Data Foundation for AI-Driven Production in Mid-Sized Manufacturing

Today on the IoT Use Case Podcast: Industrial AI and digitalization don't start with data or with artificial intelligence – they start with the question of where a company actually wants to go. Together with SCHUNK, Next Level Mittelstand, Cybus and Schwarz Digits, we talk about why German mid-sized companies now need to be far more ambitious about automation and digitalization in order to stay internationally competitive. We look at concrete examples from practice, at the importance of a reliable data foundation, and above all at shared target pictures that provide orientation.

Moving companies from debating to ambitious implementation. An episode for everyone concerned with the future of industrial mid-sized manufacturing. Enjoy!

Welcome to the IoT Use Case Podcast, the channel with the latest IoT projects from our implementation partners.

Hello, dear friends of IoT. Welcome to the IoT Use Case Podcast. We have been talking about Industrie 4.0 for more than a decade now. And yet, in many factories it is still very laborious to get reliable data out of machines and equipment – especially in a standardized way across multiple sites – and then to build productive applications on top of it. With the competitors from China, who can scale enormously thanks to their home market, and with Industrial AI, an entirely new pressure to act is now emerging. Companies want to and really have to shorten their lead times. They want to automate processes more heavily and make better use of their production knowledge. But that first requires a data foundation you can trust to work in day-to-day operations.

Joining me today to discuss this are Henrik Schunk, Chairman of the Board of Directors of SCHUNK and initiator of Next Level Mittelstand, Peter Sorowka, Co-Founder and Co-CEO of Cybus, and Daniel Traub of Schwarz Digits. Henrik brings the perspective of a manufacturing family business and of the Mittelstand. Peter explains how heterogeneous production data becomes reliably and scalably available. And Daniel adds the question of how this data can be operated sovereignly and used for cloud and AI applications. Henrik, let me start with you. Mid-sized companies have been hearing for years that they need to digitalize their production more. What has actually changed with the current discussion around artificial intelligence, and where exactly does the pressure to act show up on the shop floor?

Henrik

Yes, hello from my side as well. Over the past, let's say, five years plus, mid-sized companies have all tried to get the shop floor into shape, including where data is concerned. The topic of data – or really Software Defined Manufacturing, you could almost call it that – is now gaining a lot more momentum through Industrial AI. And that means every mid-sized company has to check for itself and review its fitness level on the shop floor, in terms of data quality.

And the ambition has to be to achieve the highest efficiency and performance at a high-wage location like Germany by stepping things up another notch. In my view, with AI, agents, everything that is being added on top now, the possibilities for improving lead times and efficiency have grown enormously. I think it is simply necessary for you as a managing director or owner to take another look at performance and to send your plant managers and production managers off to see what is actually possible.

When we talk about the space of possibilities, a lot of it comes down to: I have the data so I can work with it. And Peter, when it comes to industrial data foundations in particular – what has changed over the last while, from those monolithic IoT platforms through to today? Where do we stand right now?

Peter

We at Cybus have been active in exactly this segment for about ten or eleven years, and we have observed a great deal in the market, right from the first tentative steps. It became clear early on that most companies tried to implement individual use cases. Predictive Maintenance was always the big one, the topic that shaped every second PowerPoint slide. When can I actually predict that the next machine will fail? It sounds great. Honestly, I haven't seen a huge number of real implementations of it that truly scaled.

I think two things have happened. One is a certain return to simplicity. By simplicity I mean, first of all, that most companies in manufacturing – and I'm talking more about discrete manufacturing here, companies with a lot of machine tools, perhaps even manual assembly, so not so much the process industry – most companies don't know which of their machines are running right now and which ones currently have a fault. They don't know it live. And they don't really know it historically either, because this data is traditionally recorded on paper cards, by hand, and maybe evaluated at some point if you can read the handwriting.

And that of course makes it incredibly difficult – given that we just mentioned optimizing lead times – to find out where the bottlenecks actually are. If you don't know which problem you need to solve, you can't solve it either. So establishing a certain foundation, getting the basics in place, is I think the first point.

The second point is that most of the companies we speak with today have understood that it isn't enough to look at isolated use cases, connecting one machine to a dashboard. We need a holistic view – looking at the entire production process from beginning to end. On the one hand that has to do with end-to-end productivity analysis, but on the other hand also with use cases such as traceability, carbon footprint tracking or quality analysis – where in a long production process do quality problems occur. If I don't get all the data along the production line in a genuinely consistent form, I also won't get enough context to really analyze problems. That's the second point.

