On Wednesday, April 1, I attended GenAI in Zurich for the first time. To be honest, I didn’t know exactly what to expect. It quickly became clear: this event is technical. Very technical. And yet—or perhaps precisely because of that—I took away several things that have stayed with me since.
AI is the big umbrella over everything that seems intelligent—an AI that filters spam, classifies images, or delivers recommendations. GenAI is the cool cousin that actually invents something new. Whether text, image, or code. Instead of just sorting what exists, it creates (or rather: generates) something new. Hence the name: Generative AI.
Yes. And this is more important than many think. An AI hallucination occurs when the model writes something that is false or simply made up. But it sounds so convincing and fluent that you believe it. The AI does not verify facts. It is trained to form sentences that sound logical. The result: It lies with great conviction, without realizing it itself.
What this means for you: Anyone who blindly adopts AI outputs—whether in marketing, customer service, or consulting—bears the risk themselves. AI is a powerful tool. But it always requires a human who thinks critically.
GenAI Zurich is divided into keynotes in the main hall, use-case presentations from companies in practice, a hackathon, and an exhibition area where you can speak directly with the presenting companies. A well-balanced format that enables both insights and direct conversations.
Throughout all rooms, all presentations, and all conversations, one insight emerged that sounds so simple yet is so rarely implemented:
AI alone is not useful. It requires a concept behind it.
Logical, right? Apparently not. Various speakers independently came to the same conclusion: around 95 percent of AI projects fail not because of the technology—but because companies start with the tool instead of the problem. An AI strategy is not an IT project. It is a transformation process that begins with an honest question: “Where are we currently losing time, money, or quality?” And not: “Which tool should we buy?”
The gap between success and failure does not lie in the technology. It lies in the absence of a clear roadmap beforehand.
I was mainly in the main auditorium and the use-case room. The presentations in the auditorium were very technical. And I found one presentation on IT security particularly relevant.
Timo Bozsolik-Torres from SandboxAQ and Drilon Balaj from XY Cyber addressed something I implicitly know but was not sufficiently aware of. Shadow AI is the phenomenon when employees use AI tools in an uncontrolled manner. Without IT’s knowledge, without guidelines, often with real customer data. Most companies have no tools to even detect this. And what they cannot see, they cannot protect.
In the talks, there was a clear position: The solution is not prohibition. Anyone who starts banning has already lost. Because usage simply goes underground within the company. The correct sequence is: first create visibility, then define guidelines, then offer secure alternatives.
For SMEs specifically: Do you have a clear rule about what employees are allowed to do with customer data and AI tools? If not, that is your first step. Not tomorrow. But now.
In the use-case room, practical examples were shown from companies that do not view AI as an experiment, but already use it productively. Here is my personal summary of the most interesting contributions for me.
Migros has built an internal chatbot platform that allows teams to create AI assistants in just a few minutes. With high standards for security and data protection. Their biggest learning: The technology was never the problem. Stakeholder management and governance models took more time than the actual development. Especially a solution that meets both business needs and security standards. First concept, then tool. This was mentioned here again as well.
IKEA is dealing with a supply chain that has grown over 80 years. They are now gradually implementing Agentic AI to relieve employees in their daily work. IKEA has also not delegated everything to an AI. Not everything is deliberately handled with AI, only individual tasks, as they cannot yet fully rely on AI.
Zalando has shown how AI understands real customer needs. When someone searches for “What should I wear to an 80s party?” that is no longer a classic search. It is a moment, a feeling, a context. Zalando translates exactly that into concrete product recommendations using AI.
One statement occupied me more on this day than all the tools and demos combined. It came casually, but it stayed with me:
“AI gives you two options: You generate more revenue with the same team. Or you generate the same revenue with fewer people.”
Which option you choose is a question of strategy and values. But the decision is coming.
Three questions I ask myself after this day—and that I would also like to share with you:
AI is not a cure-all. But those who use it strategically—with a clear goal, clear rules, and the right tool—can achieve significantly more with the same team.
Do you have questions on this topic or need consulting / support? Contact us.
