Fireside chat · 10:15 AM · Thursday, May 14, 2026
As Rohan and I were talking about, it starts with the leaders. I love the comment about getting the board on board. So if your board isn't able to help you with AI, I would say if you're an executive, if you're part of a board, if you're advising your executives, your board needs to be in tune with what's going on with AI. I think that was something that really jumped out at me in that conversation. The next conversation starts taking us down that industry path. Dana, my co-host from yesterday, and Sankar are going to talk about something I think is very interesting. We don't talk about it enough is how all this data that's being created by IoT, OT, in industry is really reshaping things and how we need to dive into that. So with that, Dana, I'll hand the stage over.
Thank you. So I want to introduce Sankar, who's, I love the title as well, so Global Head of Digital PLM, Smart Manufacturing and Ops and Industrial Metaverse Practices, and I think that we could probably add a few more titles in there as well around the technology that you cover, but our fireside is actually called Industry 5.0 Meets Physical AI, so I'm missing them on the slide deck, but that's an interesting concept overall because Saurabh just made a comment earlier around how we're, you know, we're talking about how to AI, but next year it might be how-to physical AI. So I want to start with a question by asking you, so over the last few years we've been talking about the shift from co-pilots and to, you know, agentic AI, and now we're also talking about physical AI. First of all, could you define that for us? And secondly, what feels fundamentally different about this moment in time?
Absolutely. First of all, thanks for having me, Dana. Good morning, everyone. Now, when it comes to really physical AI, right, so we have been talking about AI since yesterday, but AI and ML have been here for decades, right? So, however, if you look at really the last few years, we could say the, you know, the introduction of ChatGPT sort of brought AI to a different dimension, right? So, in terms of the evolution we are seeing from an underlying infrastructure, compute, et cetera, and then the model and the performance of model and the energy consumption, et cetera, there's a big leap going on here. But then that said, if you really take a step back, look at in terms of broader application areas of AI, we could really probably categorize this into three broader areas, right? So number one is, how do you apply AI to help empower your employees and drive more productivity, right? So which is really where what we have been talking about in terms of generative AI, you know, with content creation or consumption, synthesis, so on and so forth. That's really, you know, and even code generation and so on and so forth. That's the application area one. But then when you really look at the area number two, truly from a product standpoint, how do you make your products more intelligent, cognitive, and how do you make your operations, you know, more cognitive, intelligent, right? So these are the two broader areas where AI takes this dimension of quote-unquote physical AI, right? So we are talking about, you know, for example, you know, whether it is automotive products like cars or trains, you know, mobility products like cars, trains, or medical devices or aircrafts. What if they become really more cognitive, intelligent, and sort of self-recalibrating, for the lack of a better term? On the other hand, you know, when it comes to within the four walls of your factories or warehouses or, you know, distribution centers, what if the operations are becoming more smarter, cognitive, intelligent, adaptive? That's really where, quote unquote, the physical AI aspect comes. Now, what has changed, right? So I think what has changed is, broadly speaking, one is really the core foundational technology dimension, right, so when it comes to compute, models, so on and so forth, there has been rapid evolution there, right? Now, the second dimension is, with what's happening broadly from a global supply chain disruptions and geopolitical tensions and so on and so forth, are putting really pressure on manufacturers in terms of how they can really have a resilient operating model, sustainable operating model. Particularly manufacturers in Europe and Americas are really impacted heavily. 77% of manufacturers have clearly called out labor shortage as the most important challenge that they are dealing with. Now, one hand, you have the foundational technologies evolving rapidly. On the other hand, there are tremendous business imperatives for companies to really adopt technologies. That's really the sweet spot where we are with respect to how AI could take a different dimension in the physical world to drive the next level of innovation, operating model transformation, and huge productivity.
That's fascinating. And we were talking about this earlier as well, but the use of operating models being used kind of a lot of times in the last two days as well, how does that look like inside of, you know, factories? We talk about Industry 5.0 now. What does that look like compared to Industry 3.0, and what is the operating model shift that's happening there?