The third point, and I think this is the truly decisive one: for twelve years we have been told that data is the new oil, that you just need enough data, sprinkle some analytics over it like a bit of salt, and then it will all work out and you'll have data-driven decision making and smart factories. And I think we all know that this didn't catch on in the market, because first of all the heterogeneity of the data is so high that it simply isn't as easy as in e-commerce, where I say, okay, let me analyze 10,000 shopping baskets in my Shopify and draw a conclusion from that. In a factory it isn't that simple. And secondly, sometimes the imagination was missing as to what the data could actually tell us.

But what we are seeing now is that thanks to AI and the power of Claude and all the tools we can use today, this creativity is suddenly no longer left to humans alone. I can ask Claude: think about which data I would need from my production in order to analyze it. And Claude will give me an answer that makes sense. And it will even build you a prototype with a dashboard that lets you run the analysis. And now the last remaining element is: can I actually connect what I have just built in that prototype to a real production environment?

And this whole tedious topic of IT-OT integration, which we have been grinding away at for years, suddenly becomes an implementation bottleneck. And conversely, what we are now seeing is that finally the production managers – and not just the innovation, IT and digitalization managers, but the production managers – are saying: damn it, I need this data.

You're pushing at a very open door with me there – the idea that you can use AI, generative AI, to analyze what you need and how to go about it. Many people only think of AI in terms of the final installation, the end process and how it runs. But actually letting it guide you, discussing the whole thing with the AI and working it out, that's the point. And now – and from my perspective this really is very new – AI is genuinely capable of building these applications. We recently had our own homepage rebuilt by AI. It's hugely impressive, I think it's excellent.

So this bottleneck is now IT-OT integration, this understanding and the availability of data. Daniel, over to you next: availability of data and trust in data. Where does Schwarz Digits come in? What is your approach in the context of a sovereign cloud?

Daniel

Well, a lot of valuable and important points have already come up. So, a warm welcome from my side as well: Daniel Traub, responsible for the Manufacturing Automotive vertical at Schwarz Digits. Which makes this the perfect fit here with Henrik and Peter.

At Schwarz Digits, we have very clearly set out not only to launch STACKIT as the number one hyperscaler for sovereign workloads – what exactly that means and where we position ourselves is something we will certainly explore in the course of this conversation. But we also want to clearly stand for establishing a digitally leading Europe, for building an ecosystem of partners and customers who contribute to it, so that collectively we don't just end up with a single-vendor strategy at a customer, but genuinely offer alternatives and room for options. I think that is what we stand for and what we are setting out to do.

If I now put that into the context of Peter and Henrik, and especially into the context of AI and an incredibly fast-moving industry at this point, then I do think it is important and valuable to talk precisely about where the data used for AI actually finds its home. Because one thing is very clear: we have a situation here where we are generally talking about critical domain know-how. Production knowledge, production data and production know-how are often a differentiating factor in the market.

Particularly when we look at German mid-sized industry, at our hidden champions, it really often is the case that production know-how is also differentiating know-how. And accordingly we have set out to find a, in quotation marks, safe harbor for this data. From different perspectives. First from a legal perspective – legal in the sense of with whom and in which jurisdiction do I sign my contract. And on the other hand of course in the context of: where does my data actually sit? Is it also traceable where my data sits?

And ideally, at the end of the chain, to also ask: which AI am I actually using? Peter, you brought up a great point with Claude. We are of course also working towards providing corresponding open source models, and towards a joint European effort to potentially launch Industrial Foundation Models that could then also be hosted in STACKIT. So there are many starting points. I hope it has become somewhat clear what we are setting out to do, and what the collaboration between all of us can make of it.

Peter

Daniel, you're absolutely right. Claude is of course just the current example. I think if we look at the last one or two years of AI development, we know one thing: nothing stays the same for three months at a stretch – there is basically news every week. I think a truly decisive strategic positioning for every company – not just in manufacturing, and I include our software company here and probably you as podcast producers too, Peter – is to make yourself independent of the specific AI.

So I believe we have to create the foundations, the working methods and the technology stack that let us use best of breed in each case. With the necessary decision criteria, of course: sometimes it's trust, sometimes it's performance, sometimes it's cost. But the world is going to change completely again in three months, in six months and in twelve months.

