Your First Real Task

Reading about AI changes nothing. The gap between people who find it useful and people who do not is almost never knowledge — it is having actually used it on something that mattered.

Pick the right first task

Look for all four of these:

• You do it often. Weekly, ideally. One-offs teach you nothing repeatable.
• You dislike it. Motivation matters more than optimality.
• It transforms something you already have. Remember chapter three: it is strong at reshaping, weak at supplying.
• Being wrong is cheap. Your first attempt should not be able to hurt you.

Good first tasks: turning messy notes into a clear summary; drafting a reply to a routine enquiry; rewriting something for a different audience; pulling actions out of a long thread.

Bad first tasks: anything with a client's name in it, anything with numbers that must be right, anything you would publish unread.

Do it properly once

Write the brief the way chapter four describes: context, one task, constraints, a definition of good. It will feel like overkill for something small. Do it anyway — the point is to feel the difference a real brief makes.

Then look at the output honestly. Not "is this impressive" but "is this better than what I would have produced in the time?" That is the only question that matters.

Iterate once, then stop

Change one thing. Run it again. Notice what moved.

Then stop and use the result, even if it is imperfect. People lose whole afternoons chasing a perfect output on a task that did not warrant one.

Save what worked

Keep the brief that produced a good result. Next week, start from it rather than from a blank box. Within a month you will have a small collection of prompts that work on your actual work, which is worth more than any prompt pack you could buy.

What to do after this course

You now have the foundations: what it is, why it errs, where it is strong, how to brief it, how to check it, and what not to feed it. Everything else is application.

The library goes deeper on specific jobs — automating a process, researching without inventing sources, writing in your own voice, deciding what to work on, training a team, setting an AI policy. Pick the one closest to a problem you actually have this month.

And keep the one sentence from chapter one: it predicts plausible text; it does not look things up. Almost everything sensible about using AI follows from taking that seriously.