Will AI Make Its Biggest Splash In Industrial Use Cases? — With Mark Moffat, IFS
Channel: Alex Kantrowitz
Published at: 2025-11-25
YouTube video id: EWoJSEqQPjI
Source: https://www.youtube.com/watch?v=EWoJSEqQPjI
Could AI make the biggest splash [music] not in the corporate office, but in industry? You may be surprised. Let's talk about it with Mark Moffett, the CEO of IFS, in a conversation brought to you [music] by IFS, and Mark is here with us today. Mark, great to see you. Welcome to the show. Thank you for being here. It's great to spend some time with you, and I can't wait to get into it. So, we talk about AI a lot, a lot, a lot. And >> lot, a lot, a lot. the focus is always on the corporate office. >> Yeah. Then, there's 70% of workers that are not in the corporate office. >> 70% of work is done outside of it. In industrial settings. And up until this point, I haven't heard a word about how generative AI could be used to change their work, to help them, to make these processes better. But, you have uh belief that it's actually going to be more impactful in industry than in the office. Yeah, well, if you think about it based on what we've just discussed, 70% of the world's workforce are not behind a desk, running industrial operations, running the very industries that support economic development and prosperity. You think about the nature of the industries IFS support, construction, engineering, manufacturing, uh aerospace, defense, telecommunications, energy, natural resources. These are all fundamental to driving industrial progress and economic development. And most of these organizations have workforces that are in the field, in the operation day-to-day. So, it stands to reason you don't get the full benefit of AI until you enable that workforce. But, hold on, because let's say I'm a lawyer, right? Like, hypothetically. I'm not one in real life, but in this example, I >> I'm an accountant by background, by the way. >> Okay. All right, so we can even talk about the accounting example. >> You can take reams of data and drop it in chat GPT, and then in natural language, you can query it, and it will give you answers today. Um, but if you're working on the line in a factory, or if you're managing operations uh in some industrial settings, it's hard for me to fully picture how generative AI can then be applicable in that setting. >> 100%. So, how do you do it? 100%. Well, think about an example that we're working on, real example, real use case we're working on, in development, with Boston Dynamics, with Eversource Energy, and with Anthropic, okay? Three parties orchestrated as one to make manhole duct inspections operate at a different level, okay? So, what happens in practice? We send a Boston Dynamics Spot robotic a long a 5-km manhole duct. The Eversource are required by Boston law, in Boston law, Massachusetts, that by law they have to inspect that manhole duct on a on a periodic basis, okay? Now, that robotic dog is picking up lidar, picking up video, image, gas sensors, heat, temperature, pressure, all the way through that manhole duct, okay? They're spotting issues and fractures and and problems in that environment that the human often will miss. Or, they might not get with the same level of accuracy. Something spotted a stress fracture on a transformer, immediately captured, GPS coordinates, immediately triggering a work order, looking for the spare part, dispatching a crew. That's all happening pretty much instantaneously. Now, you think about the alternative, when the humans do that, often they don't want to do that work, it's difficult to resource it, and they might not spot that, and a catastrophic failure might arise, and therefore you see issues in uptime, you see issues in transmission, all sorts of problems. So, the AI in that question, obviously, transforms the operation. By the way, this is why I'm so happy to be speaking with you, because these are use cases, again, I spoke about it in the intro, we don't hear about. But, we're about 3 months after the introduction of chat GPT, and we're starting to see these applications of how natural language and LLMs can be used to take data that has been seen previously, and actually make it useful. And we talk all the time about the ROI of AI, and this seems to be a place where you can actually have an ROI. So, just talk through the tech stack that you would have in the example that you discussed. So, you have, obviously, the robot Spot, which looks like a dog, it's making its way, you know, through this environment, yeah, and then is there machine learning with some uh visual intelligence picking up on the different things that it sees, and then it sends it into an LLM that makes decisions? Correct. Uh also orchestrated by thousands of workflows and data that IFS has transacted for other utility companies. Knowing what parts to deploy, knowing what supply chain to activate. But, the LLM, you know, a generic LLM is not going to discover what's needed in that environment. It needs specific training and adaptation to that environment in question. So, with the