And especially in manufacturing, where that kind of pace of change is basically unthinkable – we normally talk in terms of five years – I believe laying that groundwork, creating that foundation on which I can then flexibly say, today it's Claude, tomorrow it's Aleph Alpha or an open source model, or something else, is absolutely crucial for companies.

The current developments actually play right into your hands at Schwarz Digits, don't they? Take the maker of these Claude models, Anthropic: they had released this new model, Fable, and then from one day to the next it was effectively stopped by the American government and nobody had access to the model any more. A manufacturing company simply cannot afford that kind of unreliability. Issues like that, or limited compute capacity – those are interesting aspects that play into your hands. And at the same time there has always been the discussion: do I do this on-premise, meaning do I let my data leave my factory environment at all, or do I use a cloud? So where do you see the current discussion with your customers?

Daniel

Well, you have essentially already outlined the two different types of customers. On the one hand those that are still very strongly on-prem oriented. But we shouldn't forget that in German industry, particularly when we think of the large enterprise customers, we also have customers who have largely been using cloud services from the well-known large hyperscalers for many, many years. And I have to say quite clearly, I think the ideal profile probably doesn't exist, because in every case there are always pros and cons.

Clearly, for many companies it is partly about scalable global availability. I think these use cases are absolutely relevant and legitimate. On the other hand there is an increasingly strong consideration – this is what you touched on – coming out of a geopolitical momentum, as I would put it, towards greater risk diversification with regard to my IT services in general. That is simply a fact.

And accordingly, it is a matter of finding the right application case for the respective customer. On the one hand genuinely accompanying them into the cloud and giving them the confidence that data which used to be on-prem is properly secure in STACKIT, our cloud product, and also properly available in a scalable way. And on the other hand giving customers who are already very cloud-agnostic and very cloud-oriented an opportunity to diversify risk. So yes, the momentum is there.

Nevertheless it is important that we pick the right thing together with our potential customers in those conversations. The right starting point for providing a complementary offering to today's landscape. So a complementary offering in the context of a company that is strongly on-prem. But equally, if they are already in the cloud, a complementary offering for the right use cases. And we touched on this briefly earlier: domain know-how, critical topics, production data, customer data and so on.

And that opens up an enormous space of possibilities. I think the term space of possibilities came up earlier as well. And that also makes it difficult. I still remember from my time at Siemens, we went to market with a platform where we essentially said: dear customers, you can do anything with this. And that quickly becomes overwhelming, especially in the Mittelstand. Henrik, so over to you: you started this Next Level Mittelstand initiative. And my understanding at least – and I'm curious to hear how you describe it – is that it is meant to provide something of a framework, to help and guide people in picking the right topics out of this enormous space of what is possible. Could you explain a little how you go about it, what this initiative is, and what the recommended steps are?

Henrik

It is exactly as you say. The sheer variety of topics, with ever more possibilities emerging around AI agents, but also independently of agents the whole question of how I have to build my IT infrastructure, how I protect my data – all of which has already been mentioned – means that as a mid-sized company with limited resources you quickly reach your limits. And it is far smarter and more clever to say: I'll join forces with others, build a strong community, exchange experiences and learn from the best, or avoid making mistakes others have already made a second time. That is roughly the background to Next Level Mittelstand.

And the goal is for mid-sized users to join forces and tackle the digital transformation together, with the aim of becoming digital champions themselves, and within this community – with strong orientation but also a strong implementation focus – to approach digital transformation, the use of AI agents, improving shop floor performance and all the other relevant topics with an approach that requires fewer resources and is faster and smarter, and in doing so to give the topic of digitalization a strong boost across the breadth of the Mittelstand.

That is the objective, and after almost a year we have now gained 47 members who believe in this community idea, and we want to keep growing strongly. It is an open network, so IT service providers, software developers, anyone who believes they can contribute to moving mid-sized companies forward is very welcome. And that's a good thing. I'm pleased to say that STACKIT, with Daniel Traub, is also represented as a shareholder – it is a joint venture of six companies – and Cybus with Peter Sorowka is on board as a member. We are on our way, we have already achieved quite a lot, and of course we look forward to many new members.

