The Art of the Prompt Is the Art of Leadership

About 12 to 18 months ago, I came across a post that caught my attention.

The idea was simple: a founder could increasingly operate with AI “agents” representing different members of a traditional executive team — Finance, Marketing, Legal, Risk, Chief of Staff, Strategy and other functions.

The concept stayed with me.

Not because I believed AI could suddenly replace an experienced executive team.

What interested me was something more practical.

As someone who had spent much of my career in large organizations and later served as a CEO, I immediately recognized the management model.

A CEO does not need to be the best accountant, lawyer, marketer, technologist or risk specialist in the company.

The CEO needs to understand the objective, provide context, bring together the right expertise, ask good questions, challenge recommendations, understand trade-offs and ultimately make a decision.

That was how I had worked with human teams.

Could I work with AI in a similar way?

That question changed the way I began using the technology.

From asking AI questions to managing intelligence

My early use of AI was probably similar to many people's.

Ask a question.

Get an answer.

Rewrite something.

Summarize a document.

Research a topic.

Useful, certainly.

But not transformational.

The real change came when I stopped treating AI as a search engine or writing assistant and began treating it more like a professional team.

That meant giving it context.

What business are we building?

Who is the client?

What are we trying to achieve?

What constraints matter?

What have we already decided?

What assumptions are we making?

What would Finance think?

What would Legal challenge?

What would Risk worry about?

What would Marketing say the customer actually wants?

And then, importantly:

What are we missing?

The output became dramatically better.

Not because the AI itself had suddenly changed.

The way I was working with it had changed.

The prompt is not the question

This has led me to think differently about the phrase "prompt engineering."

A prompt is often described as the instruction typed into the box.

I think that definition is too narrow.

The most powerful prompt is often the accumulated context behind the next question.

If I ask AI:

“Build me a website for a recruiting company.”

I may get a perfectly reasonable website.

But suppose we first spend time understanding:

• what the business actually does;
• who the customer is;
• which candidates we want;
• what makes the business different;
• how revenue is earned;
• what should be scalable;
• what should remain personal;
• what we want the customer to feel;
• and what success looks like.

Then we build the website.

The technology is now executing a strategy rather than inventing one.

I have experienced this repeatedly while building businesses.

We might begin by discussing the business model.

Then the operating model.

Then the target customer.

Then positioning.

Then governance.

Then marketing.

Eventually, when we get to the website or communications, the work becomes easier because the important thinking has already happened.

The AI understands the context because we built the context together.

That is very different from starting with:

“Write me some marketing copy.”

This feels surprisingly familiar

The more I work this way, the more familiar it feels.

It resembles how I worked as a CEO.

A capable executive team does not simply wait for the CEO to tell it what to do.

The CEO provides context and direction.

The functional experts investigate.

They return with recommendations.

The CEO asks questions.

Someone challenges an assumption.

New information emerges.

Finance identifies an economic issue.

Risk identifies an unintended consequence.

Legal identifies a constraint.

Sales explains what the customer is actually doing.

The recommendation changes.

Eventually, the leader decides.

Then the organization executes and learns.

Historically, that cycle might take days or weeks.

AI can compress parts of it into hours or minutes.

That is an extraordinary increase in management leverage.

But the leadership requirement remains surprisingly similar.

Better AI does not eliminate the need for better questions

This is where I think many discussions about AI miss something important.

People understandably focus on how intelligent the models are becoming.

But even a very capable system can produce an unhelpful answer if it is solving the wrong problem.

A leader still has to ask:

What are we actually trying to accomplish?

What assumptions are we making?

What must be true for this recommendation to work?

What evidence supports it?

What is the strongest argument against it?

What would another function say?

What happens if our first assumption is wrong?

What decision is actually required?

These are not AI questions.

They are leadership questions.

That is why I increasingly believe:

The art of the prompt is really the art of executive thinking.

The quality of the question reveals the quality of the thinking behind it.

