FutureTalks: Modernise to mobilise: how German companies can transform legacy systems with AI
Outdated core systems are increasingly viewed as barriers to innovation – research shows that 7 in 10 German banking leaders believe legacy infrastructure limits their ability to deliver the digital experiences customers expect. At the same time, nearly half of organisations are prioritising AI and emerging technologies as part of their strategic roadmap for the next three years.
In this discussion, we explore how modernising legacy IT systems and embracing automation can accelerate digital transformation in the era of AI. Invited industry experts share insights, practical approaches, and success stories on transforming legacy systems into agile, future-ready platforms that enable innovation and growth.
Justyna Szymańska-Laskowska: Okay, good evening everyone. It’s a real pleasure to welcome you to tonight’s event: “Modernise to mobilise: how German companies can transform legacy systems with AI”, hosted by Future Processing and Perpowers. My name is Justyna Szymańska-Laskowska, I’m the DACH Business Director at Future Processing and I’m super excited to kick off what promises to be an engaging and forward-looking discussion.
As we all know, legacy systems are often the biggest roadblock to innovation, but tonight we’re here to explore how AI can help us not only modernize but mobilize our businesses for the future. And we’re not just talking theory; with our live AI challenge later tonight, you’ll see real-world application in action. We’re honored to be joined by a fantastic lineup of speakers who bring deep expertise and diverse perspectives. So please allow me to introduce Dr. Peter Zoller [Latos], leadership and transformation expert, and Rolf Löwisch, Director and Head of AI at IBM.
And helping us navigate our discussion tonight is our moderator, Krzysztof Szabelski, Head of Technology at Future Processing. Chris will make sure we keep the conversation sharp, relevant, and engaging.
Krzysztof Szabelski: That’s the promise, I’ll do my best.
Justyna Szymańska-Laskowska: Chris will also introduce the AI challenge in a few moments after the panel discussion. We cordially invite you to some more drinks, snacks, and relaxed networking. Before we dive into the talk, I’d like to extend our big, heartfelt thank you to Andy Lucas and the whole Perpowers team for hosting us in their wonderful HQ and co-organizing this event with us. Thank you for joining and we’re looking forward to an evening full of inspiration and practical takeaways. So, let’s get started.
Krzysztof Szabelski: Okay, hello everyone and welcome. Really a pleasure that all of you joined us in exploring today’s topic, which is probably one of the oldest questions we have in technology, but the question that never gets out of date: how do we get from our legacy that we created in the past to the innovation that is ahead of us?
That question probably doesn’t have a clear answer and will never have, but today hopefully we will bring insights on how to navigate that. A good food for thought for the beginning is that recently I came across a report named “Digital Transformation: What’s Next for German Banks”. It stated that 70% of respondents say legacy systems are hindering the companies from providing a digital experience their customers expect. So, is it the reason that everyone now should put big budgets for modernization? But on the other hand, huge companies are having a lot of that legacy, so there needs to be a reason why.
So I believe we need to figure out where is the right balance. Over the next one hour… we’ll try to keep it short, but we’ll have that discussion. And before that, we’ll have something really exciting: our live AI challenge. We brought it in to have that expert discussion on “why yes, why not”, but right behind the corner there is this AI which is promising—or threatening—us to flip upside down all that we are used to. We will try to answer that one question: can a single AI expert build a useful IT system along the time we are talking here? So let’s find out. Please welcome Maciej Sankowski, AI researcher and machine learning engineer at Future Processing. Maciej, thank you.
Maciej Sankowski: Thank you very much. So ladies and gentlemen, it is my great honor and pleasure to welcome you and also to lead the AI challenge today. In fact, plenty of us every day read news, listen to the radio, watch TV, and in fact, we have plenty of information that we try to analyze and align. But my main question is: do we have a huge amount of time to work under this news to deeply analyze them? In fact, I would say no.
Before we move on to the main topic, someone can say: “Okay, but I can ask ChatGPT, I can ask Gemini to provide me a list of news and also to summarize them.” And I will say yes, of course, it is possible. But there is one major problem: these models can hallucinate. Really, I read a couple of papers last time and there is plenty of information from experts in the subject of LLMs that these hallucinations are too often. So, are you really sure that this information is the one that you would like to read?
