Which work should stay yours

Some tasks are fine to hand over. Some are the ones that make you good at your work. Telling them apart is the whole skill.

The short answer

Decide what stays yours before you start, not at each prompt. Keep the first draft, the framing of the problem, and the final judgement. Hand over the mechanical middle: formatting, boilerplate, translation, checking your own work. Write the rule down where you work, because a rule you have to re-decide under deadline is not a rule.

Why one rule beats good intentions

Deciding case by case means deciding while tired, behind and mid-task. That is the same reason app limits fail: they ask you to make the right call at the exact moment you are least able to.

A rule set written down once removes the decision. It is the same structural move as charging the phone in another room.

Keep these three

• The first draft. Ten minutes, badly, nothing open. This is where your own view of the work appears, and there is no other place it appears.
• The framing. What is the actual problem, who is it for, what would count as done. A tool cannot be wrong about this on your behalf, it can only be confident.
• The final judgement. Is this true, is it good, would you sign it. If you cannot answer, the work is not finished no matter how it reads.

Hand over these

• Boilerplate and formatting you have written a hundred times.
• Translation, restructuring and tightening of text you already wrote.
• Checking your own work for the errors you know you make.
• Lookups you would otherwise open six tabs for.
• The mechanical middle of anything where you already decided the shape.

None of that was practice. Giving it away costs you nothing and buys back the time to do the three you kept.

Write it where you work

One line, visible: draft mine, middle theirs, call mine. Put it in the document template, the repo readme, the notebook, wherever the work starts. A rule you cannot see is a rule you will renegotiate.

Same logic applied to the working day is in [screen time at work](/journal/screen-time-at-work).

Where this usually breaks

It breaks the first time the ten minute draft would make you late. Once you skip it under deadline, the rule becomes advice, and advice does not survive a busy week.

The rule holds better after a day of working with nothing to reach for, because you find out what you can still do alone. That is DEAD AIR: ten sealed offline hours and a written plan for the days after.

Asked next

What should I never hand to AI?

The first draft, the framing of the problem, and the final judgement about whether the work is true and good. Everything mechanical in between is fair game.

How do I stop over-relying on AI at work?

Write one rule down where the work starts and follow the order: your rough version first, the tool second. Deciding case by case fails under deadline.

Is it cheating to use AI for the middle of the work?

No. Formatting, tightening and checking were never the practice. The part worth protecting is where you decide what you think.

Next

Join even hand free with an email and a city at the community, or apply for the 48 hours.