AI and Leadership: Why Efficiency Puts Your Own Opinion at Risk

AI and leadership – this is no longer an abstract debate about the future. Instead, it plays out every day, in boardrooms, in editorial offices and at desks where decisions are prepared. And it is precisely there that a loss reveals itself – one that does not feel like a loss, because it arrives so quietly.

Three Days Became Two Hours – and Then Something Was Missing

A speechwriter for a CEO recently put it in precise terms. Previously, he had needed several days to write a speech, whereas today, with AI support, it takes one to two hours. Up to that point, it sounds like a success story. Then, however, came the second part: the ripening time is gone. He no longer works through the topic himself, and he no longer has his own position on it. He writes, as he put it, in open space.

That sentence describes a problem that reaches far beyond the writing of speeches.

AI and Leadership: The Thinking Was Embedded in the Work

Those three days were never just writing time. During those three days, arguments were weighed against each other, formulations were discarded and objections were anticipated. At some point – usually late – a genuine stance on the topic emerged. The result was the speech. But the real product was the position.

That part is gone today. Not because anyone consciously abolished it – but because it was invisible. It appeared in no process, no job description, no invoice. It was a by-product of slowness. And slowness is something we are currently optimising away with great success.

Why This Is a Leadership Issue

What applies to a speechwriter applies to leadership on a far larger scale. Today, leaders receive decision templates that are produced in minutes: market analyses, scenarios, risk assessments, three options with a recommendation. Cleanly structured, plausibly reasoned, linguistically flawless.

Previously, a leader’s own opinion formed during the process of working through the topic – through questioning, through debate with the team, through the third attempt. Today, the finished result exists before the leader’s own position does. And that raises a question that is genuinely new: am I thinking this right now – or am I merely reading it?

The Dangerous Part: It Does Not Feel Wrong

If the result were poor, the problem would be easy to spot. But the result is good. Sometimes very good. It is structured, well-evidenced and sounds convincing – often better than one could have produced oneself in the same time.

That is precisely where the real risk lies. A finished, well-formulated position starts to feel like one’s own with surprising speed. It is not consciously adopted. One simply stops noticing that it has been adopted. AI does not implant opinions – but it delivers them so smoothly that the personal engagement with the material never happens.

The Test Comes Later

In practice, the problem rarely surfaces in the meeting where the decision is made. It surfaces afterwards:

• In the interview, when the second follow-up question arrives – not the first, which was prepared
• In the all-hands meeting, when someone from the third row asks: “And why exactly is this the right decision for us?”
• In the board meeting, when someone challenges an assumption that was never independently examined.

In those moments, no text helps. What helps is whether the topic has genuinely been worked through. People who have developed a position themselves can restate it in different words. People who have adopted a position repeat the same words.

AI and Leadership: Three Concrete Measures Against the Silent Loss

The answer is not to use less AI. That would be unrealistic – and the speechwriter does not want those three days back either. Instead, the goal is to deliberately rebuild the ripening time, because it no longer emerges on its own:

1. Your own thesis first, then the prompt. Two minutes, one sheet of paper: what do I think before I ask? Those two minutes determine whether one subsequently examines or merely adopts.
2. Use AI as an opponent, not as an author. “Formulate an argument for me” generates agreement. “Dismantle my position, find the weakness” generates thinking. The second prompt is more uncomfortable – and significantly more valuable.
3. The sentence test. Remove from every AI-assisted draft the sentences you would not defend in front of 800 employees. What remains is your actual position. Sometimes that is alarmingly little – and that is precisely the useful information.

Efficiency is not the opposite of quality. But it is sometimes the opposite of understanding. And leadership that does not understand cannot persuade. It can only present.

The more important question in the context of AI and leadership is therefore not which tasks AI takes over. It is: what thinking do we lose when we hand over the work in which that thinking used to happen?

When did you last notice that a position you were defending was not actually your own?

Yours, Nikolai A. Behr, CSP®