So, what I would like to propose is to build a tool right now—a simple AI agent that will not only collect the news from the topic you would like to see, but also ensure they are reliable sources and provide a summary. Before I start my work, I would like to ask you to join me in the selection of the topic via this QR code.
(Audience selects topics)
Okay, let’s try to concentrate on these two: “Regulatory/GDPR/AI Act” and “Business”. To make it a little bit harder, could you please tell me what kind of summary you would like to see?
Audience Member: An abstract with sources.
Maciej Sankowski: An abstract with sources, okay. Anything more?
Audience Member: By region.
Maciej Sankowski: By region, okay. I suppose that we have everything right now. Thank you.
Krzysztof Szabelski: So, big round of applause for Maciej to warm him up for the challenge. Now we can dive into our discussion.
Let’s start with something simple but not that simple. Lasse Wollatz, I know you have experience with the transformation of large financial institutions. Based on those projects, can you share with us a clue when the modernization is really necessary for an organization and when this is just a buzzword?
Lasse Wollatz: Sure. I think the easiest thing is that modernization is necessary if your processes or your legacy systems start to slow you down or slow down your value creation. You mentioned before that there are so many companies running around with really old systems and they still work… so, do they really need to change them? Apparently, they don’t. But sometimes there are reasons why you still should change them.
We have an overaging population here in Germany. I know many customers who have this very real problem that their systems cannot be maintained anymore because there is no one who can do the job. There is “Bob,” he’s 60, and he does a great job, everything runs. From the outside, it looks like a running system, don’t change it. But he’s 60, in 3 years he wants to retire early, and you look at the market and there is no one who can do this job because it’s Perl code, or Fortran 60 code, or whatever. So that’s the part where you need to think about modernizing a system even if it is running.
Krzysztof Szabelski: And you mentioned earlier to me, the products we build today, that’s the legacy systems of tomorrow.
Lasse Wollatz: That’s obviously true. But if we start building things today which add up to a larger goal, then the time at which they become the legacy system can be extended much further.
Krzysztof Szabelski: Okay. Peter Zoller, you have extensive experience in managing transformation projects and you mentioned a lot about the human side. Most projects fail because of humans. How is that influencing what we should and shouldn’t do?
Peter Zoller: I think this is a valid factor for any modernization or transformation project. Since decades, there are studies that clearly say: whenever you have a transformation project and the project fails, in 70% of those failures, it’s not technology that is the problem. The problem is a missing strategy, bad communication, or leaving people behind.
To give an example: I was at BMW during the transformation from waterfall to agile. We tried to transform how vehicle software is developed. We were in a meeting, we explained how agile works—user stories, sprints—and they agreed. Then a general manager raised his hand and said, “That’s fine, and now we have to talk about our change process.”
I said, “I don’t think we need that.” In agile, a change is just another user story. But 5 minutes later, at the end of the meeting, this guy raised his hand again and said, “Ah, and what we did not cover today: next time we need to do the change process.”
In that moment, I realized how naive I had been thinking that if I explain a new process and the person understands, my job is done. The “change process” is deeply rooted in the automotive mindset because late changes are expensive. Your mindset determines your thoughts, your thoughts determine your language. I realized I had to start changing the language first.
Krzysztof Szabelski: So, from what you’re saying, when we talk about modernization projects, we should be talking about a change project?
Peter Zoller: Even worse. Even in the phase of preparing and designing the scope, you need to involve the people that in the end have to work with those systems. You should not let the tech guys design the scope on their own.
Krzysztof Szabelski: Thank you. Rolf Löwisch, you have experience modernizing and deploying systems. What drives the balance between reliability of legacy systems and the drive for innovation?
Rolf Löwisch: Well, I’m from IBM, so I think we are at fault for all those legacy systems we have! But seriously, reliable core systems—what we call legacy—are what you interact with every day: ATM withdrawals, supermarkets, flight bookings. Reliability builds trust. Old systems are not per definition bad. In fact, we have never sold so many mainframes as we did last year, so there is still a demand.
However, complexity has grown. You have an API that has been there for 30 years, written in a language nobody knows, so nobody wants to change it. But business changes, so you want to be flexible. Modernizing does not necessarily mean creating everything new. It can mean a mixture: keeping the core system that fulfills a specific reliable service, but building new microservices or processes on top of it.