application of Spot the dog into the entirety of Eversource's operations, together, hopefully, with other utility companies as customers in North America, together with any OEM information we might have, or any other data that gives us a sense of patterns and repeat probability, you put all of that together in a specialized set of models, then the ability for you to, you know, deliver an increased ROI is just off the charts different. I want to take a moment to talk a little bit about IFS, because, of course, there's a lot of moving parts here, right? There's the robot dog, there's the Anthropic large language models, but there has to be some sort of system in the middle to route these signals to the people on the line. So, talk a little bit about the role that IFS plays here. Yeah, what I talked about this morning at the event that we're hosting today, which we've called Industrial X uh unleashed. Industrial X unleashed is all about what I describe as bringing the dimensions of the X, you know, the four dimensions or or engines, as I call them, of progress and possibility. They all need to come to bear. You got the models, you have the infrastructure and the data, you have the robotics, and then you got the reinvention partners. We often work with partners, advisory firms, top-tier consultants, because when you're changing fundamentally operations and teams and skills and capabilities, with all of that coming together, you have to move an organization from A to B. It doesn't happen by magic overnight. There's resistance to change, etc., okay? IFS has been supporting the industries we serve for decades. The entire organization is geared with understanding the intimacy of every industry we serve. And we've been developing workflow for field engineers, for asset maintenance, for ERP in these industries for over 40 years. So, we've got know-how built in, we know how to orchestrate value chains. So, IFS is sitting in the middle of those four engines of progress and possibility. Our platform, the application stack, it doesn't go away. Because ordinary workers in the field, they still need coherence. They still need to understand how they're expected to execute their tasks on a day-to-day basis. We're not asking engineers who, in another example, are but, you know, telephone tower or a power tower to use LLMs on the fly. They need to be using and digesting and consuming AI in the natural flow of their work. That's IFS's role. Okay, and so I want to talk to you about the um enhancement that large language models have played here. Right? Because IFS in the past, and correct me if I'm wrong, >> had been working on predicting when there was going to be something that had gone wrong. Yeah. Um I don't know, is the Boston Dynamics a new thing, or have you you're always working with these robot dogs? New thing. Uh new thing. We've we've looked at this market evolving, and we've recognized we can't do this by ourselves. That would be arrogant in the extreme. And we recognize there's so much innovation happening in the field of robotics, with all the capital that's flowing out there and the expertise that's available there. And particularly Boston Dynamics, because Boston Dynamics, they're not developing humanoids, or they are now, but they've come from a world where they've developed different form factors for the industry challenges. And Spot is an incarnation of that, okay? So, we recognize we have to work with those industry participants if we're going to be able to access the level of innovation that we believe our customers need. So, okay, so now you have this new data point as an input. It goes into your system, and the to me, the interesting thing is, without an LLM, it maybe would just sit there. But, what you can do now is be proactive to people out in the field, and get them to work on things fairly quickly now that the data points can be translated into natural language and conveyed to people that are there working. It's a really good point, because what we see from an LLM perspective and a natural language perspective is we've just introduced a completely new engagement layer for our technicians. So, they can use voice, they can use video, they can use photography to engage with, you know, all the intelligence that sits in the platform that draws in all these other capabilities to do the job at hand. So, if an aircraft engineer, to use another example, is faced with something they haven't seen on a jet engine as they're stripping back the jet engine to make a repair that's required by the maintenance that's set by the OEM, then they're able to use natural language to engage with all the intelligence that's built up to help them in the task at hand. And if you take it one step further, we're exploring with the use of Nvidia Jetson technology in the Metropolis platform to give augmented reality to those technicians, so that when the technician is performing a task, you know, with that edge-based chip and an LLM that can go on to that chip, still with a billion of parameters, they're able to get real-time guidance on the torque