Marc Schlingheider: “AI projects are transformation projects.”
The AI pilot is up and running. The first use cases have delivered promising results. But what comes next? Many organisations find themselves at exactly this stage. Marc Schlingheider explains what it takes to turn AI initiatives into lasting business value, why data and security matter from day one, and when organisations can realistically expect to see a return on their investment.
Marc Schlingheider leads IBM’s Data and AI business across the DACH region. Together with his team, he helps customers put AI to work, from harnessing real-time data and managing data effectively to establishing governance frameworks and integrating AI into business processes. He works with organisations as they move from initial AI pilots to measurable business outcomes.
Marc, what fascinates you about data management?
Marc Schlingheider: Data is the gold of the 21st century. Every day, billions of new data points are added, and the quality of that data plays a huge role in how successfully a company can grow and evolve. Whether it’s customer data or insights into market developments, data sits at the heart of every business. Companies need to be able to access it and put it to work. What I find particularly exciting is that organisations are at a pivotal point right now. They’re having to make real strategic decisions about investing in data and AI and about how those investments will shape the future of their business. I’m convinced that the companies that make the right decisions today, and take a serious look at what AI means for their business model, will be in a very strong position going forward.
What is currently top of mind for you in your role leading Data and AI at IBM?
A lot is happening in the market right now. Organisations have reached the point where they’ve recognised the potential of AI in their business models. The challenge now is to make AI an integral part of the way those businesses operate. Companies need to move quickly. They want to remain competitive and successful, which means making AI capabilities available in real time, supporting analysis and planning, and using those insights to make better decisions. At IBM, we’ve been investing in agentic AI and real-time data from an early stage. We recently completed the second-largest acquisition in the company’s history with the purchase of Confluent, a specialist in enterprise-ready real-time data. That enables organisations around the world to make decisions immediately, based on what is happening in their business at that moment.
Can you give us a practical example?
Take a major German automotive manufacturer, for example. Every day, across its 32 production sites, machines and people generate around three billion data points. The key is being able to use that data immediately. The company has successfully implemented a highly cost-effective solution that not only makes the data available, but also analyses it in real time.
So we’re talking about robots assembling cars?
Exactly. In this case, systems recognise when materials or tools are needed on the production line and where they’re required. An agentic AI system then sets the process in motion in the background, and an autonomous robot delivers the required part directly to the workstation. There is no need for anyone to step in or manually initiate the process. The same applies to related processes further down the line, such as ordering replacement parts. Whether the entire chain can be automated ultimately depends on having the right data available at the right time. When data is generated in milliseconds and can become outdated almost as quickly, that creates significant challenges for organisations.
Are there any other real-world examples of how real-time data analysis can create value?
L’Oréal, for example, uses Confluent’s technology to provide customers with real-time product availability information in its online shop. That data comes directly from the manufacturer, but it also needs to be accessible to third-party retailers such as Sephora and Carrefour. Managing that data and providing customers with reliable information about product availability is a huge challenge. The goal is to be able to tell customers not only whether a product is in stock, but also which retailer can get it to them fastest.
Many organisations have invested in AI and, understandably, expect to see a return on that investment. Yet according to a Forrester survey, a quarter of planned AI spending will be put on hold by 2027 because the expected benefits haven’t materialised. Why do you think that is?
That doesn’t surprise me at all. Over the past few years, we’ve launched some excellent AI projects with customers that were ultimately shelved. The main reason is that AI projects were often viewed as technology projects rather than transformation projects. By that, I mean recognising across the entire organisation that real change is needed and that AI has the potential to reshape the business model. Transformation isn’t something that only concerns the IT department. It affects everyone, from the CEO to the procurement team. Alongside that strategic aspect, organisations also need to consider the cost. AI projects can become very expensive very quickly. That’s simply something organisations need to factor in
How can organisations approach that?
