Forty minutes. That's how long she spent reading a document a machine wrote in nine seconds.

It was fine at minute three.

She kept going because if it's wrong, the machine doesn't take the blame. She does.

I've spent two years working with AI every single day, building real things with it at Frederick Studios. Everyone asks what these tools can do. Almost nobody asks what they reveal about us. That's the better question. The software is the same for everyone. The difference is human, and it shows up fast.

It isn't about technical skill, either. I've watched sharp engineers get worse results than people who barely know what a model is. The difference was never knowledge. It was what each of them was protecting.

The Gatekeeper protects accountability.

Ask why everything needs three approvals and you'll hear, correctly, that their name goes on it. They're right about the stakes. They're wrong about where to spend their attention.

Then the inversion happens. Someone who wanted speed builds a review process that costs more hours than the original work. They're not producing anymore. They're editing, and editing someone else's draft is harder than writing your own.

Worse, the review lands at the end, where changes cost the most. Bad brief, bad draft, four rounds of fixes. Nobody ever goes back to the brief.

So the relationship stays supervisory forever. They never get the good part, the moment the thing suggests something they hadn't considered. You can't be surprised by a subordinate you're auditing.

The fix isn't less control. It's control moved to the front, where their judgment actually buys something.

The Collaborator protects nothing. That's why it works.

They ask if it understood. They ask if there's a better route. They treat disagreement as information instead of malfunction.

And it compounds. Every exchange sharpens the next, because they're teaching it the shape of the problem while it shows them what's possible. This is the only mode where the work gets better over months instead of staying flat.

The failure here is quieter. Fluency breeds trust. Trust breeds drowsiness. Two hundred good exchanges make the two hundred and first easy to skim.

The Vending Machine treats it like a search box.

One sentence in. Whatever comes out is the verdict, and the verdict is usually that the whole thing is overhyped. No second attempt.

They aren't hostile. That's what separates them from the Skeptic. Nobody ever told them the first answer is a starting position.

And nobody corrects them, because admitting you got a bad result feels like admitting you asked badly. So they carry one unrepresentative sample around as a settled opinion and share it confidently at lunch.

The Impostor protects reputation.

Uses it constantly. Won't say so. Takes the credit with a small queasy feeling.

The cost isn't moral, it's practical. They can't ask a colleague for a better approach. Can't compare notes. Can't ask how anyone else handles revisions.

Everyone else is learning from each other. The Impostor learns alone, which means slowly. They plateau in month one and stay there, guarding a secret half the building shares.

The Skeptic protects truth.

They engage seriously, which is what separates them from the Vending Machine. They read closely. They assume error until shown otherwise.

Slow to adopt, slower to trust. Also the only person who catches the citation that doesn't exist before it reaches a client.

Organizations need this person and resent having them, because skepticism reads as obstruction when everyone's excited. The real risk is that being right about the failures becomes an identity. Identity is hard to update when the failure rate drops.

The Abdicator protects time. Or thinks so.

Judgment goes over the wall. The work ships unread.

That's not collaboration, it's outsourcing, and the thing being outsourced is the same faculty you need to notice something went wrong.

Skill decays without use. Six months in, they couldn't evaluate the work if they wanted to. Usually a client notices first.

Here's what ties the six together.

The Gatekeeper and the Abdicator look like opposites. They're the same error. Both are dodging the labor of judgment, one by applying it everywhere, the other by putting it down.

The axis was never pro-AI versus anti-AI.

It's whether you'll stay present in work that something else is partly doing.

Brad Frederick is the founder of Frederick Studios, a full-service creative studio that builds websites, brands, and media for the small businesses everybody else overlooks.