​

There is a researcher named Simone Stolzoff who studies the relationship between knowledge and bias.

His observation is straightforward: the more expert you become, the more susceptible you are to filtering new information through what you already believe.

Expertise does not make you more objective. It makes your assumptions harder to see.

Which creates a real problem for mid-career professionals trying to get better with AI.

Think about it, you aren't necessarily a beginner...

You have context.
You have pattern recognition.
You have strong opinions about how things should work.

And all of that is genuinely useful.

But when you sit down with an AI tool, those same qualities can push you toward a specific behavior: using AI to produce what you already planned to produce, faster.

That's not really an AI gain, but rather autocomplete with a few extra steps thrown in.

The "Expert" Trap

The Expert Trap is what happens when you bring confident assumptions into an AI workflow and never stop to surface them.

It looks like this:

You need to draft a technical recommendation.

But, you already have your own view.

So you open a tool, describe the situation, and ask for help structuring your argument.

The output validates your direction.
You move forward, and the whole thing takes 10-minutes.

And you never once asked: Am I starting from the right assumption?

The Expert Trap isn't about being wrong, but rather about never finding out what answers you aren't getting because you're blind to them.

What actually works

The professionals who get the most from AI are not the ones who use it to produce faster. They are the ones who use it to think earlier.

Here is how that works in practice:

Before I enter any technical conversation, whether it's a

client discovery call,
a conversation with an engineering team,
or a decision that requires cross-functional alignment,

I do two things first:

First: I use AI to rebuild the context around the problem before I finalize my position.

I describe what I think I know, what I think the goal is, and what I think is driving the decision.

Then I ask the model to tell me what I might be missing, where my framing might be incomplete, and what questions I have not asked yet.

Second: Then I run my thinking through the 5 Why's. Not to find the answer, but to find the assumption I skipped.

(What's the "5-Why's" and how can I implement it?​
Good question, my dear watson...

The "5 Why's is a practice of asking the question "Why?" 5 times before settling on an answer.

If you're a parent (or have every worked with kids) you know exactly why this works, because you can be 100% confident until you get asked why a million times)

This isn't a research exercise (it might take 10-minutes, max).

But it changes the quality of every conversation that follows, because you walk in with calibrated confidence instead of unchecked confidence.

The difference isn't speed. It's when you initially use AI.

Most professionals use AI at the blank page stage, when they are ready to produce.

The better move is to use it at the fog stage, before you think you know what to do.

That's when it can actually change your thinking, not just execute it.

Your job this week:

Before your next:
​
decision
meeting
deliverable

spend 10-minutes using AI to challenge your own starting assumptions.

Write down what you think you know, what you are not sure about, and what you have not asked yet.

See what comes back.
Then go into the room.

​