setting for a part of an engine. It's just mind-blowing in terms of possibility. Now, I want to talk to you about Anthropic, right? Because Anthropic is a big partner. We're here again at this Industrial X Unleashed event that you've held here in New York City. Uh New York City, great city to hold events in. I agree. And um Anthropic is providing an important layer right here with with the LLM. So, talk a little bit about how they fit into this picture. Uh you know, we're always looking uh how innovation is happening in the market. And of course, we've been tracking quite carefully the frontier models. And we were super attracted I was super attracted in listening to how Dario and the other co-founders of Anthropic have set their strategic position out. They're clearly focused on the enterprise. I think that's less of a focus than some of the other frontier models based on our experience and based on what we've seen. I'm sure it's got some relevance, but you know, Dario and the team seem to be very focused on the enterprise and they see the potential to unlock value in the enterprise. But they're also focused, I think, uniquely on the balance of risk and possibility. And I think that's so important with industries that we serve. You can't afford to get things wrong. You need 99.99999% reliability on the work execution. Otherwise, supply chains grind to halt, aircraft are grounded, or ultimately, there's risk to life. You know, that's the type of operationals that we're dealing with. Mission-critical operations. So, them understanding there's risk in the broadest context as well from a societal perspective is really encouraging. And I think that's different based on others that we see. So, those two components, I think, taken together made Anthropic a very obvious partner for us. And and we approached them, and very rapidly we got into conversation about introducing kind of world's first partnership for someone like IFS in the space of the industrial sector to partner to bring that capability to bear together. So, how much trust do you put into them? Because Huge amount. Huge amount. But there's varying levels of trust, right? The first bit of trust is I'm going to trust them to take a signal that I'm seeing in the data. Like going back to this robot example, there's a signal. I'm going to send it to someone. I'm going to trust them to get that signal accurately to the people I'm working with. The other level of trust is I'm going to trust the large language model to actually make decisions uh that would previously be made made by the company. Where do you stand? I mean, we're not relying blindly. I mean, these things don't get into production without regular stress testing Mhm. in all sorts of ways. Uh one of the ways in which we think about stress testing the capability that ultimately makes decisions in the field is running millions of scenarios. Because the more scenarios you run and the more times you run queries, the higher probability you are getting that accurate. So, the availability of compute allows that to happen. So, I think we recognize Anthropic recognize, you know, the job to be done, the nature of the operations, they are mission-critical in nature. So, we will find ways together to get the level of trust we jointly need because our brands on the line, Anthropic's brand will be on the line, and the nature of what we do is so significant. So, it's not blind trust, but the nature of the partnership and and and how we are as organizations I think creates that environment of trust to develop in that way. So, you will have to make choices eventually. But what ultimately the technology will have to make the choices. But the other thing that I would say is that we are making sure that we balance any agentic-based capability with a human in the loop. So, where are the checkpoints? What are the escalation points? What are the break points in a process that would still require you to have some level of human intervention? I don't see that going away. Where we draw that line, I mean, yet we've to discover. And uh you've actually already seen some pretty impressive results with them. Um yeah, as someone who's from Scotland, I think this is a example close to your heart. You were speaking with a Scotch company or some distillery in Grants, yeah. That um was actually only uh encountering most of their issues in an emergency setting. But with some more proactive notifications and translation into natural language, they've been able to uh get ahead of some of the problems that they would have seen previously. Can you expand upon that a little bit? Yeah, no, 100%. They invested hundreds of millions into new The team told a chemicals manufacturing facility. I call it a distillery. They were trying to obviously disguise that they were working with uh and and and Hendrick's when they were complaining about how much time they were spending on site in in the facility. I I only joke, as you know. Uh but, you know, they were experiencing a lot of problems in that facility. They'd invested a huge amount of money. Uh and we began a