What We Can Learn About Leadership from Prompting

1. If you are vague, don't be surprised by the result

One of the first things we learn when using AI is that vague prompts tend to produce unpredictable results.

“Make a presentation about our strategy” can result in almost anything.

A better prompt explains the purpose of the presentation, the audience, the desired outcome, the context, the constraints and perhaps some indication of what good looks like.

Leadership is not so different.
Leaders sometimes complain that people have not delivered what they expected. But when we reconstruct the original conversation, it turns out that those expectations were never really made explicit.

• “I thought that was obvious.”
• “They should have known.”
• “That wasn't what I meant.”

These are familiar sentences.
The question for the leader should therefore not only be Why didn't they deliver? but also How well did I define what I expected?

2. Context matters

Good prompting is not simply about giving more detailed instructions. It is also about providing the right context. This gives the prompt more meaning.

AI produces better results when it understands why something matters, who the audience is, what has happened before, which constraints exist and what the desired outcome is.

People need context too.

In fact, providing context may be more important than providing detailed instructions.

If people understand the why, the desired impact and the boundaries within which they can operate, they can make decisions themselves. They do not need to return to the leader for every decision.

That creates an interesting paradox.

Better direction can create more autonomy.

Clarity does not necessarily mean micromanagement. The opposite may be true. When purpose, expectations and boundaries are clear, leaders can afford to give people considerably more freedom regarding how they achieve the result.

3. A prompt is not the end of the process

Another lesson from AI is that we rarely get the perfect result from the first prompt.

We look at the output. We evaluate it. We clarify something.
We add information. We challenge an assumption. We ask for another version.

And gradually the result improves.

Leadership works in much the same way.

Delegation should not mean: I told you what to do, now come back when it is finished. Effective leadership involves feedback loops.

We agree on an expected outcome, allow someone to work, look at progress, learn from what is happening and adjust where necessary.

This is not necessarily control. It is calibration.

The frequency of that calibration should depend on the situation. Someone who is experienced and working on a familiar task may need very little interaction. Someone who is learning, or dealing with an ambiguous and high-risk situation, may need considerably more.

The art of leadership is therefore not simply knowing when to intervene. It is knowing how much steering the situation requires. This is an example of versatile leadership.

4. You need competence to prompt well

This may be one of the more uncomfortable lessons. AI is often presented as something that reduces the need for expertise. To some extent, that is true. AI gives us access to capabilities that we did not previously have.

But there is a catch.

If I know absolutely nothing about a subject, I may not know what to ask. More importantly, I may not be able to judge whether the answer is any good.

The same applies to leadership.

A leader does not need to be the greatest expert in the room. In complex organizations that would be impossible. Leadership increasingly means working with people who know much more about their specific domain than the leader does.

But zero competence is problematic.

Leaders need enough understanding to ask meaningful questions, recognize important issues, challenge assumptions and evaluate whether something makes sense.

Perhaps we could call this minimum viable competence.

Not enough competence to do everyone's job, but enough competence to have an intelligent conversation about it. Without that competence, empowerment can easily become abdication or abandonment. 

5. Competence also builds trustworthiness

There is another reason why a certain level of competence matters.

People need to trust the person providing direction.

With AI, trustworthiness is something we need to evaluate externally. AI does not earn our trust in the human sense. We check its output because we know that it can be wrong.

Leadership involves a different kind of trust.

People ask themselves whether the person leading them understands the situation. Whether their judgment is sound. Whether they understand the consequences of what they are asking. Whether they will recognize good work. And whether they will take responsibility when things go wrong.

Competence therefore contributes to trustworthiness, next to loyalty or benevolence and integrity. 

A leader does not have to know everything. In fact, pretending to know everything is likely to undermine credibility. But leaders need enough competence to inspire confidence that their questions, choices and judgments are worth taking seriously.

6. Good prompting is about outcomes, not only instructions

There is an important difference between telling AI exactly what to do and telling it what you are trying to achieve. The same distinction matters enormously in leadership.

