealing in a CPG product.

30:37Shane Kehoe: People want to know what are those formulas? How are those made? Whether that's Coca-Cola, that's paint, whether that's how we assemble something. So that lives there. So there is a threat vector that exists of people that that want to know that. And then the third threat vector is the threat of its own self, is that these changes can be pulled down. So when you look at AI, many plant managers, and I thought that Shane said it really well, they hold the keys to the kingdom. They're seeing a system that has the ability to make on the fly changes as a threat vector to what is really becoming coined as production resiliency.

31:22Shane Kehoe: So the idea that a tweak or a change in temperature in the oven as you're cooking something can change how much water vapor is left in kibble as you're making dog food or if you're making wallboard. So that change in an undocumented way also reduces their production resiliency. So when you think about those three vectors, that's what you're really, really working into. And so, of course, you've got to connect this. you want to understand that there's a lot of challenges around getting everything at least on the same timestamp. You can have, and Shane mentioned it before, it's very easy to have a pump that's got 30 different names and 30 different systems.

32:04Shane Kehoe: That's just how they're built. But there's also an underlying thread that these have got to be secured, they've got to be fortified, and they've got to be reliable. And I think that that all fits into that production resiliency kind of piece out of that.

32:19Alan Hinchman: I'll just add one more piece to this, because it just came up very recently and I'm starting to see this over and over. What's really interesting is in a lot of manufacturing customers, like just take what Alan just said about security and threats, but in a lot of manufacturing customers, they don't even know who owns OT security. Like it's as simple as that. So is it owned by IT? Is it owned by OT? You know, and again, the plant managers, like what we've been saying, own a lot of the keys to the kingdom. So they'll say OT security is under our charter. But IT are saying, no, it's not. It's under our charter.

32:59Alan Hinchman: And they've got very different responsibilities. You know, so who owns it? You know, and it could be as simple as a lot of companies are even struggling with that question.

33:10Shane Kehoe: Well, in it comes down to the battleground is there's a Windows 95 system on a press in the back that's isolated. Does that piece of equipment have to be upgraded, and is that upgraded supported by the software? And that's where the IT and the OT become no T. Nobody wants to do anything. Nobody wants to touch it because neither of them want to take the responsibility for this antiquated piece of equipment going down. And we have done a number of projects where things that were never supposed to be on the network now have to be networked. And you really run into those challenges because what you're going to find out is somewhere there's a machining unit in the back of this that's going to tell you, here's the tolerances that I made this specific part to.

34:04Shane Kehoe: You're going to want those and it's not going to talk to you. And who owns that? Yeah, I mean, so the challenges are... Plenty. So far, we've covered legacy architecture, acquired architecture. We've talked about security in many different forms. We've talked about culture and change management. All of that is funneling into the person, the keys to the kingdom, the plant manager. Alan, you said something that I flagged when listening to you on a recent podcast. You said at the end of the day, somebody, in this case, like the plant manager, has to have the intestinal fortitude to actually mash the button and turn it on.

34:41Shane Kehoe: Now, I think you were talking about AI-generated programming in that instance, but I think it applies here. What type of confidence or what type of trust do you have to have in these systems? What would you need to see to be able to mash that button when you know on the other end the potential risk is your, to your point, giving up the recipe to Coke. Yeah, and we talk about it quite a bit, is depending on the industry you're working in, and many of them that Gray Matter and WWT work in, there's a real clear and present danger to badly written code. If you think about the closer that you get to Houston and the ship channel, the more serious that becomes where now it's not just, hey, I put something into into AI, generated a few thousand lines of code and it just doesn't look right to me to I just blew up a cracker in a petroleum plant.