Yeah, I think that's a great question, right? It's one of the most overused and sometimes misused word, right? So operating model, and then everybody talks about operating model transformation. Now let's take a step back, right? So when it comes to industrial revolution, right, so industry 3.0 kind of focused on automation, right? And then 4.0 focused on more connectivity, right? So how could you really bring connectivity to what you do things and then sort of drive next level of automation? We are talking about industry 5.0, and 5.0 here, right? So what we mean by industry 5.0 is how could you bring physical AI to sort of apply not only from a product's dimensions but from your factories, assets, and operations dimension with seamless human machine integration to drive next level of innovation and business performance. Now, this physical AI and an industry 4.0 to 5.0 is sort of truly a catalyst when it comes to operating model transformation. I'm using this word deliberately here, right? So, for example, right, so what if companies like, you know, we are already seeing this in automotive industry with what Tesla is doing, right? So it's no longer really, you know, sort of point A to point B, but how could you, you know, really, you know, bring AI, physical AI in action when it comes to what Tesla is already doing? Physical AI today is already in action with respect to what Hitachi is doing in rail industry, right? So the rail cars that we make, for example, now have about 1,200 sensors and kind of bring service aspect as we sell it to owner-operators. Thereby, we are pivoted from sort of providing rail cars and signaling systems to providing mobility as a service. You see that with industrial equipment providers like Caterpillar of the World as well. So it's truly a catalyst when it comes to operating model innovation.
What can other firms learn from this specific context as well as the industry that they can take to their own? So as they're beginning to think about operating model changes, what are the lessons that they can take away?
See, I think with, you know, we are seeing, right, so in the stock price today, companies in software space are going through, I think, turbulence time, right, so for the lack of better term. Not necessarily all validated personally, you know, in my view, but if you look at really what's happening from a stock market standpoint, the data point is indicating differently, right? But there is going to be truly a disruption there with respect to what AI is doing, right? So the way we are going to approach software engineering is thoroughly changing right in front of us. In similar fashion, when it comes to product engineering, manufacturing, and deployment of those products in industrial setup, that is changing rapidly, right? So whether we like it or not, that is changing. That's going to continue to change. So obviously, you know, companies that are into making products, physical products or providing services, need to take a step back, look at it in terms of what this means for their, you know, for their core foundational vision mission and what they are trying to do from a business standpoint and how they could really potentially use this as a pivot to sort of move to next level, right? So that's really how I look at this. And I see a lot of our customers as well, you know, are approaching this way of how they could really use this technology shift to drive some foundational operating model innovation. But it's going to take time, right? So it's going to take time because the reality is a lot of companies are already struggling with what they are doing. Some of these aspects require capital investment as well. So that becomes even more problematic, right? So because when you talk about really factories becoming smarter, when you're commissioning a brand new factory, the story is different. But we are dealing with a lot of factories that were constructed, commissioned long back. How are you going to modernize that? The cost of modernizing versus perhaps cost of commissioning from scratch, there is a dialogue to be made, you know, discussion to be made there.
We've been talking a lot about process debt over the last two days, and I like Manish Sharma's kind of comment too around these mega processes that need to be a new result before you begin to address them internally. In this world, in your world especially, organizations are still layering on AI into fragmented processes overall. Are they kind of underestimating what needs to kind of happen to, you know, move to the next level? And what is the impact of layering AI in more of these broken processes?
I think that's a great question, right? So if you go to, say, ChatGPTs of the world, and if there is hallucination, the impact of that versus you are attempting to apply AI at scale in your factories to make things, you know, if that goes wrong, imagine the impact, right? So not only from a human safety standpoint, but from your operational performance and so on, and customer commitments and so on and so forth. So this is an environment where, while experimentation is super important, you need to really do that in a very controlled, predictable manner, right? So the reality today is, are organizations ready to embrace this? The fact is, we all know, right? So they are not. I think in my view, it is okay, right? So there has to be always a starting point for something. But how do you really quickly, you know, assess in terms of where you are and what you need to meaningfully do to be ready for embracing these transformations? I think that is going to be super important, right? So I always recommend, you know, doing two things in parallel. Number one is companies need to comprehend in terms of what this means for them, right? So this technology shift and the opportunity potential because of which, you know, because of the technology shift, what this means for them, I think it is super important to comprehend that, which means it is okay to go on AI pilots, but now we are seeing pilots don't scale because one thing companies are not doing in parallel, which is as you are trying to figure out what AI could do for you and trying to pilot and comprehend, how could you have a parallel path to be able to rapidly assess from your people, process, data, and underlying technology standpoint, how ready are you to embrace this? So particularly when it comes to physical AI and operational setup, whether it is on the product side or factory side, asset side or operation side, we are talking about convergence of a few things, right? So one is, you know, as defined as designed information to as planned, to as manufactured, to as sold, and then as commissioned, to as serviced. So, you know, these pieces of information need to seamlessly come together in an authoritative fashion to be able to sort of employ AI at scale, whether it is at your product level or operation level or from a supply chain or warehouse or logistics level. Now, are you ready from a, you know, underlying data infrastructure standpoint? Are you ready from the data readiness standpoint? Are you ready from, you know, your people standpoint in terms of embracing this, right? So, you know, I firmly believe that, you know, unlike, you know, the software apocalypse that we are at least hearing from pundits. Physically, in operational setup, is pretty needed, right? So because today, the bottleneck is really the labor availability, right? So particularly in Europe and Americas, you don't have labor, right? So I mean, 77% of companies, I'm repeating this fact, 77% of companies clearly called out that this is their topmost challenge today, right? So we're already operating in an environment which is a huge bottleneck. How could you bring technology to address that? That's the potential that we have here.