Weak delegation often consists of transferring activities:

• “Make this report.”
• “Organize that meeting.”
• “Contact these customers.”

Stronger delegation defines an outcome:

“We need to understand why customers are leaving and decide what we should change.”

The second formulation creates room for judgment.

This is where leadership differs fundamentally from command-and-control management. The leader defines the desired impact but does not necessarily prescribe every step towards it, unless they work with novice people.

A useful leadership question might therefore be:

Have I explained the task, or have I explained the desired impact?

The latter is usually more powerful.

7. Constraints can enable creativity

Good prompts often contain constraints.

• Keep it to two pages.
• Use these data.
• Don't change these elements.
• Focus on these three audiences.

At first sight, constraints seem to reduce freedom. In practice, they often make productive freedom possible. The same applies to organizations.

People need to know where they have freedom and where they do not.

A leader might say:

“This is the result we need. These are the principles we will not compromise. This is the budget. This is the deadline. Within those boundaries, decide how you want to get there.”

That is very different from either micromanagement or complete laissez-faire leadership.

8. Sometimes the problem is the prompt, not the person

This may be one of the most useful lessons.

When AI produces something disappointing, experienced users increasingly ask themselves:

Was my prompt good enough?

That reflex would be useful for leaders too.

When someone does not deliver what we expected, our first reaction is easily to evaluate the person.

• They misunderstood.
• They lack initiative.
• They didn't think it through.
• They aren't senior enough for the task at hand.

Sometimes that assessment is correct.

But before reaching that conclusion, there is another question worth asking:

What did I contribute to this outcome?

Was the objective clear? Did I provide sufficient context? Did we agree on expectations? Were the constraints understood? Did I check understanding? Did I provide feedback early enough?

Leadership means taking responsibility not only for the instructions or input we give, but also for the conditions under which others are expected to perform.

9. The response can improve the prompt

There is also a more subtle lesson.

When we work well with AI, prompting becomes a conversation. The response gives us information that changes our next question. Sometimes we discover that our original question was wrong. Sometimes an unexpected answer opens a possibility we had not considered. The same should happen in leadership.

Leadership should not be a one-directional process in which the leader formulates the perfect instruction and everyone else executes it. People know things the leader does not know. They see customers, technologies, processes and problems from different perspectives.

So good leaders do not only prompt. They allow themselves to be prompted back.

They give direction, listen to the response and are willing to adjust their own thinking.

That is particularly important in uncertain environments. When nobody knows exactly what the right answer is, leadership becomes less about transmitting answers and more about organizing an intelligent process of questioning, experimenting, learning and adapting.

10. You cannot outsource accountability

Finally, there is perhaps the most important lesson of all. When I ask AI to produce something for me, I remain responsible for what I do with the output.

“I didn't write it, AI did” is not a convincing excuse if I use incorrect information to make an important decision.

The same principle applies to leadership.

Delegation transfers responsibility for carrying out work, but it does not make leadership accountability disappear.

Leaders cannot say:

• “I delegated it.”
• “My team decided.”
• “I wasn't involved in the details.”

Leadership means creating the conditions for others to act while remaining accountable for the system in which they act. And that may be the deepest parallel between prompting and leadership.

11. Leadership is not about writing the perfect prompt

Of course, people are not machines. They have ambitions, emotions, experience, doubts, ideas, values and agency. They can disagree with the leader. They can see things the leader cannot see. And unlike AI, they need meaning, recognition, fairness, belonging and trust.

So leadership cannot be reduced to prompting.

But prompting can remind us of something we have perhaps always known.

Effective leadership requires us to be clear about the impact we want to create, provide enough context, define useful boundaries, give people room to act, maintain feedback loops, and have enough competence to ask good questions and evaluate what comes back.

And perhaps there is one final lesson.

The objective of a good prompt is not to demonstrate how clever the person writing the prompt is. It is to increase the probability of a useful outcome.

The same should be true of leadership.

Like I wrote 10 years ago in my book on sustainable leadership: leadership is not about the leader. It's about the progress a team achieves.