Navigating the Future of Operations
Manufacturing and production operations are under a big microscope right now with businesses feeling the pressure to become more flexible and still meet their other key performance indicators. To talk about what those businesses are and could be doing, we have Mark Hindsbo back on for the Future-Ready Podcast.
Nick Finberg
00:10
Hello and welcome to the Future Ready podcast from Siemens. I’m Nick Finberg and I’m joined today by Mark Hindsbo. This is part two of our discussion on software-defined and operations software in the first part, we talked about how the physical and digital worlds are coming together and why that’s creating a need for more adaptable software-defined production. Today, we’re building on that, looking at what it really takes to make this work in practice, from building a strong data foundation across the factory. to moving from reactive insights to more proactive agentic systems and ultimately how companies need to continuously reinvent themselves to stay competitive. We’ll also dive into how these ideas translate into real market opportunities and how value is created across the full life cycle from design to optimize. So, Mark, maybe looking at a wider view of the entire value chain for software to find from design to realize to optimize, what do you see as our key strengths? What differentiates us in delivering value across the end-to-end life cycle of a product?
Mark Hindsbo
01:16
I mean the main reason that I came to Siemens was because we do combine the physical and the real world. So, we have the entire if you want life cycle including operating our own factories so we’re kind of forced to drink our own champagne or eat our own dog food depending on whether it’s good or bad. But I see people coming from the pure software world that I’ve been in. and they have you know say digital tools for product development, but they never use those tools themselves right and it’s purely digital then we have others that might be hardware-only companies, right? And they produce hardware, but they don’t have the virtual world. I think we have the unique strength in that we have tools for Designing products and we have you know PLCs and drives and so on that go into the factory floor. We operate our own factories. And we’re able to stitch that together in an end-to-end workflow. I think that is our unique strength. I see we have colleagues slash competitors in the industry that are on one side of that fence or the other side of the fence. but very few that that can stitch the entire thing together. So, I think that’s one yellow sort of Siemens advantage. There are a couple of others. We are an industrial player, and, in this space, you have to roll up your sleeves and get some proverbial oil in your fingers as well, So I think that’s a that’s a real advantage as well. There are, again, call it generic. We’re operating system developers. We are developing an industrial operating system that has to survive on the factory floor. that has to have failover in milliseconds, that has to be FDA approved for their processes and so on. Having that industrial heritage, having the close connection to the customers that we have, I think is another you know, real advantage for us versus say we’ve seen IoT players that say, hey, we’re going to build an IoT system for every single thing in the world, you know, super horizontally I do think in the industrial space you have to be steeped in the space. And then a and then a willingness to reinvent. yourself. And look at how many times Siemens has reinvented itself in its history, right, from where we started and where we are today. So that that I think is a cultural trait that is an advantage. as well. And then and then it’s also the breadth of partnerships because none of us are going to do it alone. This has to be this is a heterogeneous you know world. It’s an open world where our systems have to interconnect. And I think we on one side, for instance, we have really, really deep partnerships with the industrial equipment makers, the people that make metal bending machines and CNC machines and 3D additive printing machines. And on the other hand, we have partnerships with the hyperscalers, with you know the LLM developers of the world and so on. So, we kind of sit and benefit from input from both of both sides of the spectrum.
Nick Finberg
04:49
In our last episode, Mark, you mentioned what some of our customers are doing and even how we are doing this for ourselves. What do these market opportunities look like? How are businesses growing?
Mark Hindsbo
05:02
There is a today when you look at my space, operational software, there’s about a you know $30 billion market, $13 billion Euro. you know depending on the exchange rate market for that today, that’s gonna double over the next five years. And it’s it it’s not going to double because of us, right, who supply the tools. It is it is doubling because the other things I talked about, because our customers need to have more adaptable production facilities, need to have more sustainable, need to go near shore, right? We’ve all we’ve been through a period of call it centralization where we put a lot of manufacturing in in in one location and now it’s sort of coming back again and becoming more distributed. Those things are only going to happen if we if we build a software-defined foundation for running our production facilities.
