You Don’t Change the Mindset. You Change the Maze.

AI adoption is often treated as a people problem. Sometimes the real problem is the system people are being asked to work inside.

By DeQuan · Predicai · October 29, 2026

A friend sent me back to Who Moved My Cheese? recently.

If you’ve worked in sales long enough, you’ve probably lived some version of the book.

The station that used to be full starts running low. The old territory stops producing. The list that always worked gets quieter. The play you could run almost on muscle memory starts returning a little less every quarter.

And still, you go back.

Not because you’re lazy.

Because you know the route.

You know where to look. You know how long it takes. You know what success used to look like there.

The book is good at diagnosing what happens next: eventually, someone has to leave the old station and go back into the maze.

But a friend asked me a question that I don’t think the book fully answers:

How do you make someone actually want to go hunting somewhere else?

I’ve been thinking about that question for a while, because I’ve had to do it myself.

I didn’t need to relearn how to hunt.

I needed to learn a new terrain.

Moving into a different industry taught me something I didn’t fully appreciate before.

A lot of the instincts travel.

How to ask questions.

How to read a room.

How to find the real decision-maker.

How to recognize urgency.

How to follow a signal.

How to build trust.

Those things don’t disappear because the market changes.

But the terrain does.

The language changes.

The buyers change.

The signals change.

The rules change.

The things that look like opportunity in one market can mean almost nothing in another.

So you’re not starting over.

But you also can’t walk into a new environment and assume the old map still works.

I had to keep the parts of the hunt that were useful and build a new way of navigating around them.

And increasingly, AI became part of that.

Not because AI knew the terrain for me.

Because it changed the cost of learning it.

Research that once took hours could happen much faster.

I could map organizations, study new terminology, compare sources, prepare questions, test messaging, organize what I learned, challenge assumptions, and preserve useful patterns for the next time.

But that only became valuable when I stopped treating AI as something I occasionally asked for help and started building a system around how I used it.

The model is not the system.

And in this story, the AI isn’t the cheese.

It changes the maze.

We keep treating adoption like an attitude problem.

This is where I think the cheese metaphor becomes useful for AI.

A lot of organizations introduce a new capability and immediately focus on the person.

Use the AI.

Experiment more.

Try the new tool.

Be curious.

Change your mindset.

Then we leave almost everything surrounding the work untouched.

The information is still scattered.

The approved use cases are unclear.

The old workflow still has ten steps.

The performance measures still reward the old behavior.

Nobody has documented what works.

Employees don’t know what the AI is allowed to do.

They don’t know which outputs they can trust.

Every useful result creates another round of manual checking.

And then we wonder why people keep walking back to the old station.

We moved the cheese. We didn’t redesign the maze.

That is not always resistance to change.

Sometimes the economics of the system still favor the old behavior.

If the familiar way is easier, safer, better understood, and still rewarded, then returning to it is a perfectly rational decision.

Which means the leadership problem may not be:

How do I change this person’s mentality?

It may be:

How do I change the conditions around the work?

1. Map the new territory before asking people to explore it.

The most expensive part of a new hunt is often the beginning.

Where do I start?

What information matters?

Which sources can I trust?

What does good look like?

What can go wrong?

Who has already figured some of this out?

AI can dramatically reduce that first-mile cost.

But handing someone a chatbot is not the same as handing them a map.

A useful system brings together context, trusted sources, instructions, examples, constraints, and a clear objective.

Instead of saying:

“Go use AI.”

You can say:

“Here is the workflow. Here are the sources. Here is what the system can help with. Here is what still requires your judgment. Here is what good looks like.”

The maze immediately becomes cheaper to enter.

2. Make the first piece of new cheese small and fast.

People repeat what rewards them.

That matters in AI adoption because the first experience is disproportionately important.

If someone’s introduction to AI is an enormous transformation project with unclear benefits, they may spend months experiencing nothing but friction.

If the first use case removes an hour of repetitive work next Tuesday, the lesson is different.

Start with something bounded.

One workflow.

