How far could human thinking go because AI exists?

I keep coming back to that question.

Not because I'm worried about whether humans can keep up with AI.

I'm interested in something much bigger.

How far could human thinking go because AI exists?

That's really the territory I've been exploring through Human Frontier Thinking™ — the discipline of continually expanding how humans think, learn and act at the edge of change, discovering capabilities and possibilities we did not previously know existed.

And right now, AI gives us an extraordinary place to practise it.

For the last few years we've concentrated, understandably, on AI skills.

Can people use the tools? Can they prompt? Can they automate? Can they work with agents?

We need all of that.

But I think the question is moving.

Organisations don't simply need people who can use AI.

They increasingly need people capable of thinking, judging and exercising agency while working with it.

And there is growing evidence that we need to pay attention to how those human capabilities are developing.

We're going to need people who can think well at precisely the moment it becomes easier not to

IBM's new Designing the Thinking Organization study, conducted with Oxford Economics, surveyed 1,500 Chief HR Officers (CHROs) and senior workforce executives and 8,800 employees.

It found that 60% of employees worry AI is eroding their skills. Among those concerned, three quarters believe the erosion has already begun.

Critical thinking was the capability most frequently identified as deteriorating.

Now look at what organisations say they need.

57% of CHROs identify critical thinking and problem framing as among the most important AI-era skills.

48% identify human judgement.

And 71% say the ability to supervise, validate and override AI is an essential workforce capability.

There's the tension.

We're going to need people who can think well at precisely the moment it becomes easier than ever not to.

I don't think AI inevitably makes us worse thinkers.

My own experience suggests almost the opposite.

But it depends enormously on how we work with it.

Using AI and thinking with AI are not the same thing

I've been experimenting pretty intensively with AI for some time now.

Some of my best experiences haven't been when AI has given me a better answer.

They've been when something in the interaction has caused my own thinking to move.

A challenge I hadn't considered.

A connection between two ideas.

An assumption exposed.

Something I've said comes back slightly differently and I think, hang on…

And off I go again.

That's the bit that fascinates me about HI:AI®.

I'm not interested in discovering how much of my thinking I can hand over.

I'm interested in:

How much further can I think because another intelligence is available to think with?

And that brings me straight back to Human Frontier Thinking™.

Because if we want to become more capable alongside AI rather than simply more efficient because of it, I think there are human capacities we need to develop deliberately.

Edge Sensing: notice what is changing

AI itself is changing so quickly that knowing how to use today's tools won't be enough.

We need to notice what is changing around them.

What's happening to our work?

Our profession?

Our customers?

Our assumptions?

What has suddenly become abundant that used to be scarce?

What is becoming more valuable because AI exists?

And what are we still doing because we've always done it?

That's Edge Sensing.

You can't adapt to something you haven't noticed.

And you certainly can't shape what comes next if you're still looking backwards.

Courageous Curiosity®: question what you know

AI gives us access to an extraordinary abundance of answers.

That makes the quality of our questions even more important.

What has changed?

What are we assuming?

What doesn't quite fit?

Whose perspective is missing?

What if the opposite were true?

What else could be possible?

That's more than learning how to write a clever prompt.

It's remaining curious when experience is telling you that you already know.

And that matters particularly for experts.

Expertise gives us enormous value. It gives us knowledge, patterns and experience.

But it can also make “I know this” remarkably comfortable.

Courageous Curiosity® interrupts that with:

What if something has changed?

Independent Thought: don't just ask AI first

This may be one of the simplest things we can all practise.

Sometimes, think first.

Before you open the AI, ask yourself:

What do I think?

What have I noticed?

What does my experience tell me?

What doesn't make sense?

What might I be missing?

Get your own thinking onto the table.

Then bring AI in.

Challenge me.

Find the holes.

What's the strongest argument against my position?

What evidence would change my mind?

What haven't I considered?

Now you're doing something quite different.

AI isn't establishing the intellectual starting point.

You are.

And the machine gives your thinking something to push against.

Because acquiring an answer isn't necessarily the same as developing your thinking.

Dynamic Deliberation®: judge what matters

AI can give us more information, analysis, scenarios and options than we've ever had before.

That doesn't necessarily make decisions easier.

Someone still has to decide what matters.

What evidence carries most weight?

What's noise?

What's the trade-off?

What are the consequences?

What would change my mind?

And eventually:

What decision am I prepared to own?

That's judgement.

Research from the Alan Turing Institute's Centre for Emerging Technology and Security into AI and professional skills in law and healthcare makes this point particularly well.

The effects of AI on professional capability depend on what we delegate, whether people remain cognitively engaged, how they are trained and supervised, and whether they retain the capability and authority to challenge AI outputs.

That's an important combination.

We can put a human at the end of every AI process we design.

But if that person no longer understands enough to challenge what they're seeing, what exactly have we protected?

I've been using this distinction for some time:

Human-in-the-loop is not the same as human-judgement-in-the-loop.

The Human Override® isn't the ceremonial presence of a person.

It requires an active human capable of saying:

No. I see this differently. And here's why.

Co-Creation: don't just delegate, think together

This is probably the capacity I've experienced most directly through my own work with AI.

I don't want AI simply to do things for me.

I want to jam with it.

I bring an idea.

It challenges it.

Something comes back.

I disagree.

We explore another direction.

A connection appears that neither starting point contained.

I test it somewhere else.

It changes again.

That's co-creation.

And I think it's fundamentally different from delegation.

The test I increasingly use is:

Did AI merely give me something or did the interaction enable me to think somewhere I wouldn't have reached alone?