discussion with them, and that led to the development of a new uh field capability where we were able to analyze equipment and failed points uh and issues uh using the Anthropic model to index uh you know, the issues at hand that have resulted in a more proactive maintenance uh set capabilities. So, how would that work in practice? Like, can you help flush it out a little bit for someone trying to think about what this would look like on the ground? Yeah, so somewhere in the distilling process, if there was uh temperature reading or a valve or a throughput that wasn't reading accurately, uh the team were able to get on top of that super quick through the telemetry. They're able to use video and photography uh to look at the issue at hand very quick, and able to put that into the context of the engineering drawings and diagrams to be able to diagnose the end-to-end and what was happening through effectively the value chain of creating the spirit at hand. That's happening with large language model. >> Happening right now today. With an LLM. Correct. I keep asking because to me it's just like when we think about this technology, we never think And I keep going back to this, we never think about it for these applications. >> certainly don't think about it for making whiskey, that's for sure. No. And an LLM I think well, I mean, part of the issue is LLM large language model. You know, it's got It goes way beyond language now. It goes beyond all sorts of data. So, I think the very description of LLM needs to change. In fact, I think now we're talking about world models. I think world models are a more accurate description of what we're talking about. >> Right. And the inputs that you're getting, the video, the photos, that all feeds into these models' perceptions of the world that you're placing them into. >> think about the wide modality of data you can take on board from a manufacturing facility, this being one example. Like flow rates, temperature, gas, vibrations, uh the throughput measure. There's so many ways in which you can take and catalog and ingest data. The hard bit is about how you put it all together and how you make some meaning of it. Now, let's talk a little bit about labor here. Because these questions always come up when, for instance, you have a Boston Dynamics robot like I saw uh today walking around the factory floor looking for spills uh and probably very efficiently through computer vision making a notice that, "Hey, there's a spill." and maybe saving, you know, a lot of time or potential like uh batch of product. Uh but previously, maybe that was something that a person did. So, how do you think that this will impact labor moving forward? So, I'm an optimist. Uh and I read with great interest and agreement the all we're going to experience is growth. Uh economic growth for sure. The predictions are 1% percentage point increase to global GDP growth in the coming years. Uh and I look at, you know, research like the World Economic Forum who did the future work study at the beginning of this year, and they concluded that 170 million new jobs will be created by 2030. Okay? Uh 92 million of existing jobs will be displaced. My mathematics tell me that's 7 to 8 million of net new jobs, incremental jobs and employment by 2030. Okay? And what we're experiencing is, in all senses, economic growth. So, you think about growth, you think about some of that research, and you think about some of the previous general-purpose technology shifts that we've been through, albeit this one is different. All of them have resulted in growth and more labor and more employment and more business models, and I think the same is absolutely true here. When I put that into the real world, so put that to one side for a second, every customer that we deal with, whether it's in North America, whether it's in Europe, whether it's in Asia, in the industries that we serve to a greater or lesser extent, are dealing with labor shortage today. And that's driven by aging workforces. It's also driven by reindustrialization of Western economies. It's driven by global supply chains moving. You know, the reindustrialization in the United States, manufacturing coming into this country. You know, these plants need to be built, these data centers need to be built, transmission grid networks need to be improved. I mean, you look right across the footprint of the industries we serve, it's growth, it's demand on capacity. The only way we're going to be able to deal with that is by using digital workers and by robotics, in my view, as well as increased growth of labor. It's almost as if the Japan example is playing out for everybody. Like, Japan has been the one that's been way ahead of this because it has an aging workforce, it has a shortage of high-skilled labor, and they have really tried hard uh to automate, and they're far ahead of everyone, I think, on automation and robotics. They love robots over there. >> They do. They're good at it. >> is that a preview for what's going to happen with us? I haven't studied it in detail, but directionally, from