Could you pick out one concrete topic? I mean, at the Hannover Messe you had a demonstrator, for example, where you showed one of your automation cells in interplay with Schwarz Digits and Cybus. But just an example that comes to mind spontaneously, something where you say: this is something we already know works and that would genuinely help mid-sized companies – tapping into data and then making it available via the cloud. Where is the benefit? What's an example you really like?

Henrik

One example where the spirit becomes very visible is certainly – and this was an early topic – Product Carbon Footprint. Every mid-sized company has to address this topic, or really every company. It is not an easy topic to get to grips with, and in the community we started a working group dedicated to PCF, with the objective – and after two or three months we had it done – of being able to report a PCF for a certain number of products, and essentially to develop a recipe for how any other company can apply this themselves.

And we were then able to test it nicely in the community on a case where a mid-sized company had been notified of an order from an automotive OEM who had made it a condition: I want a PCF for the product, a PCF value reported for the product I want to order from you. And on the basis of our recipe we were then able to provide a PCF within two hours. And that led to the order from the automotive OEM. That is the proof that it worked well.

Other cases – you mentioned it – include a microfactory, where we say that a standard automation cell from SCHUNK can be connected to networks very quickly in a self-contained case, via the data logistics from Cybus and the secure cloud from STACKIT, and can enable decentralized units to be created within a future production network.

We showed this at the Hannover Messe, and the idea is that in a highly standardized way, connecting the automation cells of the future can be accomplished faster and give SMEs a time advantage. Those are two examples.

Peter

I would frame it much more broadly. One core problem we have in Germany, in industry: this is a country of great engineers, and they are used to wanting to solve problems themselves. And we have an enormous not-invented-here problem. An enormous one. I have been running Cybus since 2015, and since 2015 I have been at so many companies that come up with their own specifications, that spend three years on vendor selection for problems that are not unique to their company. These are all digitalization topics, all data topics, all IT topics. You can then inflate them a bit more with cybersecurity and the notion that our production process is so special.

And of course the managing directors and executive boards of all these machine building companies don't come from a digitalization background and have to rely on their IT departments and their specialists to assess all of this. And if you ask me, all of that only holds us back, because everyone builds the entire technology stack individually from the sensor and the network access upwards, and acts as if the issue were a matter of connectivity or something similar.

Instead, I believe technology isn't our problem at all. For ten years now, technology has not been our problem. Anyone who says that technology and technology scalability are somehow holding us back in digitalization hasn't been paying attention. We have so much technology available. Netflix is able to serve millions of people with the same video streams without a single stutter. That is scaling. If we connect 200 machine tools and want to look at how fast the spindle is turning, the data volume that produces is not scaling by comparison. The technology has long been ready. There are hundreds of options, including open source, including for free if need be.

The problem is the business breakthrough, the understanding of how decision-making processes have to be set up on a data basis, the willingness to get data out of the Excel spreadsheets on the production manager's desktop and make it globally available, to make yourself transparent as a shift supervisor towards your superior – because data always says something about performance, and that means benchmarking, and then the works council comes in. All of that holds us back.

If we now take Next Level Mittelstand as a bridge, as a network for exchange where it isn't the technicians but the management who can exchange views and share success stories, then we might arrive at an understanding of what is possible and where we can learn from each other, without reinventing the wheel every single time. Because there are plenty of wheels, and in Germany we no longer have the time. I think we all agree on that – we cannot carry on like this for another three years.

Just last week I heard it several times: China used to be cheap, now China is good as well. We have a huge problem if we carry on like this, with us engineers stewing in our own juice. And that is why I believe networks like Next Level Mittelstand are absolutely decisive, so that we can pool our forces a bit.

I think that's excellent. I really like that spin. I would actually have gone into the details now – how data is transmitted and how your platform can enable scaling. But I'd rather stay at this general and, from my perspective, extremely important altitude and bring in a topic I know well from my own time at Siemens. As you say, many data projects fail not because of technology but largely because there is no shared understanding of where you actually want to go. In other words, no target pictures.

Data transparency alone is not a business case, but it does create a broad foundation for an incredible number of applications – not one killer use case, but many smaller and larger levers you can pull right across the organization: savings, efficiency gains and so on. And what has changed from my perspective: developing target pictures used to be an expensive consulting project. And that's why it didn't happen. People simply didn't have the time. Today, with generative AI, such target pictures can be developed extremely quickly. With my company SMC we do it in a one-day workshop. And I'd like to hear from you: how do you view these target pictures? This shared understanding of where you want to go – how valuable is it? And in practice, why doesn't it happen? Why isn't there a sense of, look, we have an understanding of where we as a company, as a mid-sized manufacturer, want to go?

