AI as a Mirror: How Leaders Can Use AI to Challenge Their Thinking

After more than 2,000 hours working intensively with generative AI, one idea has stayed with me:

AI becomes much more interesting when we stop asking it only to produce — and start using it to reflect.

Most conversations about AI still begin with productivity.

How can I write faster? Summarise faster? Research faster? Produce more?

Those are useful applications. But for leaders, I believe there is a more consequential opportunity.

AI can help us examine how we think.

Not because AI knows us better than we know ourselves. It doesn't.

Not because it has access to some hidden truth about us. It doesn't.

But because our prompts, questions, language, assumptions and reactions contain traces of the way we see a situation.

AI gives us a new surface against which to examine them.

Used well, it becomes a mirror.

And a mirror can reveal things we were too close to see.

1. The invisible architecture behind our decisions

Every important decision rests on assumptions.

Some are explicit:

This market will continue to grow.

This person is ready for the role.

This transformation needs to happen now.

Others are less visible:

This is how our industry works.

This is what my organisation expects from me.

This is the kind of leader I am.

This option is too risky.

Over time, assumptions can become beliefs. Beliefs become habits of thought. And habits of thought influence what we notice, what we dismiss and which possibilities we consider realistic.

That creates an invisible architecture around decision-making.

The problem is not that we have beliefs and assumptions. We need them to navigate complexity.

The problem begins when we stop seeing them as assumptions and start treating them as facts.

That is where AI can become useful.

2. Making the invisible visible

In 2025, I described AI as a possible “mirror of the subconscious.”

I would formulate it differently today.

AI does not reveal our subconscious.

It can, however, reflect the language, assumptions and patterns we give it.

That distinction matters.

Imagine that I am considering a major professional decision. Instead of asking:

“What should I do?”

I can give AI the situation, my reasoning, the options I see, what attracts me, what worries me and what I currently believe to be true.

Then I can ask:

Which assumptions am I treating as facts?

Where does my reasoning contradict itself?

What am I potentially underestimating?

What would someone who strongly disagrees with me say?

What information would change this decision?

What question am I not asking?

Suddenly, AI is doing something very different from generating an answer.

It is helping me examine the architecture of my own thinking.

3. The quality of the mirror depends on what we put in front of it

There is an important limitation.

AI is not automatically challenging.

It can also be extremely agreeable.

If I present a poorly framed idea and implicitly ask AI to validate it, I may receive an articulate explanation of why I am right.

That feels good.

It is also dangerous.

A mirror can become an echo chamber.

This is why effective use of AI at executive level requires more than good prompts. It requires the willingness to expose our reasoning to contradiction. It is also the practical discipline behind Lead with AI.

Instead of:

“Help me demonstrate why option A is the right choice.”

Try:

“Build the strongest possible case that option A is the wrong choice.”

Instead of:

“Improve my strategy.”

Try: “Assume this strategy fails in 18 months. What were the three assumptions we should have challenged today?”

Instead of:

“What do you think of my reasoning?”

Try: “Separate facts, interpretations and assumptions in my reasoning. Then identify where the evidence is weakest.”

The difference is fundamental.

In the first case, AI assists.

In the second, AI challenges.

For a leader, that can be considerably more valuable.

4. From noise to structure

Leadership decisions rarely suffer from a lack of information.

More often, the difficulty comes from too much of it.

Data. Opinions. Previous decisions. Organisational politics. Experience. Expectations. Fear. Opportunity. Time pressure.

Everything arrives at once.

One of the most useful roles I have found for AI is helping separate that complexity into layers.

A simple sequence can be powerful:

Deconstruct. Put the situation on the table: facts, doubts, contradictions, options, constraints.

Reveal. Ask AI to identify patterns, assumptions, gaps and tensions.

Challenge. Pressure-test the reasoning. Ask for counterarguments. Change perspectives. Look for disconfirming evidence.

Reconstruct. Rebuild the situation around what survives the challenge.

Decide. Return the judgement to the human.

That final step matters.

The purpose is not to outsource the decision.

It is to improve the conditions under which the decision is made.

5. AI can expand perspective. It cannot provide judgement.

This is where I draw an important line between human and artificial intelligence.

AI can process extraordinary amounts of information.

It can generate alternatives.

It can spot inconsistencies.

It can simulate perspectives.

It can help us articulate something we have not yet managed to express.

But it does not carry the consequences of the decision.

The leader does.

Judgement includes things that cannot simply be delegated to a model: context, experience, responsibility, values, timing, relationships and an understanding of consequences that may never appear in the prompt. It is the human side of The Leader’s Operating System.

This is why I don't see the future of leadership as human versus AI.

I see it as a question of allocation.

What should the machine help us see?

What must the human ultimately judge?

The more capable AI becomes, the more important that distinction becomes.

6. From reflection to better decisions

There is a temptation to make conversations with AI intellectually fascinating.

You can explore an issue for hours.

Generate ten perspectives.

Create sophisticated frameworks.

Challenge every assumption.

And still do nothing.

Reflection only creates value when it eventually changes something.

A decision. A conversation. An experiment. A behaviour. A direction.

This is why my own use of AI has progressively moved beyond prompting and productivity.

I use it as part of a larger loop:

Learn → Think → Decide → Act → Observe → Adjust.

AI can accelerate several parts of that loop.

But movement is what closes it.

The quality of an AI conversation should therefore not be measured by how impressive the answer sounds.

A better question is:

What can I now see, decide or do that I couldn't before?

7. The leader still leads

The more I work with AI, the less interested I become in the idea of AI replacing human thinking.

The interesting question is whether it can raise the quality of human thinking.

Can it expose an assumption before it becomes an expensive mistake?

Can it introduce a perspective that was missing from the room?

Can it help a leader distinguish signal from noise?

Can it challenge a strategy before reality does?

Can it turn an intuition into something explicit enough to examine?

Sometimes, yes.

But only if we use it that way.

AI doesn't automatically make us wiser.

It can accelerate shallow thinking just as easily as deep thinking.

The leverage comes from the quality of the dialogue — and from the quality of judgement on the human side of it.

That is why, for me, AI is becoming less interesting as a tool for producing answers.

And much more interesting as a way of producing better questions.

An idea developed in motion

This article builds on a three-part series I first published in 2025: Changing Belief Systems, Human + AI — The Subconscious Dialogue, and The Trajectory of a Leader.

The central intuition was already there: AI could become more than a production tool. It could become a mirror for thinking.

Since then, after continued experimentation with generative AI and its use in real leadership situations, some of those ideas have become stronger. Others have changed.

I no longer describe AI as a mirror of the subconscious.

I see something more precise — and, for leaders, more useful:

AI can help make our thinking visible enough to examine, challenge and improve.

The machine can expand the field of view.

The judgement remains ours.

Alignment emerges through movement.