what I understand, I think so. And and I think again, we come back to an environment where I genuinely believe there's going to be employment growth over the long term. The nature of the roles will fundamentally change. No question. And I think, you know, that's a responsibility that falls onto, you know, governments, uh higher education to be thinking about what's the shape of the labor force to come, and how do what does one plan for it. The nature of the jobs will be different. And there's no question, World Economic Forum said 92 million jobs will be displaced. Uh and those individuals in those roles will need to be thinking about how they enhance their skills and how they develop to take on new roles. Right. Um and going back to this question about high-skilled labor, I mean, that's a theme that comes up all around the world is that there's more work to be done by high-skilled laborers than there are high-skilled laborers. Agreed. >> And so, is this where this could sort of step up, fill that productivity gap, the company's able to make more, deliver what the market actually needs, and that's where you see the growth. Yeah, and if you make it super practical and very right down to basics, an aging workforce is you've got individuals who've got 40 years worth of what I call fingertip experience. You know, they've just got innate knowledge that in some cases you can't really develop over 40 years. So, how do you how do you ingest that in modern technology? I mean, through super basic means. I mean, you interview these people and you know, for days on end. You follow them in their day-to-day work. You have cameras on them. You have an ability to ingest all of the reports, all the write-ups, all the work they've ever done. You've got to suck as much knowledge as you can into LLMs, into models. And I think if you do that, depending on the nature of the industry, with any other data you can get from OEMs, etc., then you are finding the way to create models that are very specific to that role to be done. And, you know, I often talk about doing the hard yards and doing the hard work and getting down to first principle thinking, and that's often what that requires. Okay, but there is this quote. I have to bring it up because I did hear it here today. Uh the quote is, "In the factory of the future, you're going to need a man and a dog. The man will feed the dog and the dog will make sure nobody touches the machines." I didn't hear that. I wasn't in the room for that one. I'm assuming that was in the Boston Dynamics session. Right. Yes, it was. So, do uh do you think that that's the future or is that kind of a joke? No, I think in some cases it absolutely is. Uh I I interviewed I was on stage with Muhammad Candy from PwC, the global chairman of PwC, this morning and he talked about a personal experience he'd had in South Korea recently where he went to a factory to observe the manufacturing process and I don't know what company it was or what line of work it was, but literally as they arrived, they turned the lights on, he could see the whole operation automated with robots, with drones, with automated production lines. So, it's categorically going to be the case in many uh industries and that would be an example perhaps of the 92 million that I talk about. Right. >> but it's not going to be true of every single uh industry. So, as somebody who works very close to industry, uh you obviously are keenly aware of the need for power and how much these large language models are planning to take out of the ecosystem. What do you think the future is going to be on that front? Well, I think the future is going to be well, on two fronts. Uh obviously increased power generation is one way to solve the energy crunch. The other way to solve the energy crunch is by getting more of what already exists. Mhm. And I don't know if you were listening to Sabine from Siemens, uh the CEO of grid software at Siemens, but the stat that caught my eye was that there's 150 billion uh output a year that's lost in the US as a result of outages, right? So, if we deal with those outages, and I don't know what the power consumption issues are there, but if we deal with those outages with those types of numbers, something's telling me that we've got more energy available. So, how do we look at existing infrastructure and squeeze more out of the existing infrastructure? How do we avoid downtime? How can we get the throughput increased on existing infrastructure? And then overlay that with new power generation, and there's all sorts of ways clearly you can do that. Uh you know, introduction of new nuclear technologies is obviously one that tends to be more carbon efficient. There's clearly carbon intense hydrocarbons that can be used. Uh there's a proliferation of different things. Uh I also feel the renewables needs more attention. Some of the technological developments with solar and particularly predominantly coming out of China make the cost of generating wattage for that technology is significantly lower than it's ever been. So, it's going to take a very wide