My advice is to look at how others have done it. Start small, experiment, invest a bit of money, build a pilot and see what happens. How do people respond to it? Where are the gaps in the system? The AI itself will generally work. The real challenge is making the data available. If the data is locked away in silos or hasn’t been prepared in a way that AI can access and use it, the project won’t deliver the expected results.
IBM has also undergone a major AI transformation. Can you tell us about that?
A few years ago, we launched AskHR, our AI-powered HR assistant. Today, we’ve automated 94% of all HR processes. One of the best examples is promotions. In the past, the process could take a long time because of the way a large organisation is structured. With AskHR, we’ve reduced the time required by 75%. In practice, I simply enter my request in German or English through the user interface, have a brief conversation with the assistant, and the agent sets the process in motion. The same applies to transfers, pay rises and sick leave reporting. It covers the entire employee lifecycle, from hiring through to departure. And the results have been phenomenal. Over the past two and a half years, those productivity gains have amounted to 4.5 billion.
Once you’ve seen success with one use case, what’s the next step?
Once an organisation has committed to the transformation and seen the benefits in a specific use case, it can start looking at other areas where agentic AI can add value, such as IT support. A sales manager, for example, might use the same interface to obtain pricing information or handle more complex enquiries. My advice is simple: start with one use case and build from there.
How do you address AI-related concerns, such as the risk of bias in HR processes?
Alongside the ethical considerations, it’s just as important to think about security. That’s what we refer to as enterprise readiness. We need to consider who has access to the platform and who can access sensitive information. We’re often dealing with highly sensitive data, and if something goes wrong, the consequences can be significant, potentially even threatening a company’s future. Organisations need to think not only about data quality, but also about data security. From the outset, AI systems need to be designed with strong governance and the right controls in place to ensure they are used securely and responsibly. At IBM, these governance capabilities are built into the technology.
Let’s talk about data silos. What advice would you give organisations looking to implement AI?
From my perspective, one of the most important questions is what data we’re actually using and where it comes from. AI is only as good as the data you feed into it, to borrow a well-known phrase: garbage in, garbage out. At its core, every AI project is also a data project. The challenge is bringing the two together. Some organisations take on a major data project and bring everything together in one place, creating what we would call a data lake. For certain applications, that can be the right approach. For the kinds of AI use cases we’ve been discussing, however, organisations often rely on real-time data that remains in different systems. In those cases, the focus is on creating an intermediate data layer, whether that’s a virtual environment or another tool that can securely move data between systems. The AI can access the data through that layer, while the technology in between controls who can access what, and when. In most cases, the data itself remains where it is, and the AI accesses it only when needed. That approach can be costly and complex, but it’s often essential for a successful AI implementation.
What does it take to keep an AI project on track?
The biggest problems arise when people don’t buy into the change. Take a situation where a CEO wants to implement an AI project in sales, but nobody is on board. It’s the responsibility of leadership to explain the value of the AI model and make it clear what a transformation project like this means for employees. That might be reducing costs, improving efficiency or evolving the business model. Only then should the focus shift to implementation. During that phase, it's important to bring people along, get them excited about the project and give them the opportunity to try it out and see how it works in practice. If you get both the communication and the technical implementation right, you’re already a long way towards making the project a success.
How can organisations balance data privacy with the need for AI to access information?
A key concept here is digital sovereignty. At its core, it’s about whether an organisation remains in control of its data at all times. Whatever the use case, maintaining control over both the system and the data is essential. That’s why the AI models we provide are designed so that customers can decide how they want to run them, whether in their own data centre, in the cloud or in a hybrid environment. Most organisations today operate in a hybrid world. The important thing is that they can define the framework themselves and retain control over their data.
What are the three most important things a mid-sized company should focus on over the next 12 months if it wants to make meaningful use of AI?
First, ask yourself what you want to achieve. Define the business value you’re looking for. The use case should follow naturally from that. Second, identify the right tools. That also means working closely with the business teams that own the data and making an honest assessment of whether your organisation has the capabilities needed to take on a data project of this kind. And third, start with a small use case and simply give it a try.