Nick Finberg
06:00
Yeah, it’s that constant. And balance of efficiency versus flexibility. Yeah. could you talk a little bit about how this kind of thing is gonna take shape with the industry? Maybe some of the maybe first from the operation side but also like the design and planning. How’s what are the first steps?
Mark Hindsbo
06:19
I it it’s gonna take place you know very pragmatically, very incrementally. Of course, there will be you know occasional moonshots there will be you know occasional visions. But I mean we did this is a playbook we’ve seen over and over again in all of these shifts. So, whether you take the industrial revolution or whether you take the internet or whatever it could be, right? It’s it does end up being More of a gradual process. So, what do I mean by that? Give an example of how we’re implementing it in our own factories. We do have an automation layer in place. Then we take a specific your product line and say what are the issues with this product line? What would we like to achieve towards you know us being a more effective manufacturer. Then we take a specific production line within a factory that’s producing that product and we build the, you know, for instance, the agentic AI capabilities for that specific line, for that specific product, with an idea that it can be generalizable and that it can scale, but we solve one specific use case. We’re not necessarily trying to boil the ocean and say. Every single problem in the entire world for our entire factory blueprint all at once, right? It it’s pretty focused, but it’s focused against the vision so that every step we take you know, builds towards that vision. If you don’t have the vision, it can quickly go wrong, right? execution without vision in my mind is random walk or drug walk, right? You just you know right which is a classic mathematical problem right but if you if you know where you’re aiming to then the random walk doesn’t become random. It might still zigzag a little bit but it goes in one general direction.
Nick Finberg
08:14
Exactly. You’re correcting rather than just moving around like one of those vibration robots that kids play with when they’re their first learning.
Mark Hindsbo
08:22
Yeah, yeah, yeah, yeah, exactly, which is the classic drunken walk up. but no and I see our customers doing the same by and large. There are some there are of course some greenfield opportunities where you’re clearly starting from a blank slate where you might be able to take two steps forward instead of only one or only a half. But by and large, they’re being equally focused on where do we believe that we could get value out of the technology today? How does it help our business today? And how does it point to a generalizable future Otherwise; we get this risk of it’s just technology chasing stuff for technology’s sake and we never get the value out of it.
Nick Finberg
09:06
Okay. So, kind of like what you were talking about earlier of Data helping to optimize the production processes itself. We’re using that ourselves. Is that kind of how we’re helping reinforce some of our own strengths with software defined?
Mark Hindsbo
09:22
Yeah, so again back to where I said if we want this adaptable production facility, if you don’t know what’s going on, there’s no way to adapt, right? So, it does start with information, and the factory space is heterogeneous. None of us are going to put in new machines every other day just because they have a new IoT sensor on it. So, you need to be able to build a data connectivity, a data fabric layer that can reach all the way back to 30 or 40 year old machines. And then on the flip side, you know, integrate with the newest machine that came yesterday and create that transparency. And that is non-trivial, absolutely non-trivial. But that’s sort of foundational layer where we’re investing both in the software assets, but then also cutting our own fingers on actually implementing it on our own factory floor. And then you need to connect that to an information layer on the top that can analyze that data, contextualize that data, and give you insights. And I think we are well through that journey, which is more Let me call it the reactive side, right? So, I give you an alert, something is wrong. You need to take a look at it. Even our AI today is mainly reactive. So, I might have in my manufacturing intelligence system from Siemens, I might get an alert. I might now ask my AI co-pilot. What can I do about that? You know, what’s wrong? Can you please give me context? Oh, it’s this machine, given the signals that we’re getting, this might be this and this valve that is deteriorating or clogging. Most likely a clogging. Now go get someone to unclog the valve, right? That that’s reactive if you want. I’d say that’s sort of where we are right now. The next step is then being pre-act proactive, right? that we have these more agentic systems that are constantly polling and saying, not I have a problem right now, but I might be heading towards a problem, but if I reconfigure things, I can fix that. I can reroute production over here while I maintain this. That’s the next step on the ladder. And we’re working on all of them, right? The PLC layer of automating that. The you know HMI layer that drives your processes, the scheduling system on top, the analytics layer that sits on top of that, right? And that’s also part of our uniqueness that we can bring all those layers together.
