Building a Sensible AI Habit
The previous lessons each covered one task. This one is about the overall pattern — how AI fits into a working week without making your work worse or your skills softer.
The delegation test
Before handing a task to AI, ask: could I evaluate the output? Delegation works when you can judge the result. If you could write the email yourself and just want it faster, AI is a good assistant — you'll catch its mistakes at a glance. If you're asking it to do something you couldn't check — legal wording, code in a language you don't know, claims in a field you've never studied — you're not delegating, you're gambling.
This test explains most AI success and failure stories at work. The people getting real value use it for work they understand deeply. The horror stories — lawyers filing AI-invented citations, developers shipping code they never read — all involve output the person couldn't or didn't evaluate.
Keep the thinking, delegate the typing
A useful line to draw: AI handles transformation, you handle judgment.
- Turning your bullets into prose: transformation. Deciding what to say: judgment.
- Reformatting notes into action items: transformation. Confirming who owns what: judgment.
- Summarizing a report: transformation. Deciding what to do about it: judgment.
The risk of heavy AI use isn't that it fails — it's that judgment atrophies quietly while transformation gets outsourced. The fix is simple: for skills that matter to your career, keep doing some reps manually. Draft some emails cold. Read some documents in full. Not for purity — to keep your evaluation sharp, because evaluating AI output well requires the skill you'd otherwise lose.
A weekly pattern that works
- Default to AI for: first drafts from your notes, summaries for triage, reformatting, explaining unfamiliar concepts, and prep questions before meetings.
- Default to manual for: final judgment, anything sensitive (last lesson), personal messages, and one or two deliberate practice reps of your core skills.
- Review everything that leaves with your name on it. You're the byline; the AI is not.
Where to go from here
Our Prompt Engineering course covers getting better results from these tools, and the Prompt Library has ready-made templates for most tasks in this course. Start with one recurring task this week — status updates are a good first pick — and build from there.
Key takeaway: delegate what you can evaluate, keep manual reps of skills you can't afford to lose, and put your own eyes on anything with your name on it.