Seventeen minutes and fifty-two seconds. That’s how long my camera was dead in the middle of a podcast recording last week — and we didn’t notice until the session was over.
Three options: re-record and throw away a conversation that couldn’t be replicated. Quietly cut to my co-host for those seventeen minutes and say nothing. Or reconstruct the image from the same session and label it clearly.
We chose the third. And the choice itself sat with me for days — because it’s the same choice I see Mittelstand leaders making right now, in a hundred quieter contexts, without ever quite naming it.
The Real Question Isn’t What AI Does. It’s Who Decides.
This week I shared two pieces of content that, on the surface, look unrelated: a discussion about entry-level job decline in Germany, and a disclosure about an AI-reconstructed video frame in my podcast. But they’re asking the same thing.
When something becomes easier, faster, or invisible because of AI — who is accountable for the outcome?
The answer is always a human. But we’re getting better at not acting like it.
The Timeline That Changes the Narrative
The entry-level job decline in Germany started in spring 2022. ChatGPT launched in November of that year.
The hiring freeze came first.
Analysis of Indeed data, reported by Dana Hajek and David Gutensohn for DIE ZEIT, shows that postings for entry-level roles in AI-exposed occupations have fallen by more than two-thirds. Meanwhile, apprenticeships and dual-study programs sit roughly 40% above their 2020 levels. The bifurcation is stark and accelerating.
But here’s what nobody wants to say out loud: the decision to stop hiring juniors was made in boardrooms and budget reviews — by humans — before generative AI was a plausible operational tool. What AI did was arrive just in time to become the explanation.
One contributor in a LinkedIn roundup this week put it precisely: the debate converts a leadership decision into a technological fate.
That sentence is worth reading twice. Because the moment we accept that framing, something important disappears — the accountability for the choice.
I’ve been reducing it to three words: Aufgabe. Urteil. Wer macht’s? Task, Judgment, Who owns it. More on that below.
When AI Makes Things Invisible, Decisions Become Easier to Hide
Back to the camera failure.
When we realized we had seventeen minutes of black footage, I genuinely sat with the options for a day. Reconstructing a face — even your own, from your own recorded audio, using a still from the same session — felt like crossing a line. Christoph and I talked through it carefully.
What settled it wasn’t a policy document or a legal review. It was the subject matter. The episode was literally about learning to recognize when something isn’t right. If we quietly generated my image without disclosure, the episode would have been undermining its own argument while broadcasting it.
The rule we landed on was short enough to fit on a napkin: audio is never generated. Video only if the alternative is deletion. Always labeled in-frame, not in the fine print.
That rule is human. The AI simply executes it.
But here’s the uncomfortable part: I said publicly that you would almost certainly not have noticed. The reconstruction was clean. The disclosure was voluntary. Which means hundreds of similar decisions are being made quietly, every day, by people who see no particular reason to label anything.
The AI didn’t choose silence. A human did.
The Accountability Gap — A Framework for Leaders
I’ve started calling this pattern the Accountability Gap: the space that opens between what an AI tool does and who is responsible for the outcome.
The gap is not malicious. It’s structural. AI tools are fast, smooth, and increasingly plausible. They make it genuinely easy to not notice that a decision was made — about what to include, what to omit, what to present as original, what to label.
For Mittelstand leaders, the Accountability Gap shows up in at least three places:
- Talent decisions: “We didn’t really hire juniors because of AI disruption.” — You stopped hiring. AI arrived later.
- Communications: Board decks, investor narratives, and performance reviews polished by an LLM — with no one clear on who owns the judgment layer.
- Process outputs: Automated analyses, AI-summarized due diligence, model-generated recommendations — with the model listed nowhere in the provenance chain.
In a transaction context — and I see this regularly — acquirers and investors are starting to ask: who actually made this decision, and can you show me the thinking? “The AI generated it” is not an acceptable answer. It is an Accountability Gap on a term sheet.
Aufgabe. Urteil. Wer macht’s?
This is the three-question framework I now run against any AI-assisted action before it leaves my desk.
1. Aufgabe (Task): What is the AI doing here, precisely? Generating? Summarizing? Reconstructing? Replacing a human step?
2. Urteil (Judgment): Who evaluated whether this was appropriate for this specific context — and are they named anywhere in the output?
3. Wer macht’s? (Accountability): If this output is wrong, misleading, or questioned six months from now — who is responsible, and do they know it today?
Three questions. Ninety seconds. The discipline is not about slowing down AI use. It’s about ensuring a human is standing behind the output with their name on it.
What I’m Taking Forward
The junior hiring crisis and the podcast disclosure feel like different problems. They are not. Both are about a temptation that AI makes structurally easier: the temptation to let a tool absorb the accountability for a choice that belongs to a person.
We are not going to un-deploy generative AI. Nor should we. But the leaders I respect most right now are the ones who use AI aggressively and own their decisions loudly — who say “we chose not to hire juniors because of X” instead of “AI is changing the workforce,” and who label what is generated instead of hoping no one notices.
That is not naivety about technology. That is integrity in the age of AI.
Five Things You Can Do This Week
- Pick one AI-assisted process in your business and answer the three questions: Aufgabe, Urteil, Wer macht’s? Write the answers down — not for compliance, for clarity.
- If your company stopped hiring junior talent in the last two years, write the actual reason on paper. Was it a deliberate cost decision or a strategic bet? Own it explicitly, or revisit it deliberately.
- Draft a three-sentence AI use rule for your team: what the tool does, who reviews it, who signs off on the output.
- In your next leadership meeting, replace one instance of “AI is changing X” with “We decided X” — and notice how differently the room responds.
- Ask one junior or mid-level person on your team which skills they have stopped being asked to develop. That answer will tell you more about your talent pipeline than any AI readiness index.
Which decision in your organization has quietly been reclassified as an AI disruption — and what would change if you put a name next to it?