This week I wrote for Tagesspiegel Background that Germany is having the wrong AI debate. We focus on infrastructure and models, while the real decision happens at the company level: can we turn AI into measurable value in our operations?
That’s the question I could only hint at there. Here I answer it.
Why two-thirds of German CEOs don’t see an AI impact
The numbers are clear. 94% of generative AI pilots deliver no measurable return. Two-thirds of German CEOs report they see no measurable effect from AI. Only 11% achieve measurable revenue gains. 42% name insufficient employee enablement after implementation as the main problem.
- 94% no return – GenAI pilots with no measurable ROI
- 2/3 of CEOs – see no measurable AI impact
- 11% revenue lift – achieve measurable revenue increase
- 42% enablement gap – cite insufficient employee enablement as the main issue
But this diagnosis is misleading.
Not because enablement is unimportant. Because “more training” is the wrong answer to the wrong question.
The real question is: what have companies actually deployed? A tool. Or a new way of solving problems?
The misunderstanding that blocks everything
What most people mean when they say “introducing AI”
License Microsoft Copilot. Run a few trainings. Launch a pilot with one department. Then hope the investment pays off.
This is the old technology mindset. Push a button, expect results. With many previous technologies it worked: roll out a CRM, train staff, make processes more efficient. Done.
AI works differently.
Why the old mindset fails with AI
AI is not a classic IT rollout. It’s an enablement technology. Its value doesn’t appear right after deployment. It emerges from people’s ability to apply it to specific challenges.
When AI stays abstract, it mostly produces one thing: excuses.
“We can’t.” “Too risky.” “Data protection.” “No use case yet.” “We need IT approval first.”
The result: endless pilots. Investments without impact. Employees who feel AI creates more burden than relief.
What really matters: application intelligence
What is application intelligence?
Application intelligence is the ability to spot real, day-to-day problems and judge whether and how AI can help. Not “What can the tool do?” but “Which of my current problems will it solve better than before?”
That sounds simple. It isn’t.
Past know-how and familiar problem-solving approaches only hold up so far in the AI era. We have new possibilities. To use them, we must first remove the barriers we built ourselves.
The barriers you’ll usually find
I see two recurring barrier types in companies.
The first is structural-operational: processes designed so AI has no natural place. If every AI use case requires an IT ticket, AI will never enter daily work. This is not a technical issue. It’s a design problem.
The second is cultural. Beliefs that are honest reflexes to a technology that hasn’t yet shown how it concretely helps: “We don’t have capacity.” “This isn’t for our company.” “I’m not technical.”
The path to application intelligence runs through both. Adjust structures and enable people. Not when capacity appears, but so capacity can appear.
Efficiency is the path. Not the goal.
What do we really need efficiency for?
I disagree with a common claim: adopting AI for efficiency isn’t wrong. But it’s not enough.
Efficiency is the path. The goal is value creation, innovation capacity, and competitiveness.
When AI speeds up operational tasks, it creates something more valuable than saved hours: space. Space to pursue things that previously lacked time. To explore new ideas. To serve customers better. To improve products.
Simply doing the old things faster isn’t sufficient. If you introduce AI only as an efficiency machine, you miss the actual lever.
What the data shows
87% of German mid-sized firms plan to expand AI. Only 47% have started upskilling programs. That shows intent — but it’s disconnected from the goal.
Companies running structured AI upskilling more than double their ROI: 42% versus 21% for companies without structured programs. The difference isn’t the technology. It’s enabling people.
- 87% expansion plans – German mid-sized firms plan AI expansion
- 47% upskilling – have started upskilling programs
- 42% ROI – with structured AI upskilling
- 21% ROI – without a structured program
Bringing people along as a competitive advantage
Why AI adoption is also an employer issue
One aspect often missing from strategy papers: what happens to people when they experience a company managing this change with them?
The goal isn’t to pull employees away from their expertise with new tools. It’s to rely on the same people and their expertise — and give them new ways to achieve more with it.
Employees who control AI and see their knowledge become more effective through AI don’t feel replaced. They feel empowered.
What that means in practice
Two-thirds of organizations are not culturally prepared for AI transformation. Only 53% of employees feel prepared. This isn’t a technology problem. It’s a leadership problem.
Companies that invest in enabling their people now will be more productive in two years and more attractive as employers. In a market where AI skills become hiring prerequisites and two-thirds of HR leaders no longer consider candidates without AI knowledge, application intelligence is not a nice-to-have. It’s a basic requirement.
- 2/3 organizations – culturally unprepared for AI transformation
- 53% employees – feel prepared
- 2/3 HR – no longer hire candidates without AI skills
Three steps that decide now
How to succeed with AI adoption in practice
- Identify bottlenecks, don’t chase hype The most common mistake in AI projects: start with the technology. Start with the problem. Which three tasks in your team take the most time and add the least value?
- Deliver the first concrete use case Don’t just evaluate. Don’t park it in a pilot. Deliver it.
- Learn from delivery, not from training Application intelligence doesn’t emerge in a classroom. It grows from real experience with real challenges. Teams that apply AI to actual problems build what lasts: the ability to solve the next problem themselves.
1. Identify bottlenecks, don’t chase hype. The most common mistake in AI projects is starting with the technology. Start with the problem. Which three tasks in your team take the most time and add the least value? That’s where real AI impact begins, not in the showroom.
2. Deliver the first concrete use case. Don’t just evaluate. Don’t leave it in pilot mode. Deliver it. On a real task, with a real team, in a real process. The first successful use case matters most because it proves it can work and builds the trust needed for everything that follows.
3. Learn from delivery, not from training. Application intelligence doesn’t form in a seminar room. It forms through concrete experience with concrete challenges. Teams that apply AI together to real problems build the lasting capability to solve the next one themselves.
Frequently asked questions about AI adoption and application intelligence (FAQ)
What does “application intelligence” mean in day-to-day work?
It’s the ability to analyze real tasks and bottlenecks so you can decide whether and how AI helps. Instead of debating tool features, the focus shifts to the problem to be solved. Teams assess value, risks, and fit with their own value chain.
Why aren’t classic trainings enough for AI adoption?
Trainings teach tool knowledge. Impact appears only when people apply AI in real processes. Without structural changes and learning loops, what’s learned remains abstract. That creates pilots without transfer — and frustration instead of progress.
What’s the difference between efficiency and impact with AI?
Efficiency mainly frees up time by speeding routines. Impact appears when that time is invested in value creation, innovation, and better customer experiences. Making the old faster leaves the real lever unused.
Which barriers prevent AI impact in companies?
Operational designs that give AI no natural place, for example cumbersome ticket processes, often slow things down. Cultural beliefs like “no capacity” or “not our topic” also block progress. Both require action: adjust structures and intentionally enable people.
How do I start pragmatically with the first use case?
Start at the bottleneck, not the tool: which tasks consume time and deliver little value? Implement a real use case with a real team in an ongoing process and keep learning cycles short. This builds trust and a repeatable pattern.
What role do leadership and HR play in AI adoption?
Leaders address cultural and organizational readiness — not just technology. Companies that invest in enabling people and set clear guardrails increase productivity and employer appeal. HR increasingly makes AI competence a hiring requirement, which helps steer direction.
Conclusion
Germany is running the wrong national AI debate. But at the company level, the decision is ours.
The crucial question is not whether we can build AI. It is whether we can make our organizations and people AI-capable. That doesn’t require new tools or another pilot. It requires a new methodology and the courage to apply it.
Those who start today will not only gain efficiency. They will create the space for what comes next.
Interested?
Let's find out together how we can implement these approaches in your organization.
Schedule a conversation now