35:43Shane Kehoe: Like that's that's where for that comment comes from. So AI generated code is an incredible tool, an incredible time saver. It's going to change the way things are priced, the way work is done. But you also still have to retain someone that signs off whether this is happening. And it's the same thing. What we're going to see in this space, if we begin to go three, four years, five years into the future, what we're going to see is less hard-written code and more situationally generated AI. Here's what we're going to go do. And when you look at, you know, NVIDIA and some of those, that's where that's where they're beginning to say this is where the space is going to go.

36:26Shane Kehoe: You're not going to have hard written control code. What you're going to have is a device that's got some constraints. And here's how we go make this inside of that. And it's constantly evolving. And so not to be a futurist on this, but someone is going to have to have the courage to to mash the button. But also that becomes how far do you get separated from knowing that this thing is right? And that's the question that we've been that we've been pondering in our organization and looking at how do you hire? How do you do the change management? How do you have safeguards? How do you not continue to build things?

37:04Shane Kehoe: I think the Russians learned a valuable lesson on how to how to not make the same mistakes after they blew up Chernobyl. And it's a great example that was built off of a compiling of mistakes that didn't that folks in their design phase didn't know were mistakes. They were building off of the prior generations and eventually had a fatal flaw. And I'm not saying that everything goes to that level, but I am saying there is opportunity for that. And that's why when you're thinking through these projects, that's why I think there's a huge opportunity for the folks that live inside of our walls for WWT and for Gray Matter to be some of those folks to say, look, we've seen this a few times.

37:46Shane Kehoe: Let's make sure that you've got the right resources, that there's a framework on how you're going to approach this. It seems like those would be the obvious things, but they're not obvious. I tell everyone there's a reason why there is a don't put your hand in the wood chipper sticker. it's because the obvious doesn't always get to be said. And I think for us, that's what's going to be the future of our businesses, is shoulder to shoulder, hey, we can help you understand exactly how these far more complex systems than they've ever dealt with are going to be. Shane, I'd be interested in your thoughts on that too.

38:19Alan Hinchman: The opportunities in manufacturing are off the charts because, like I've said multiple times, manufacturing just hasn't been getting the investments that it should be getting because manufacturing at the end of the day produces the product that keeps the company in business. So the opportunities are tremendous. But it is very much that full-walk run type model that we talked before. It's about identification of a use case. It's identification of a high-value use case. And it's how do we solve for that use case in a fashion that brings everybody on board in a highly secure, highly governed way and shows an immediate path to value.

38:55Alan Hinchman: You get into that journey, you start that flight wheel, you get the next use case, you get the next use case, you get the next use case, and it becomes very, very scalable and repeatable. And then you've got something that you can take out to market. People start to understand it. All of these manufacturers are looking for help. Every manufacturer I talk to is looking for guidance. They all are getting tremendous, tremendous pressure from their internal C-suites, their internal boards on what they're going to do around AI, and they don't even know where to start. They don't even know how to have the conversations.

39:27Alan Hinchman: They don't even know who to put in the room. Like we just, I just had a workshop a couple of weeks ago with a big manufacturing customer. And it was the first time the IT side of the org had ever been in a room with the OT side of the org. And they actually came out of that meeting at the end of the day and said, we should do this more often. Manufacturing is the wild west, but that's the opportunity and that's the challenge.

39:51Shane Kehoe: Yeah, well, certainly a lot of work ahead, but hopefully some of the practical insights that Shane, both you and Alan shared helped put us on a path towards realizing, you know, the next great era of industrial manufacturing. So Shane, Alan, thank you so much for taking the time to join us on the AI Proving Ground podcast today. It was super enlightening. Enjoyed the discussion and hopefully we'll have the two of you on again soon. Okay. Thanks to Shane and Alan for joining us today. the lesson here, AI earns more responsibility by proving itself in the operation. Start with a problem worth solving, set clear limits, and expand when the evidence supports it.

40:28Shane Kehoe: This episode of the AI Proving Ground podcast was co-produced by Naz Baker and Kara Kuhn. Our audio video engineer is John Knobloch. My name is Brian Felt. Thanks for listening. We'll see you next time.