One useful problem.

One measurable outcome.

Let someone experience the benefit before asking them to believe in the transformation.

This is one reason I keep coming back to repeatable skills and procedures rather than endless prompt libraries.

A successful experience should be reusable.

The first win should make the second win cheaper.

3. Connect the new behavior to something the person already wants.

Nobody wakes up wanting to improve the company’s “AI adoption rate.”

People want something else.

• More time.
• Better work.
• A larger opportunity.
• Less administrative burden.
• More confidence going into a meeting.
• A new skill.
• A promotion.
• A problem they have been tired of solving manually.

The question shouldn’t only be:

Where can AI be deployed?

It should also be:

What is this person already trying to accomplish?

Then build from there.

People are much more willing to enter a new part of the maze when something they genuinely value is on the other side.

4. Change the scoreboard.

You cannot tell people to work differently while measuring them exactly the same way.

If a system rewards familiar motion, people will produce familiar motion.

If you say experimentation matters but only count the output of the old process, experimentation will remain extracurricular.

If someone builds a workflow that eliminates ten repetitive steps but the organization only measures how quickly they perform those ten steps, the system is fighting its own improvement.

The scorecard tells people where the organization actually wants them to go.

AI changes the unit of work.

Eventually, measurement has to follow.

5. Write it on the wall.

One of the ideas I’ve always liked from Who Moved My Cheese? is Haw leaving messages behind as he moves through the maze.

For teams, that is more important now than ever.

When someone discovers a better way to research an account, prepare a meeting, verify information, structure a handoff, review an output, or use an AI tool safely, that learning should not disappear into their personal workflow.

Write it down.

Turn it into a skill.

Turn it into a playbook.

Build the source rules into it.

Add the constraints.

Define the handoff.

Capture what failed.

Make the next person’s first hour cheaper than yours was.

One person explores. The system remembers.

That is how individual experimentation becomes organizational capability.

The new maze needs guardrails too.

There is an important difference between making exploration easier and simply making AI do more.

A system also has to establish boundaries.

• What may the AI know?
• What may it recommend?
• What may it change?
• What may it release?
• Where does the evidence come from?
• What happens when sources disagree?
• Which checks can be deterministic?
• What requires judgment?
• When should an exception move to a person?
• What can be rolled back?

These questions are not barriers to adoption.

They are part of what makes adoption possible.

People trust a new path faster when they understand where the edges are.

This is why I think so much of the conversation about AI adoption starts in the wrong place.

The goal is not maximum AI use.

The goal is dependable work.

Sometimes AI should explore.

Sometimes it should recommend.

Sometimes it should execute.

Sometimes the right system should stop and ask.

Operating AI means knowing the difference.

Go into the maze first.

There is one other leadership lesson here that matters to me.

Haw’s eventual move through the maze works because he learns that the uncertainty he imagined was worse than much of what he actually encountered.

Leaders can shorten that learning curve for everyone else.

Go first.

Try the workflow.

Find the dead ends.

Learn where the information is weak.

Figure out what requires more review than expected.

Build the first version of the map.

Come back with evidence instead of enthusiasm.

That matters with AI because people can tell the difference between a leader encouraging experimentation and a leader who has actually done the work.

The second person does not need your motivation speech.

They need the map you brought back.

Mindset follows experience.

I started with a question about mentality.

How do you get someone to become eager to hunt somewhere new?

I’m less convinced now that mentality is the right place to start.

Because experience changes mentality.

Make the first move cheaper.

Make the territory clearer.

Connect it to something worth pursuing.

Reward the new behavior.

Preserve what people learn.

Put boundaries around the places where mistakes matter.

And let the system make the better behavior easier to repeat.

That is the deeper lesson I take from the maze now.

You can tell people to embrace change.
Or you can change the conditions that make the new behavior worth choosing.

AI gives us extraordinary new capabilities.

But capability alone does not change how work gets done.

The tool is never the whole story.

The maze around it is the system.

And if you want people to move, sometimes the most important thing you can do is redesign the maze.