# Skill-Preservation AI Use Audit

Use with [What Deteriorates When You Outsource Thinking](/articles/what-deteriorates-outsourcing-thinking).

## Before you start: the four questions

1. Is this a skill I am actively building, or a task I just need finished?
2. What does an honest, unaided attempt look like, and can I timebox it (5-15 minutes)?
3. What does the assisted phase add that the unaided attempt could not?
4. How will I check, later, that I can still do the unaided version?

If you cannot name the specific skill you are protecting, this task probably belongs in "delegate
freely" rather than this audit.

## Your tracking table

| Task | Baseline (unaided attempt, dated) | Assisted phase — what AI added | Independent check (date + result) | Review cadence |
| --- | --- | --- | --- | --- |
|  |  |  |  |  |
|  |  |  |  |  |
|  |  |  |  |  |
|  |  |  |  |  |

## Suggested review cadences

- **Weekly** — for skills you use almost daily (a language you're learning, a tool you use constantly).
- **Monthly** — for skills you use regularly but not daily (persuasive writing, estimation, structured
  arguments).
- **Quarterly** — for skills you use occasionally but want sharp when needed (debugging an unfamiliar
  system, a rarely used calculation).

## The order-of-operations rule

Attempt unaided first, even briefly and even badly, before opening an AI tool for a task you are
building the skill on. Research on essay writing with LLMs found that participants who attempted a task
unaided before getting AI help retained more and re-engaged more of the brain regions associated with
the task than those who started with AI assistance from the beginning. Order matters more than total
effort.

## Not for high-stakes first attempts

Do not apply "unaided first" to tasks with real safety, legal, medical, or financial consequences if
your unaided attempt is not yet reliable. Practice on low-stakes reps; verify high-stakes work regardless
of who or what produced the first draft.

## Review this audit

Revisit your tracking table at each scheduled check-in. If an independent check shows real decline in a
skill you care about, increase the frequency of unaided reps before increasing assisted ones.