We keep talking about, you know, that AI is not a technology challenge. It's also a people, culture, leadership challenge as well. Where does that sit in kind of your environment as well? And how much are you seeing culture being this inertia for progress as well in the factory setting?
Yeah, so, you know, always, right, so whether it is AI or any other, you know, we talked about really, you know, cloud and digital and it's always really, you know, at the end of the day starts and ends with people, right? So it's people, process, data, and technologies, particularly when it comes to physical world, right? So, you know, when we talk in the context of physical AI, industry 5.0, and so on and so forth, there are even additional dimensions that we need to be aware of, right? So for example, there are stringent labor laws and so on and so forth that you need to be aware of. The pace in which technologies are being applied in, for example, China, versus the pace in which it is being applied in Western countries, the huge difference is really the complexity with respect to regulations and so on and so forth as well. So I think, you know, first of all, leadership needs to be aware of what the potential of this and what this means for their people. And what you're talking about is really not only employees, it's more of a community as well that you're dealing with when it comes to operations world. What this means for them. And not necessarily to displace the worker community but to enable and empower them. Today, I've been to hundreds of factories. When you talk about really what people are doing in the shop floor, is there a better way to sort of empower them, for example? The concept of AGVs, AMRs, forklifts, et cetera, have been here for a long time, right? And people are still using it. But the complexity in terms of operating them, is there a way to sort of reduce it to empower them further, right? So, you know, first of all, you know, there has to be, you know, understanding of what this means for companies and what this means for the people. And how do you sort of, you know, approach this in a way that you empower people and sort of you bring this convergence of, you know, I'm going to use this word augmented intelligence, right? So, I think somebody used this yesterday as well. With people to be able to sort of empower them. What that means for productivity. I think we don't really have a demand problem, at least at this time. What we have is the supply problem. So, of course, the equation may change down the line, but how do you sort of be ready to cater to the market needs that exist today and also pivot to the flexibility and resilience you need in your operating model down the line? Yeah.
I think we're kind of in this interesting phase of AI and agentic development as well where we're talking about this human-AI collaboration. And for most of us in the room, we're kind of collaborating with, you know, digital agents and Claude and whatever it is. But in your environment, it's also physical kind of capabilities in AI as well. So what is that cultural shift that's happening in factories when you're beginning to leverage more digital and physical tools as well?
That's a great question. Let me give you an example. So, you know, Hitachi, you know, is a $80 billion company. And one of the things that we do is in the mobility space, right? So we make railcars and signaling systems and so on and so forth. When it comes to quality inspection of your railcars, right, so whether it is surface level quality like welding and paint jobs and so on and so forth, it used to be people walking around the railcar and inside and under to be able to understand in terms of identify what kind of quality defects they had, right? So what we have done now is to sort of bring robotics and AI to empower them to automate this process so that they no longer sort of simply go around and, you know, do this manually. Rather they get to really act on, you know, now Boston Dynamics' Spot with our AI algorithms and vision models are doing the quality inspection, bring the quality events, and now people basically analyze and act on it, right? So of course there is opportunity to further automate that, but the point I'm making here is the team was super happy, right? So they still, you know, spend the time that they need to spend, but they spend it really in a more productive and creative fashion rather than doing monotonous work. So that's really an example of how people on the shop floor react, but let me be very direct as well. There are going to be implications in terms of, like in every other sector, right? So in white collar as well as in blue collar, there are going to be impacts. I think there is no different way to say this. But having said that, how are we going to really, as I say, if you're running a company, for instance, you have this talent sort of bandwidth and potential that exists now because of what technology can do. What are you going to do with this available capacity? What else can you innovate? I think that's really how I would, it's like the classic half-glass empty. So companies can look at that time from that standpoint to be able to sort of leverage the opportunity.
