Nothing will humble you faster than trying to teach something you've become exceptionally good at. 

What feels obvious to you feels overwhelming to someone else.

You think you're explaining the process.

They think you’re speaking a different dialect. 

You walk away wondering, How is this so hard to understand?

Then they walk away wondering, Why is this so hard to learn?

Neither of you is wrong. You're just working from completely different reference points. 

The better you get at something, the harder it becomes to remember what it felt like to be a beginner.

The decisions that once required conscious thought become automatic.

The questions you used to ask stop occurring to you.

The tiny details that create exceptional outcomes no longer feel like individual steps, they just feel obvious.

Psychologists call this unconscious competence.

It's the final stage of learning. The point where you've repeated something so many times that you no longer have to think about how you do it.

Which is incredible for you... and insufferable for the people you're trying to train.

You assume people need more effort, more accountability, and repetition. 

Most of the time, they need more context.

You're asking someone to execute a process they can't see because half of it never leaves your head.

Every shortcut, judgment call, and tiny decision that separates a mediocre outcome from a great one has become so automatic, you don't even realize you're making it. 

That's why documenting your work feels so frustrating. You're trying to explain decisions you stopped consciously making years ago.

Ask me how I know…

Some of the longest meetings in our company revolve around one question: "Nat, explain how you decided that."

They already know what I did.

They're trying to understand why I did it.

And after spending hundreds of hours trying to pull judgment, taste, and pattern recognition out of my own head, one thing became painfully clear:

If I couldn't consistently extract my own thinking, I could never expect someone else to learn it.

So we built a process called The Extraction Loop.

It's a simple four-step process that turns instinct into something another person can understand, repeat, and improve. 

The solution isn't to become a better teacher. It's to become better at extracting your own thinking. 

Open ChatGPT or Claude and turn on voice mode.

Pick something you've done often enough that you don't really have to think about it anymore. It could be a conversation, a decision, a workflow, a meeting, a creative process, or the way you solve a recurring problem.

Now explain it out loud as if you're training someone to do it exactly the way you would.

Explain what you did, why you did it, what you noticed, the shortcuts you've learned, the mistakes people usually make, and the judgment calls you made along the way.

Don't worry about making it organized. Just talk. 

Your goal isn't to build documentation yet. It’s to capture everything your brain has stopped noticing before you try to turn it into a process.

Once you've emptied your head, let AI organize it.

Ask ChatGPT to turn your brain dump into a step-by-step workflow in the exact order each step happens. Your goal isn't to accept the first draft. It's to give yourself something concrete to react to.

As you read through it, look for what's missing.

Keep refining it until the workflow matches what actually happens in your head.

AI is great at organizing information but your job is to make sure it's organizing the right information.

Before you worry about perfecting the workflow, make sure everyone is speaking the same language.

Go back through the process and identify every word, concept, or standard you're assuming people already understand. Then ask AI to pull them out and define each one as clearly and specifically as possible.

Don't stop at what something is. Define how you recognize it, how you measure it, and why it matters.

If a word can be interpreted five different ways, it will be.

Without shared definitions, you won't get consistent execution. People can't follow the same process if they're working from different assumptions.

A simple gut check: if you handed this process to five different people, which words or phrases would each of them interpret differently?

Once you've documented the workflow and defined your terms, try to break it.

Ask AI to roleplay someone who's never done this before.

Have it look for gaps, assumptions, and places where it would get stuck. Encourage it to challenge your thinking and ask every question a new team member would ask before they could successfully complete the task.

Then keep refining the process until there are no more unanswered questions.

The goal isn't to write the perfect SOP, it’s to remove as much guessing as possible before another human ever has to guess.

Every time you do this, you're taking judgment that only exists in your head and turning it into something another person can understand, repeat, and eventually improve.

By the end, you have more than a set of instructions. You have a clearer record of how you make decisions, what you notice, and what separates an acceptable outcome from a great one.

The first couple times you do this, it'll feel painfully slow. It should.

You're forcing yourself to notice work you've spent years teaching your brain to automate.

But eventually something changes.

You stop waiting until someone asks, "How did you decide that?" and you start capturing your judgment while you're making it.

And documentation stops feeling like a chore because you're no longer trying to reconstruct your thinking after the fact. You're capturing it while it's happening.

So remember:

The purpose isn’t to just document the work. 

It’s to make your expertise transferable.