While the wave of change driven by AI is gaining momentum, many small and medium-sized businesses and government agencies are still standing on the shore, undecided. Jürgen Weber, founder of weber.cloud in Balingen and a member of the supervisory board of the Baden-Württemberg AI Alliance, sees this firsthand every day in conversations with clients. We spoke with him about why real AI projects fail, what a meaningful first step actually achieves, and why free credits are often the most expensive option.
An Interview with Jürgen Weber:
Mr. Weber, you say that German small and medium-sized businesses haven't really taken the changes brought about by AI on board yet. What do you mean by that?
Imagine a tsunami. At first, the ocean is completely calm; it even recedes a little. Everything seems normal. And then the wave hits—at a pace that makes it impossible to reach safety if you’ve waited too long. I know that’s a harsh image. But that’s exactly the pace we’re experiencing right now with AI. Most companies haven’t even noticed the wave yet. Any company that doesn’t have an AI strategy and comprehensive AI-based automation in two or three years will simply cease to exist in many industries. That may sound dramatic. But in my conversations with clients every day, I see just how far behind we are in this area.
So where exactly do German small and medium-sized businesses really stand right now?
We see three patterns. Some companies are doing nothing at all. Others have individual employees who have long been using ChatGPT or Claude, but in an uncontrolled manner—often bypassing the IT department—and storing company data in third-party systems. And then there is a small group that takes a structured approach. This third group is significantly smaller than it should be. A recent study shows that 74 percent of medium-sized companies are still in the very early stages of AI adoption. Among companies with 20 to 99 employees, just under one in four officially provides AI tools to its employees.
That sounds like there's huge potential for "shadow AI"—that is, AI use that doesn't officially take place at all.
That’s exactly what’s happening. According to a 2025 Bitkom survey, the use of personal AI tools at work is widespread in 8 percent of German companies, and there are isolated cases in another 17 percent ( Source: Bitkom Research, representative survey of 604 companies with 20 or more employees, October 2025 – bitkom.org). The percentage of employees who use AI secretly has doubled within a year. This is understandable. The tools are available, they provide immediate help, and no one asks for permission first. But this is precisely how company data ends up in systems over which the company has no control. This isn’t a theoretical risk—it’s everyday reality.
What advice do you give to companies that want to resolve this situation? Where do you start?
The first sensible step is almost always the same: a chatbot powered by company knowledge. That might sound like an old hat, but it’s incredibly effective. You take everything your company knows—including the quality management manual, operating instructions, design information, and business data—and feed it into an AI model. Then you can ask questions about your own company, citing sources so that every answer is verifiable. We often do this as a proof of concept. Customers are blown away when they see what’s possible for the first time.
Why does this simple use case, of all things, work so well?
Because AI only realizes its full potential when the company feeds it all its critical data—not just individual documents or texts. Business data, patents, knowledge databases—everything. Only then does the system become truly useful. And that’s exactly why the question of where this chatbot runs is so crucial. You don’t want this data to be stored in a U.S. cloud where U.S. authorities can access it under the CLOUD Act. But you do want it in a 100% sovereign German environment.
You mention pilot projects that comply with data protection regulations. What happens if this issue is underestimated?
Then exactly what we’re seeing so often these days happens. A company launches an AI project—perhaps even a successful one—and at some point, the data protection hammer comes down. The project is shut down. This not only kills the individual project but also damages internal acceptance of further AI initiatives for the time being. This often sets companies back a year. So here’s my advice: Raise awareness, develop a strategy, and only then get started. And make sure to comply with data protection regulations from the very beginning—especially for a German company without an American parent company.
Many U.S. providers offer very attractive introductory deals—free credits, low token prices. Why are you skeptical about that?
Because it’s a proven strategy and has nothing to do with charity. You’re familiar with this pattern from other areas of the cloud world: lure customers in with low prices, integrate deeply, then crank up the prices. That’s exactly what’s happening right now with AI. Jensen Huang said as much himself at Nvidia GTC 2026. The new high-speed models are expected to cost $45 per million tokens, while the premium tiers will cost $150. Currently, you’re paying a fraction of that for comparable models. Anyone who builds their entire business model around these tools today will pay dearly for it in 1–2 years.
What is your recommendation for the first few weeks? What specific steps should small and medium-sized business owners take right now?
Three things. First: Raise awareness within your own company. Who is already using AI, for what purposes, and with what data? This overview alone is valuable. Second: Define a pilot project that delivers clear benefits and is set up in compliance with data protection regulations. A chatbot powered by company knowledge is almost always a good place to start. Third: Explore options beyond U.S. hyperscalers and U.S. AI model providers. There are regional, sovereign alternatives. We’re one of them, but not the only one.
What is your mission as a member of the Baden-Württemberg AI Alliance?
I want AI to take root in the minds of regional businesses and government agencies so that the tsunami-like wave doesn’t sweep everyone away. The German economy isn’t exactly in the best shape right now anyway. And this transformation hasn’t really taken hold yet. When it does, it will affect many. That’s exactly why the AI Alliance is important. It brings together businesses, local governments, and academia—regionally, in a networked way, and on equal footing. That’s exactly what’s needed right now.
About Jürgen Weber
Jürgen Weber is the founder and CEO of weber.cloud, based in Balingen. The owner-managed company has been operating its own data center for over two decades and is one of the few German providers with a fully sovereign infrastructure—ranging from an autonomous system to its own fiber-optic link and hardware. Since 2010, weber.cloud has also been building an infrastructure connecting Europe and China. Today, the company operates its own sovereign AI cloud based on freely available models. weber.cloud is a member of the Baden-Württemberg AI Alliance.
Jürgen Weber has been a member of the supervisory board of the Baden-Württemberg AI Alliance eG since June 2026.