Daniel

Let me start briefly with one perspective, because many years ago I was also part of a central Industrie 4.0 project at one of the big tier ones. And I think in general something is striking again in the AI context that already struck once with Industrie 4.0. It is an extremely fast-moving industry with a continuous stream of new players in this space, which sometimes leads to – and this is my personal experience – companies tending not to make a decision at all, because they don't want to make the wrong decision.

That also feeds into your question of how important target pictures are. I think target pictures are extremely important. One question we could gladly explore together is: how important is this use-case ROI orientation compared with the courage to build a motorway? To put it in pragmatic terms. And I think that's the point where a change in mindset has to happen.

You just mentioned this killer use case. Even if it isn't handed to me on a silver platter in an Excel spreadsheet, that doesn't mean I don't have to do anything. And this general courage to move forward and dive into the topic is, I think, an important point.

Peter

May I jump straight in there? On ROI, you're pushing at a wide open door with me. I recently had an example at a customer. I toured the plant and we talked about how you would have to improve this here and that there. And then I saw a board with paper cards where machine fault reasons were being recorded. The time and the fault reason were written on them by hand and then they were slotted in. And I said that would be the first thing I would tackle.

That would be the very first thing I would simply digitalize with a digital process. So that you can evaluate it, so that it doesn't fail because of whether you can read the handwriting, so that you can perhaps even respond to fault reasons automatically, and so on. Look at everything that follows from that. Because I have customers who have digitalized it. It is so digital there that the fault pops up automatically in Teams for the right person, and suddenly the response time for fixing the fault has dropped from an average of 30 minutes to an average of two minutes. And that kind of thing creates productivity.

But at that moment, the customer's answer was: yes, but the paper doesn't cost anything, so there's no ROI. And I said, that's where we have a mindset problem, where we have to start setting certain paradigms as our default. I mean, if you ask me, Peter, about target pictures: for me the absolute target picture is the fully autonomous factory.

Maybe it is a theoretical target picture, but it is the target picture to begin with. The autonomous factory that steers itself completely on its own. AI agents manage production, call the AGV, robots load the cell. Maybe there is a human in the loop somewhere, or someone in the background, but the largely autonomous factory is the target picture. Just as with autonomous driving, I define different levels of autonomy within this target picture. Of course we don't start with the fully autonomous factory – just as with autonomous driving there are driver assistance systems.

For me, level 0 is the factories we have today. They are automated but not autonomous, though they may be connected. That's what I first have to establish. I first have to establish connectivity so that I can achieve basic transparency at all. Then I can do level 1: for me that is advisory, the factory that advises me. I have an AI that watches along, looks over my shoulder, looks at the data and makes suggestions to people. Adjust this here – but the action is taken by the human.

Level 2 is where the AI can start to act, perhaps even to control machines and automate processes, but still with a human hand on it. Then there is level 3, where the AI already has the mandate to control certain things itself, and level 4 would be fully autonomous. And you can break that down, and in an autonomous factory – it really is that simple – there is one hundred percent certainly no paper.

And then I can start dealing with topics like that. But it takes courage, it takes budget, and in the end it takes a CEO and also a CFO who says: okay, I'll release the investment here. And that is difficult these days in a world where many customers tell us they don't even have enough budget to keep IT operations running. Where am I supposed to find innovation? But that is exactly why I say that reinventing the wheel now and wasting time is definitely the wrong way to go. You have to define a target picture like that. And then you have to move forward boldly.

A great model, and very concrete with the different levels. Henrik, now from the perspective of mid-sized companies: how do you see it? Where does it fail, and how might target pictures help?

Henrik

I think Peter nailed it. Peter, I'd like to discuss what you just said with the Mittelstand at one of our next roundtables or meetings. How could we perhaps even develop such a target picture together and hold out a helping hand to SMEs and mid-sized companies? I think we have to learn to think big. We must not stay small in small things.

That starts with us as leaders and with the managing directors and owners, allowing space, saying: look, you have a plant here, what should this plant look like in five years? Show me your ambition and your vision. And I often find that because day-to-day business is always there and always takes priority, too little innovation can happen on the shop floor and too little forward-looking vision takes place. And for Germany as a location that is a problem and an enormous risk.

