Stefan Latuski on transforming public sector services at scale

| Interview

With more than 100,000 employees and millions of citizens relying on its services, Germany’s Federal Employment Agency (Bundesagentur für Arbeit, or BA) operates at a level where even incremental changes can have far-reaching effects. But demographic changes and rising demands on the organization made it necessary to fundamentally rethink how its processes and services work.

Stefan Latuski, BA’s chief information officer (CIO), is leading that effort by addressing the complexity of the agency’s legacy systems rather than focusing on the technology alone. As part of McKinsey’s Rewiring the public sector series, Latuski speaks with McKinsey Partners Dominik Hepp and Katharina Sickmüller to discuss how BA is simplifying and standardizing its ways of working, using AI to accelerate modernization, and creating the organizational conditions for technology to deliver.

This interview has been edited for length and clarity.

McKinsey: Your organization has set a bold ambition to become Europe’s most modern public sector service provider. For an organization with more than 100,000 employees that serves millions of citizens every day, what does that ambition mean in practice?

Stefan Latuski: The Federal Employment Agency is currently facing a multitude of different challenges. On the one hand, if you look at the economic situation in Germany, unemployment rates have been growing significantly over the past couple of years.

On the other hand, we have a significant demographic change happening within the Federal Employment Agency. Roughly 35 percent of the entire workforce is going to go into its well-deserved retirement within the next few years. That means there will be way more work to do and way less of an available workforce to cope with that work.

That also means that we have to automate and digitize nearly all of our business processes. There’s a saying: “It’s better to change yourself rather than being changed by external factors.” So we introduced a huge transformation program a couple of years ago: the Future Model. That program is about analyzing nearly every single business process, deriving optimization measures from that, and then implementing various activities steadily and gradually.

McKinsey: So technology is at the core of bringing that ambition to life and making it a reality: with AI, with cloud, with modernizing the core systems and core platforms underneath. With all of those moving parts and that development, especially on the technology side, what are your main priorities to make the transformation a success?

Stefan Latuski: Without technology, all of this is obviously not going to work. For us, it’s two topics. First, artificial intelligence is one of the key technologies we believe has the capability to transform nearly every single business process. If you take a look at a high-level business process of the Federal Employment Agency, it’s applying for services, then proving the applicant’s eligibility for a service, and then paying out the benefit. So nearly every single step of that business process can be automated using artificial intelligence.

Second, given the size of the organization—we have more than 100,000 employees—there is no way to automate other than using cloud services. But considering we are part of the critical infrastructure in Germany, there are a ton of different laws and regulations that can create a high barrier to entry for using the cloud.

We have to cope with all of those challenges. But over the past couple of years, we have still found a good way to introduce various new technologies and make that transformation happen.

McKinsey: Every transformation comes with its flagship programs and initiatives. Is there one example of a program that captures the essence of your transformations with cloud and with AI?

Stefan Latuski: CONTACT is a project to migrate a multitude of communication templates—nearly 50,000—that have been stored in a legacy application called BK-Text (see sidebar “Modernization in action: More about the CONTACT program”). The program is about migrating that legacy application into a new environment. Theoretically, we could migrate 50,000 documents, document by document, into a new system. Honestly speaking, we tried that twice before and failed. So we were aware right from the beginning that this time we had to go a different way to be successful.

Together with McKinsey, we started that program approximately one and a half years ago and found a way to handle this problem completely differently. Instead of migrating one document at a time, we went the way of significant standardization and gradually moving away from a document transformation toward more of a business process transformation. We introduced processes and workflows and found commonalities across the different communication templates. So instead of migrating 50,000 different documents, we ended up migrating a few hundred.

McKinsey: CONTACT ended up being much more than a technology migration. It seemed like a simplification effort. What did that teach you?

Stefan Latuski: We were coming from a world of roughly 50,000 different communication templates, which we transformed into a few hundred communication processes. That requires a significant amount of standardization.

Honestly speaking, standardization is not only a technology challenge. More and more, it is becoming a challenge of changing people’s behaviors and mindsets. CONTACT taught us that we have to go down that path of change management, which we kind of learned the hard way.

McKinsey: AI has been part of every step of the migration and is now transforming the platform itself. Where has it created the most value for you?

Stefan Latuski: We used artificial intelligence in nearly every part of the project (see sidebar “Accelerating modernization with AI”). It’s hard to say there was just one thing or activity that brought us the most business value. But there is a multitude of activities where it had great impact. For example, we used artificial intelligence to analyze the content of those 50,000 different templates, which otherwise would have required a significant amount of manpower.

The legacy system, which was more than 20 years old, consisted of millions of lines of code. So we also used artificial intelligence to analyze the code, derived the business processes out of that code, and then translated those processes into a new system environment.

McKinsey: Looking ahead, agentic AI is the next big step. Are you already exploring how it can help you in your transformation?

Stefan Latuski: Agentic AI is one of those key technologies, especially in the public sector. Our bread-and-butter business is verifying documents, proving eligibility for any kind of service, and then delivering those services and benefits accordingly.

In this case, every single business process step could be transformed by using artificial intelligence, not only generative AI—in terms of analyzing documents—but also agentic AI to automate those processes and reduce manual efforts to an absolute minimum. This is one of the technologies where we are placing our biggest bets.

McKinsey: With the growing use of AI, what are your thoughts about risk management and governance in the public sector?

Stefan Latuski: Our role in the public sector requires us to use data and technology highly responsibly and to ensure that no sensitive information is leaked or used without people’s consent. For us, using AI in a responsible way is not about avoiding it. Our focus is on creating a secure environment in which AI can be used confidently and at scale. For our organization, this means that governance and data protection are built into the processes where we use AI from the outset.

Our first live generative AI use case on the cloud shows what this means in practice. The anonymization runs on premise—so technically, no personal data ever leave our own environment—and those data are processed only by models inside the European data zone. We walked that complete data flow by our data protection unit, the staff council, and the federal data protection commissioner, step by step, until everyone could see exactly where a given citizen’s data sits at any moment. Of course, establishing this approach took time, but it was worth the effort. We now have a pattern we can reuse in the future instead of reopening the same discussion for every new use case.

McKinsey: Based on your experience, what lessons would you share with other public sector leaders who are embarking upon a digital transformation at this scale?

Stefan Latuski: First, a transformation should be approached as a simplification effort rather than a technology replacement exercise. The greatest value often comes from reducing complexity before implementing new systems.

The second is that business transformation, technology transformation, and human-centric change management need to move together. Separating them creates friction and limits a transformation’s potential impact.

Third, AI should not only be viewed as a future capability. It can also be a powerful accelerator of the transformation itself, as it is happening, by helping organizations understand and modernize complex legacy environments.

And fourth: Build every program so that the next one inherits it right from the start. CONTACT, for example, was never only about templates or customer communication; rather, it was about creating significant productivity improvements and a dramatically simpler foundation for future innovation. The cloud setup, the anonymization service, the agnostic data model, the integration layer, and the AI tool chain are all reusable and are now building the starting point for modernizing our core benefits systems—where even greater productivity gains are possible for an organization of our size.

Explore a career with us