That, for me, is HI:AI® at its best.

Not human intelligence competing with artificial intelligence.

Human and artificial intelligence interacting in a way that allows the human to develop too.

Adaptive Action: do something with what you've discovered

Human Frontier Thinking™ isn't an invitation to sit around having fascinating conversations about the future.

Try something.

Experiment.

Act.

Notice what happens.

Learn.

Change course.

Try again.

Nobody has the definitive handbook for working with AI because the technology will have moved again before the ink dries.

We develop capability by interacting with change.

That's agency.

Not simply responding to the future as it arrives, but believing we can participate in shaping what happens next.

The best AI users may already be showing us something

Microsoft's 2026 Work Trend Index found something I especially like.

Its more advanced AI users aren't simply the people using AI for more things.

They're more deliberate about the relationship.

Among Microsoft's “Frontier Professionals”, 53% deliberately pause before starting work to decide what should be done by AI and what should be done by a human, compared with 33% of other users.

And 43% deliberately do some work without AI to keep their own skills sharp, compared with 30% of other users.

I think there's something in that.

The future-ready human may not be the person who uses AI for everything.

It may be the person who knows when to use it, when to challenge it and when to put it down and think for themselves.

Artificial intelligence may become abundant. Human attention isn't.

There's another part of this I'm increasingly interested in, because of what I see in my work with senior leaders.

Most aren't short of information.

They're not short of things to do.

Quite the opposite.

And now AI can produce even more.

Twenty ideas.

Five strategies.

Another analysis.

More information.

More possibilities.

More output.

All before you've finished your coffee.

But AI hasn't given us another brain with which to consider it all.

So I've become increasingly interested in Attention Allocation.

What deserves my attention?

Where does deep human thought add value?

What can AI handle?

What can I ignore?

Where does somebody need me, rather than another piece of output from me?

Perhaps becoming more capable in the AI era isn't about learning to process ever more.

Perhaps part of it is becoming much better at deciding what deserves our attention in the first place.

Which takes us to Intelligence Allocation

We've always allocated resources.

People. Money. Time. Authority.

Now we have another form of intelligence available to us.

So how do we allocate the work and the authority between human and artificial intelligence?

What should I do?

What should AI do?

Where should we think together?

Where does human judgement matter?

Who gets to decide?

That's what I've been calling Intelligence Allocation.

And there is a deceptively simple Human Frontier Thinking™ question underneath it:

What is the best use of the human here?

Importantly, the answer isn't always human.

If AI can remove pointless work, let it.

If it can analyse something far better than I can, use it.

If it can find patterns I can't see, brilliant.

But make that allocation deliberately.

Because some work gives us more than its output.

It develops us.

Don't protect the task. Protect the capability the task was developing.

Think about how expertise develops.

You try.

You get things wrong.

You see another example.

Then another.

You begin noticing patterns.

You develop judgement.

Eventually, something that once required enormous conscious effort becomes almost intuitive.

Some of the routine work AI is beginning to absorb has historically been part of that developmental journey.

The Turing research raises exactly this concern in professional settings: if AI takes over formative work, we also need to think about how future professionals acquire the experience and judgement that doing that work once developed.

That doesn't mean keeping pointless work because it's character-building.

Please, let's get rid of plenty of that.

It means:

Don't protect the task. Protect the capability the task was developing.

If AI takes the task and the capability still matters, we need another way for humans to get the reps.

Which means sometimes we should deliberately do the hard thing.

Think it through ourselves.

Sit with uncertainty.

Create something before asking AI what it would create.

Make the judgement.

Have the conversation.

Get something wrong and learn from it.

Not all friction is valuable.

But not all friction is waste either.

This can't all be left to the individual

We can't tell people to become better thinkers while simultaneously redesigning work so they have fewer opportunities to think.

Organisations have choices to make too.

IBM found that only 26% of organisations clearly define work as human-led, AI-assisted or AI-executed.

Those that do report better quality and lower risk, although those findings are associations rather than evidence that the work allocation itself caused the improvement.

So the leadership conversation needs to move beyond:

How many people are using AI?

Towards:

Where could AI create capacity?

What should humans continue practising?

Where does human judgement add value?

What happens to the capacity AI releases?

Where does authority sit?

What are our people becoming better at?

And:

Are our people becoming more capable as our organisation becomes more AI-enabled?

That's a much bigger ambition than AI adoption.

So perhaps we've been asking the wrong question

We've spent a lot of time asking what humans need to learn about AI.

I'm increasingly interested in what humans might learn about themselves because AI exists.

How much more curious could we become?

How much better could we become at noticing change?

At thinking independently?

At judging?

At creating?

At working with another intelligence without surrendering our own?

At turning possibility into action?

These aren't human capabilities we need to preserve in aspic because machines are coming for them.

They're capacities we can deliberately develop.

That's at the heart of Human Frontier Thinking™.

AI gives millions of us access to another form of intelligence with which we can question, explore, challenge, create and think.

We could use it simply to make what we already do faster.

And sometimes that's exactly what we should do.

But there's a much bigger possibility here.

What could humans become if we deliberately developed ourselves as fast as we are developing our machines?

There is more inside us than we know.

Let's find out how much more.

For the past two years, Carol has been testing a question: can working with AI make her a better thinker? She co-creates with Chad as a creative thinking and ideation partner, and Claudio as an objective sparring partner - challenging assumptions and testing consistency and clarity. It is a live, ongoing experiment in HI:AI® and Human Frontier Thinking™.