The Great Flattening: what AI does to middle management

The starting point is a worn-out sentence: “AI is not taking your job away, the person who uses AI is.” The episode takes it seriously enough to test it against figures.

What the figures yield

The IFO figures are a good example of why studies on this topic rarely create clarity. 20 to 30 per cent of companies expect job cuts. Others reckon with more jobs.

Both stand side by side, and both can apply, because companies grow differently. A business whose market is growing converts a productivity gain into more output. One in a saturated market converts it into fewer staff.

The headlines about redundancies at large technology companies rarely tell the whole story, because several developments coincide there: a correction of over-hiring during the pandemic, redeployment into other areas and actual automation.

What can be said robustly: a productivity gain changes who carries which responsibility, regardless of whether the overall number rises or falls.

Why middle management

The episode's actual finding concerns one level, and the reasoning is structural.

As examples the episode names Amazon under Andy Jassy and the merging of IT and human resources at Moderna. The second case is the more interesting one, because it shows that the shift affects not only levels but also boundaries between areas.

What that means for skills

Vibe coding appears in this episode as evidence of the same movement: tools that deliver quality assurance alongside the code. What used to be a division of labour between roles becomes one process.

The skills in demand shift accordingly: away from pure specialist knowledge, towards soft skills and a feel for how to negotiate with machines.

The second part sounds like a stock phrase and is concrete. Working with a model means formulating an intent so that it is understood without follow-up questions, assessing the result and correcting course. Those are the same skills you need in order to hand a task to a person.

That is exactly why it is no surprise that the people who cope best are the ones who were already good at delegating.

The example from practice

As its own experiment the episode brings in a self-built AI advisory board: persona prompts that debate like Steve Jobs, Tim Cook or Angela Merkel and deliver a recommendation for action at the end, including voice cloning so that it sounds like a conversation.

On the everyday side stands the resolution of the opening: how Manus organised a complete musical weekend including dog-friendly accommodation and a charging point for the electric car.

Both examples describe the same thing from two directions. What used to be research and coordination has become an assignment. Whoever did that research before was useful for it. What counts now is the ability to pose the assignment correctly and to check the result.

Conclusion

The episode's title is pointed and the movement behind it is real: hierarchies are becoming flatter because the share of preparation is falling and with it the span of control is rising.

For your own role an honest split is worthwhile. How much of your working time goes into preparation, meaning reports, consolidation, presentations? This part changes first.

And how much goes into decisions, conflicts and coordination against resistance? This part remains and gets denser.

The episode's conclusion puts it in a nutshell: anyone who does not engage with AI will not be replaced by AI, but by the colleagues who do.