I think we have to flip the switch. We have to free up funds. We have to free up investment. And we have to give people from the top down the opportunity to make proposals and to think big. As Peter just described correctly, I keep seeing at SCHUNK as well that return on investment is the first thing people think about, and that as a result big plans or visionary directions stall somewhat.

And for me it is always a nice example – and I think many mid-sized companies will nod when they hear this: the decision to purchase a new machine is made on the staircase. There is relatively little discussion of return on investment, because the machine produces chips, after all. And I think in IT we have to achieve a similar shift in mindset. If we really want to get our factories converted to software defined, then we have to make concessions on return on investment and first create a basis.

And that basis means I have to allow investments even when you can't see an immediate return, or when someone finds it hard to express a return. Because if I don't do that, then my only option in X years is to relocate my site to some low-wage country. And none of us actually want that for Germany. So we know it now: as owners, as entrepreneurs, we have to step up and, to secure our locations and for sustainability for our children too, give IT and digitalization more momentum and free up investment.

Peter

May I put that into perspective briefly? I find that making concessions on return on investment sounds as if it doesn't pay off. And I believe the opposite is the case. I have so many examples of customers who said that on the day we saw the data live on the dashboard for the first time, we immediately solved problems we hadn't known we had.

Last year I had a customer who said that with the data on the dashboard they discovered within a week that if they leave the machine in a different loading state in the evening than before, they can ramp up faster in the morning. And just like that they had gained 30 minutes of productivity and could produce more. But they hadn't been aware that this was a bottleneck before.

And on these topics – most companies don't even know their OEE figures. They don't have the KPIs that would let them say, we should be able to get another 3% here. And that is why it is a chicken-and-egg problem. If I don't know my problem, I can't say I want to solve it. And then I also can't say where the return on investment will come from.

Last Friday I was at a customer who showed me his fully connected plant, a pure assembly plant. They produce large construction machinery, pure assembly, purely manual work, so not even automated. But they have digitalized everything. Every call button used by an operator – I need support here – is a digital event in their data backbone. And with that data, which they have had for three years, they were able to run analyses and say, okay, at this workstation we have an above-average need for support. Then they changed processes, solved problems and increased throughput.

Increasing throughput means either – and this depends on the economic situation – I can produce more if it is also bought, or during short-time work I can maintain productivity with fewer operators, or with the labor shortage, because I have difficulty finding employees, young ones, since more are retiring than are coming up from below, I can still achieve the same or even higher productivity output with fewer staff. Those are the effects where the ROI is phenomenally high. But I have to create the preconditions.

And planning the ROI in advance is often difficult. That's why I find our discussion really strong – I've enjoyed this enormously. I'd happily continue it, but I notice we'll otherwise run into trouble on time. So let me do one final closing round. My understanding would be that Next Level Mittelstand is a platform for exchange where you can really discuss things like this, where target pictures are also on the agenda as an upcoming topic. Any further points to pass on: where can people reach you, how can they engage you, how should mid-sized companies get started – so what would be concrete steps for our listeners to get going or to get in touch with you?

Daniel

You can of course reach us via the Next Level Mittelstand website. I think we are all also approachable beyond that via LinkedIn and so on, or of course through our respective companies. What I would like to add is a genuine invitation to everyone to take part in something that also has value for society. Henrik raised something extremely important. He talked about the fact that we really do have the opportunity, in the context of AI and everything that is emerging.

Peter talked about how fast the world is turning. This is not a lost cause we are in. And we should all hold on to it collectively, in the sense of an ecosystem, and work together so that we also give something back to our society. Especially in the context of AI, of infrastructure and so on.

If we spend another 10 or 15 years financing technologies whose ownership doesn't lie with us, then we will end up with a different problem. Because then we won't have anyone left who can actually take an AI agent to the next level – we will simply be consumers. And that isn't fit for the next generation. Anyone who has heard the term knows it is something the Schwarz Group highlights as one of its key points of identification. And I think it is something that matters to all of us. So the invitation to participate through Next Level Mittelstand, or to get in touch directly – yes, that's my call to action here.

Excellent. Then many thanks to all our listeners. Please share this episode. I genuinely believe that every mid-sized company in Germany should hear what we discussed today. I'm firmly convinced of that. So thank you for listening, thank you for sharing, and thank you for continuing the conversation.