array of things to deal with the energy crunch. Yeah, it's almost I was speaking with someone recently, this is completely different, but talking about the brain computer interface, and the brain computer interface in a way can be a parallel track that helps alongside medicine. Like you have the typical medicine that we've had with doctors, and the brain computer interface can be something that helps um for instance, if you're losing sight, you have two options. Get a contact lens or glasses, or um maybe there'll be a future where you can have a small probe put in and it will deliver that information to your visual cortex and you'll be able to see. And I it's a weird, but maybe there's some parallel here where in industry, you can either make more power Mhm. the traditional way or you can optimize. You can use technology in out-of-the-box ways, ways that haven't been thought about until recently, and make everything in industry more efficient, and there you go. Yeah, multimodal. Uh I mean, everything. Pushing everything all times to solve problems. I think it's the way that I would uh think about that. What you begin to describe there just blows my mind. I mean, you can bet I'll be the first in line when that happens. Yeah. It it's already in production, but I think it's a long way until it will just be easy to pop it in. >> Yeah, I might wait until a few more people have tried it. >> too. I mean, I did meet the first Neuralink patient, Nolan Arbaugh, Wow. >> and we played a video game against each other. Yeah. >> Uh he's quadriplegic, so he can't use a computer with his hands, uh but he was able to think on the screen and the cursor would move and click. Wow. >> And he beat me in that game. And one interesting thing that he said >> Wow. um was that he can uh uh think about mo- motion and the cursor will already be there. So, there the his prediction is there'll be a league of in video games of people who that actually have this ability because people who have to actually move the thing with their fingers will never be able to do Do you think you're beginning to predict the IO device that's coming from OpenAI? I It look, if OpenAI eventually wants to plug into our brains, um that might be a time where I'm saying that they actually are too big. >> [laughter] >> I agree. Okay, um but but one thing they do talk about is AGI. Mhm. Um do you think that we're anywhere close to that? And does it even matter honestly, because people think to get real economic value out of AI, you need to have AGI. Do you believe that? I mean, I think for the for the immediate opportunity in front of the customers that we see, I mean, AGI isn't a necessary step right now. I mean, actually, let's get to the practical reality. I was with a customer on Monday earlier this week, a commercial aviation customer, and they were talking about what they got today in their technology stack. And they've got IFS in the core helping them with their maintenance, their planning, their repair, their overhaul. And then they've built up like 80 boundary applications, organically built, all serving mission-critical purpose, all dealing with regulatory requirements with the FAA and others, okay? And that, you know, core technology isn't yet on the cloud. Okay? >> Wow. So, you know, and and that's a representation of many customers that I meet. I met another customer this week. Uh I met the CEO, the CIO, and the COO of a ports terminal business. Very similar picture. Still got some of their estate on prem, right? That's the hard nose reality where a lot of companies are today. So, to get them from legacy application architecture to cloud to enable the availability of AI, and then to go to age AGI is like there. Yeah. So, I think the reality is whilst it's super exciting and it's got all sorts of positive implications for mankind, for businesses, you know, a lot of businesses that we encounter, they've got some basic stuff to get right first. So, when we talk about just let's end where we started, which is ROI on AI. When we talk about AI um and the ROI of this technology, a lot of people have wondered, well, like is AI going to have to replace all coders for there to be a justification for the massive valuations we're seeing. And when I hear you talk about the way that AI can be applied and deployed in these industrial settings, I'm like, wait a second. This might be the place where the ROI comes. If it makes factories and people working in industry much more efficient, that's invaluable. I I I couldn't agree more. Uh and the reality is that most frontier models right now, I think, based on what I can deduce, are making most of the revenue from consumer applications. With the exception of Anthropic, who are distinctive, I think, and focused on the enterprise, and I think that's one of the reasons we're working with Anthropic. All right, Mark, thanks so much. Great speaking with you and I appreciate you opening my mind [music] to this new application of the technology. Well, thank you for entertaining me and engaging and being here at [music] Industrial X. My pleasure.