The Dinner Table That Woke Me Up
I was sitting at dinner last week after an intense stretch with AI-first builders—people running agent fleets, connecting live databases, using tools like Claude Code as an execution layer—when someone dropped a line that stopped me mid-sentence. They estimated that fewer than 2% of German professionals are actively using that kind of tooling today.
I can’t verify the exact number, and I wouldn’t stake a business decision on it. But emotionally? It landed like truth.
Because I had just spent a week deep inside a world where people were debating memory architecture and audit agents—and I had momentarily forgotten that most professionals are still weighing whether to use a chatbot to summarize a PDF.
That gap is not a footnote. It is the defining business risk of the next three years.
The Mental Model That Is Already Obsolete
Here is the story most leaders are telling themselves: We gave our team AI access. We bought the licenses. We are covered.
I understand why that feels like progress. A year ago, it was. But the tools have moved faster than the mental models.
What I watched builders do last week had nothing to do with prompting a chatbot. They were connecting tools to real files and workflows, building lightweight applications without a developer, turning a rough intent—“check all these supplier contracts for renewal clauses”—into a working system within hours. No IT ticket. No six-month project. Just someone who understood how to orchestrate the right tools in the right sequence.
The old question was: What can I prompt?
The new question is: What part of this workflow should still be done manually?
That inversion is not subtle. It changes how you think about headcount, process design, transaction readiness, and where your real leverage is.
The Three-Layer Gap
When I look across the SME and Mittelstand landscape, I see the same pattern repeating. Most organizations are stuck at Layer 1. Very few have reached Layer 3.
Layer 1 — Chatbot Thinking AI is a productivity add-on. Individuals use it occasionally to draft emails, summarize documents, or brainstorm. There is no shared structure, no workflow integration, no institutional learning. Usage depends entirely on individual curiosity.
Layer 2 — Process Augmentation AI is embedded into specific workflows—reporting, research, first-draft generation, meeting summaries. There are some shared prompts, some governance, some repeatability. This is where most progressive SMEs currently sit. It creates real efficiency but limited leverage.
Layer 3 — AI-Native Execution AI is part of the operating layer itself. Agents handle file inspection, data checks, workflow routing, and audit trails. Humans supervise, decide, and course-correct—but they are no longer manually executing steps that a well-configured system can handle. The organization builds institutional capability, not just individual fluency.
The honest truth: most DACH companies are hovering between Layer 1 and early Layer 2. Layer 3 is not science fiction—I watched people build at that level last week—but the gap between where most companies are and where the frontier is moving feels wider every month.
Why This Is Also a Transaction Readiness Problem
I work with companies preparing for ownership transitions, capital events, and governance upgrades. And I am starting to see AI capability—or the absence of it—show up as a value and risk factor in those conversations.
Not yet explicitly. But the signals are there.
A business that runs on manual, undocumented workflows is harder to value, harder to transfer, and harder to scale post-transaction. A business where key processes live inside one person’s head—rather than in a governed, repeatable system—carries concentration risk that sophisticated buyers notice.
When AI-native execution is done well, it does something beyond efficiency: it makes the business more legible. Processes are documented by design. Outputs are auditable. Institutional knowledge stops being trapped in individuals and starts living in systems.
That is what I mean when I talk about Fit-for-Transaction. It is not only about clean financials. It is about whether the operating model can survive and scale without the founder or key person in the room.
AI, built thoughtfully, accelerates that transformation. AI adopted carelessly—chatbot licenses with no structure—does almost nothing for it.
A Framework: Manual → Standard → System
I use a simple sequence with the operators and CEOs I work with. I call it Manual → Standard → System, and it deliberately slows people down before it speeds them up.
Manual first. Before you automate anything, do it by hand. Understand the edge cases, the human judgment calls, the failure modes. If you cannot do it manually, you cannot govern it automatically.
Standardize second. Once the manual process is clean and documented, standardize the steps. Build a shared template, a checklist, a repeatable structure. This is where most companies should be spending energy right now.
Then, and only then, build the system. Connect tools, introduce automation, layer in agents. Because now you know what the system is supposed to do—and you will notice when it breaks.
This sequence matters because the failure mode I see most often is companies jumping straight to systems before they have earned the right to automate. They get brittle automation on top of messy processes, and then blame the tools when it does not work.
The tools are not the bottleneck. The operating discipline is.
One Human Moment
I had a conversation recently with a CEO—strong operator, serious about his business, genuinely curious about AI—who told me he had tried three different AI tools in the past year and abandoned all of them because they “didn’t deliver.”
When I asked what he had tried to build with them, he described essentially asking each tool to replace a task he had not first documented or standardized. The tool would produce something imperfect, the team would lose confidence, and the experiment would quietly die.
He was not wrong to be skeptical of the output. But the problem was not the tool. The problem was that he had skipped Manual and Standard and gone straight to System.
We spent an hour walking back through one core workflow—manually. By the end, he had something more valuable than any AI output: a clear picture of where the real complexity lived, and where the genuine opportunity for leverage was.
That clarity is the actual foundation. Everything else is built on top of it.
What You Can Do Next Week
-
Audit your current AI usage honestly. Ask your team how they actually use AI today. How often? For what tasks? Is there any shared structure, or is it all individual and ad hoc? You need a baseline before you can build.
-
Pick one workflow and do it manually with fresh eyes. Choose a recurring process—a report, a research task, a supplier review—and walk through every step as if you are documenting it for the first time. Write down where human judgment is genuinely required versus where it is just habit.
-
Identify your Layer 1 → Layer 2 move. Based on that manual walkthrough, find one step you could standardize with a shared template or structured prompt. Not automate yet. Standardize. Get the team using a consistent approach.
-
Have the “builder vs. user” conversation with your leadership team. Ask them: are we using AI, or are we building with it? What would it take to move from one to the other? The conversation itself is useful, even before you answer it.
-
Protect your humans where it matters. As you increase AI-native execution, be deliberate about where human judgment, client trust, and ethical oversight must remain central. AI expands what is possible. It does not replace what is irreplaceable.
The gap between AI observers and AI builders is real, and it is growing. But the opportunity is also real—and it is not yet closed.
The companies that figure this out in the next 18 months will look very different from the ones that do not. Not because they found a magic tool. Because they did the unglamorous work of making their operations legible, governed, and systematically improvable.
That has always been the work. AI just makes it more urgent—and more possible.
Where does your team sit right now—Layer 1, Layer 2, or somewhere on the way to Layer 3? I’d genuinely like to know what is working and what is holding you back.