Nick Finberg
12:01
In one of my previous conversations there was a really interesting idea brought up around augmenting HMIs. that when an operator walks up to a machine, there could be an active prompt of, hey, there was an issue while you were away. It wasn’t critical, but you might want to investigate. Is this an example of that integration of layers?
Mark Hindsbo
12:25
So, it’s exactly that. And I think the interesting thing is I’m glad you brought this example up is How do you then present it in context? And that’s where I think we do have strength as well. Because we have an HMI system. that already sits on the shop floor, right? It’s designed in some cases to be able to have oil on it, to vibrate. It sits on the machine next to the software-defined emergency stop button. So, it’s not like we’re saying to the to the operator or the person on the on the shop floor, oh you now have to run up in front of your laptop in your office and get your insights. No, you’re getting the insight in the context of what you’re doing right now, and you’re able to take action in the context of what you’re doing right now. So, it’s not you know an AI co-pilot. That sits in some magic cloud somewhere, right? It’s a copilot that sits in the context of your process and also in the context of the way that you are interacting with that process.
Nick Finberg
13:28
With that example, I’m realizing I’m not entirely sure of the difference between software and software defined. Or is there a difference?
Mark Hindsbo
13:36
Let me maybe give an analogy. So, if I think about my car that that I have, right? That used to be a very hardware-defined vehicle. Today the difference between the model that I have and the next model up, which is about 50 horsepower stronger in the engine. is not necessarily the engine. Mechanically, those two engines are almost a hundred percent the same. But the software system that controls those that engine and sets up you know the firing speeds, the compressions, everything and configures are different. So, I think that’s an example of software defined hardware. Whereas take the entertainment system in my car, that’s almost like just a pure, you know, software experience right. It’s it it’s just yeah. I mean of course there’s hardware involved because there’s satellites in space that give me my GPS system and so on, but it tends to be a little more on the you know software spectrum, my engine might tend to be a little more on hardware side of the spectrum. But it’s not that for me to get my You know, my engine reconfigured, a lot of that can be done dynamically on the fly. ABS brakes are another great example of that, right? What’s the braking speed of them? That’s software-controlled hardware as well. So, I think we see that, you know, all over the place, but I don’t. There’s lots of grayscale, right? What’s embedded what’s embedded development, what’s not embedded development. you know what’s software defined, what’s hardware defined, that doesn’t sort of, to be honest, concern me that much.
Nick Finberg
15:20
So, looking to the future, what do you think is going to be possible the next two to five years? It can, it can sometimes be hard to balance the long-term investment potential and the near-term results of a new technology process or mindset. How do you how do you think about this when you’re talking with your team and with customers?
Mark Hindsbo
15:37
So, you know I kind of remind myself a little bit, I think it was William Gibson who said the future’s already here, it’s just unevenly distributed. What does he what did he mean? He means that there are these front runners, right? So, we had PepsiCo on stage at CES talking about what they’re doing. We just talked about what we’re doing in our factories. There are we don’t have to wait three to five years to figure out what will happen at scale three to five years from now. That’s happening today. And we can work with our customers to make it happen for them today. It won’t happen at scale, uh, it it’ll happen in pockets. But where the thing that typically happens with a lot of these technology shifts is it doesn’t allow stuff to happen that you couldn’t have done before. It just becomes two orders of magnitude cheaper and less complex to do it. So, the internet did not enable any communication or electronic commerce or whatnot that couldn’t have happened before it was just fairly cumbersome to do it before and it became democratized. So, I think we are in the same journey of we are democratizing the adaptability of in in my case of production. So, we have to pinpoint with our customers where will you get value out of it today? What is your proof point that is your highest value return of investment? And how can you then use it to scale as you come down the cost curve, so to speak, and it becomes a more commonplace activity? So that’s you know how I think about it with also with my teams because there’s classic hype cycle. I have no doubt, for instance, that AI will be a very valuable technology similar to the way the internet was, but we tend to overestimate them in the short term and underestimate them in the long term
Nick Finberg
17:40
So, you need time to engineer on that that solution. How is it going to integrate?