What do you have to do to foster that culture of innovation and wanting to kind of shift beyond what you're currently doing as well?
Sorry, ask that again?
What are you doing in terms of driving that culture where people are excited to be leveraging these tools?
Yeah, I think the classic saying of what's in it for me, right, so that's there from a personal standpoint, whether you're talking this at the executive leadership level or the operational leadership level or the workers level, right, so as long as we are able to sort of articulate in terms of what this means for them and how this is going to empower them, that kind of automatically drives the adoption mindset. And the other dimension is, as I said, if we are able to sort of take the monotonous dimension out of really people's equation and give that bandwidth to them to do what they need to do, imagine the type of outcome that potentially we could tap on. So that's really how I would suggest that companies think and kind of drive the change agenda.
I would love to open it up to the audience as well if anybody has any questions for Sankar.
My question is in terms of the cost of physical AI if you're running all your models from the cloud. So any insights in terms of how people can connect equipment for various providers, various chipset providers to make it simpler and cheaper?
Yeah, the cost of running physical AI, particularly from a, that is, you know, Edge compute has been here for quite some time, but we are going to see more of that, right? So in real world environment, physical world environment, going back and forth to cloud, not only from a cost standpoint, but from a latency standpoint and security standpoint, et cetera, it's going to be problematic, which is the reason why, and you look at really even NVIDIA, for instance, right? So they bring the whole stack to run on cloud. So it's not going to be cheaper, right? So from a cost standpoint, there is always going to be, you know, it's a capital investment at the end of the day, right? So the way to think about this is particularly if, you know, when companies commission a brand new factory, the equation becomes a little easier to articulate and justify. But if you're going to really do this in existing factories or, you know, existing warehouses or, you know, whatnot, the analysis equation is slightly different. But to your point, I believe that with the combination of edge and cloud, you could manage that better. Yeah.
Any other questions? Marisa?
So when you're looking at manufacturing, a lot of the times it's hard to keep training, upskilling for multiple reasons, right? Sometimes out of the norm, how busy it is, you have unions. I would say if you're working on the manufacturing floor, sometimes it's very difficult with the training, the upskilling for multiple reasons, whether you're going through the fact that that's a bit hard to do around the day job, you might have unions to work through. There's a lot of factors, and we've talked so much about training the last day and a half now. Just curious if you have any tips or anything you could share on how you're helping to kind of shift that culture and get people trained and want to be a part of going through these trainings?
That's a great question. In fact, AI becomes a key enabler for that worker training today from a shop floor standpoint, right? So connected worker, you would have heard this word in the recent times. There are companies that leverage, you know, AI and sort of AR, VR, and so on and so forth to drive next level of operator training. So the churn in shop floor from a people standpoint is significant. So you always train people. So that's really the fact of the matter today. So companies are dealing with this today. But with what is happening from a technology evolution standpoint, the concept of training itself is kind of changing. So from training to sort of empowering people to simulate on the job virtually to acquire what they need to acquire and then become more accustomed to it. And then there is a gamification as well that's going on, which kind of motivates people, right? So nobody wants to listen to a lot of slides. Yesterday we were talking about, and I think we had the right, so, and he said, I'm gonna walk you through, I have only 90 slides. I think a lot of us laughed at him. He literally had 90 slides, but then he did it in a way that was super, you know, interesting. But the reality when it comes to really, you know, how to operate a complex equipment or a sit on the shop floor, you know, taking people through a set of slides and comprehensive documentation versus providing them an environment where they can virtually simulate to learn, people prefer the latter. And thanks to AI, that is becoming possible.
Thank you so much. We can ask questions outside, but thank you. Thank you, Sankar.
All right. Thank you so much.
If you don't have an account, Register here |
With the exception of our Horizons reports, most of our research is available for free on our website. Sign up for a free account and start realizing the power of insights now.
Our premium subscription gives enterprise clients access to our complete library of proprietary research, direct access to our industry analysts, and other benefits.
Contact us at [email protected] for more information on premium access.
If you are looking for help getting in touch with someone from HFS, please click the chat button to the bottom right of your screen to start a conversation with a member of our team.