Challenge matters more than agreement

There is another lesson from working with both executive teams and AI.

You do not want a team that simply tells you that you are right.

The best people I worked with were prepared to challenge.

They could understand the leader's intent without becoming intellectually subordinate to the leader's preferred answer.

The same principle applies to AI.

One of the most useful instructions I can give is:

Do not validate my current view. Tell me why it may be wrong.

Or:

Give me the strongest opposing argument.

Or:

What would cause this plan to fail?

That changes the relationship.

AI becomes less useful when it acts as a sophisticated form of confirmation bias.

It becomes more valuable when it helps expose assumptions.

The objective is not agreement.

The objective is a better decision.

The human still owns the decision

This is the part I would not outsource.

AI can research.

It can analyze.

It can compare alternatives.

It can identify risks.

It can draft.

It can challenge.

It can suggest a decision.

But somebody still needs to decide what matters.

Not every risk deserves the same weight.

Not every objective can be optimized simultaneously.

A financially attractive decision may conflict with reputation.

A faster decision may create unnecessary risk.

A technically elegant solution may be wrong for the customer.

Those are judgments.

And judgment carries accountability.

The leader must still be able to say:

This is the decision. These are the trade-offs I accepted. And I am responsible for the result.

AI can increase the leverage of judgment.

It should not eliminate responsibility for judgment.

Curiosity may become a competitive advantage

For much of my career, senior executives were valued partly for experience.

Experience still matters enormously.

It provides pattern recognition.

It helps identify when something feels wrong.

It provides context that cannot always be found in a database.

But AI makes another characteristic increasingly important:

curiosity.

The person who asks one good question gets an answer.

The person who asks the next five good questions may develop an entirely different understanding of the problem.

Why?

What are we assuming?

What else could be true?

Explain that differently.

What would the customer say?

What does the data contradict?

What am I not asking?

That curiosity creates iteration.

And iteration is where much of the value appears.

The strange next question

My experience with AI has been overwhelmingly practical.

It has helped me research, develop business models, challenge strategy, build websites, draft client materials and think through unfamiliar subjects.

But the better the collaboration became, the more another question began to bother me.

Today, the human largely determines the inquiry.

I provide the context.

I establish the objective.

I ask the next question.

I challenge the answer.

I decide when something doesn't make sense.

I decide when we have learned enough to act.

But what happens when increasingly capable AI becomes better not only at answering questions, but at determining which questions should be asked next?

What happens when it can independently:

• identify the weakness in its previous conclusion;
• create competing hypotheses;
• determine what evidence is missing;
• design the experiment;
• interpret the results;
• improve the method;
• and repeat the process?

At that point, AI is doing more than answering human prompts.

It is becoming increasingly capable of directing the inquiry itself.

That question eventually led me away from the immediate business applications of AI and toward a much larger question:

Where does this trajectory lead?

That is how I eventually arrived at what I now call The AI Conundrum.

For now, the opportunity is enormous

The longer-term questions are important.

But they should not obscure what is possible today.

We are currently in an unusual period.

A thoughtful individual with experience, curiosity and judgment can access extraordinary analytical capability at extremely low cost.

Small teams can perform work that once required much larger organizations.

Founders can explore disciplines outside their own expertise.

Executives can test ideas much more rapidly.

Professionals can prepare more deeply.

People can learn faster.

The differentiator, at least for now, may not be who has access to AI.

Increasingly, everyone will.

The differentiator may be who knows how to work with it well.

That means establishing intent.

Providing context.

Asking better questions.

Challenging the first answer.

Bringing multiple perspectives into the discussion.

Recognizing when something doesn't make sense.

Making the decision.

And remaining accountable for what happens next.

Those are not new leadership skills.

What is new is the amount of intelligence a capable leader can now bring into the room.

Perhaps the art of the prompt is not really a new discipline at all.

Perhaps it is simply the old discipline of leadership — expressed through a very different kind of team.