Krzysztof Szabelski: That makes sense. Question is, where are the challenges?
Lasse Wollatz: Data. I haven’t found a single company yet which has clean data. If you think you have clean data, you’re an illusionist. But you need to have data available.
Rolf Löwisch: I would add to the data point. When we talk about data and modernizing, there’s an understanding that data is on the cloud. It is not. More than half of enterprise data is not on any cloud; it’s on hard disks in data centers.
Peter Zoller: I would add metadata. Sometimes the real knowledge of what the data origin is, is in the heads of people. If those people leave, you lose that. Secondly: Psychology. In large corporations, it’s easier to get funding for a “new cloud architecture” (bullshit bingo) than to ask for €100k to extend a legacy system, because legacy isn’t “sexy”.
Krzysztof Szabelski: So how do we decide? Go big and modern or take small steps?
Rolf Löwisch: Modernization itself is not a strategy. You need a bigger vision. For example, at IBM, we started applying AI in HR. Initially, satisfaction dropped because the chatbot was poor. But we stuck to it because the goal was to change the operating model. We connected the AI to backend systems so it could actually do things, like transferring an employee. Now, we answer 94% of requests via the agent. This allowed us to change the backend systems (ripping out Workday for SuccessFactors) without the users noticing, because everyone was using the frontend chatbot.
Krzysztof Szabelski: Basically, having a vision and going there step by step.
Rolf Löwisch: Yes. And we learned a lot about governance and risk along the way.
Krzysztof Szabelski: Let’s touch on AI. Is AI a driver to modernize now, or is it optional?
Lasse Wollatz: That’s a tough one. We don’t quite know where we will end up with AI. But to start with AI, you need clean data. It’s not just about data, but processes. It’s like the industrial revolution: first we replaced horses with steam engines, but the real change happened when Ford built the assembly line around the machine. We need to prepare people.
Rolf Löwisch: AI is too powerful to ignore. There is an early adopter advantage because you go through the learning curve.
Krzysztof Szabelski: But 90% of organizations say they don’t see benefits yet.
Rolf Löwisch: Because they need to find the ROI. Often the big value is just cleaning the data first. Or focusing on user-centricity.
Peter Zoller: My hope is that AI is the first technology that will overcome the issue of humans having to adapt to the machine. AI offers an interface that speaks the language of the user. This is the chance to throw over the old way where users have to learn new menus every update.
Rolf Löwisch: Totally agree. It’s often the unsexy use cases, the boring stuff nobody wants to do, that create immediate value.
Krzysztof Szabelski: (Checking on Maciej) Are you finishing?
Maciej Sankowski: Finishing.
Krzysztof Szabelski: Looking 5 years ahead: do we design systems to be self-sufficient or to help people?
Lasse Wollatz: Both. We need automation for the aging population, but also easier interfaces for humans.
Rolf Löwisch: In 5 years? Nobody knows. But look at the public sector—pension funds or justice systems. AI helps speed up processes like reading documents, though humans still decide. Also, back to the HR example: our HR experts transformed into change managers and digitalization experts.
Peter Zoller: It’s a question of mindset. If your mindset is to use AI to fire 50% of people to be efficient, your competitors will do the same. But then, who are the customers who will buy your products if 50% of Germany is unemployed? Leaders have a task to help people develop new skills.
Krzysztof Szabelski: Is AI helping us speed up modernization delivery today?
Lasse Wollatz: Yes, it’s helping (e.g., reading legacy code), but it’s not a silver bullet. You still need to bring the company and data along.
Rolf Löwisch: AI coding helps, but developers need to understand architectures and frameworks. I think the bigger potential is in “agentic” workflows—a new layer on top of existing applications that integrates them flexibly. The next big learning step is governance and monitoring of these agents.
Krzysztof Szabelski: Last round of thoughts. One thing to remember?
Lasse Wollatz: Start today, don’t expect results tomorrow. And don’t just buy the first thing with “AI” in the name.
Peter Zoller: Start now to test it, get a feeling for what is possible. Focus on user interfaces.
Rolf Löwisch: Be curious. Start your learning curve. Look for the unsexy use cases where there is value.
Krzysztof Szabelski: Good. Thank you very much.