Mark Hindsbo
17:46
And some things are great dreams, but they’re not quite the reality yet. you know I mean I can be as visionary and excited I would say as most others but then it’s dialing it back and say okay What’s the reality today and where can we deliver it today?
Nick Finberg
18:06
Are there any interesting conversations you’ve had recently with customers or with industry experts in your in your travels? about what these first steps are looking like. How are they positioning themselves now to be able to not only capture the value today with those very easy point problems that they have but also understanding it as a framework to do more in the future.
Mark Hindsbo
18:32
So, and we talked about some of it already. I think one thing that I am seeing across everything is we’re all realizing that data is crucial to the journey. So, if we don’t have access to the data. And if we don’t have a way of having that connected, if we don’t have a way of getting it more or less in real time, and I don’t necessarily in some cases if it’s a train that has to break, you better have the data in milliseconds. But other times, I mean real time might be 10 minutes from now or next day or whatever. It kind of depends on the problem, right? But getting that layer in place is a crucial first step that people are taking. And that is something that we can absolutely Do at scale. And we’re also doing it with our customers. We announced a data alliance recently with our machine builders and others. So, if you look at these large language models, if I’m a bit cheeky, they have been trained on Twitter feeds and YouTube videos. So, they’re really, really good at things that are related to that space. But we need to train them on time series data from machines. We need to train them on 3D CAT data. And that’s part of what Siemens is also investing in and where the industrial context comes in, right? We announced that we are Driving research into foundation models and development of foundation models that are trained on an industrial context So those are also things that we are doing today and that I think will enable things going forward.
Nick Finberg
20:09
You mentioned the discrepancy between brownfield and greenfield earlier. How does that fit into to this shift? how are Brownfield operations able to kind of start integrating this data structure?
Mark Hindsbo
20:21
I mean fair fairly easily. But most customers that I talked to, as I said, have this spectrum of machine. They have stuff they installed yesterday that comes with all the latest and greatest. But you can retrofit your old machines with a vibration sensor or a temperature sensor or whatever. the appropriate mechanism might be. and then you just I mean it’s always a cost thing, right? You can’t retrofit it with everything. So, decide what are the key signals that we need to act upon and let’s instrument those for our specific use cases. And in some cases, you can even simulate things. So, if you have you know the temperature from three different points you might be able to simulate if you know the process of what’s going on in inside or you, I had a really, really great conversation a while back with a big manufacturer of ball bearings. And based on the vibrational patterns that they measure in the macro level on the outside, because they have these very sophisticated simulation models, of their bearings and their bearing balls and the oils and the viscosity that it runs in, they can very accurately diagnose even in Brownfield, right, what’s going on. So, it’s more to say that that there it the brown field can absolutely move forward as well. It’s not a rip and replace, take everything out and you can’t start before you have a full set of new machineries that would be simply not feasible.
Nick Finberg
22:00
Thank you so much, Mark. It’s a real pleasure to sit down with you for these discussions and I really look forward to covering more with you. on your world and some of the challenges our customers are facing in a bit more detail. But until then, for the audience, be sure to check out our other episodes where we talk with Rainer Brehm or Alastair Orchard about software-defined automation. and operations on the shop floor or tune into some of the other episodes around artificial intelligence. there are a few from my colleague Spencer Acain, as well as a longer episode first recorded with The Economist. See you next time.
