AI Guides › Getting Started

Choosing Your First Real Task

By Nigel Guy · 2 min read

Once the practice sessions are done, the next mistake is picking the first real task by ambition instead of by whether you'll actually be able to tell if it went well. People reach for the biggest, vaguest thing on their list — "help me rethink my whole strategy," "write my business plan" — and come away either falsely impressed or quietly unable to judge what they got.

The mechanism: what makes a task checkable

1. You know roughly what a right answer looks like already. Not the exact wording, but the shape — the facts it should include, the structure it should follow, the tone it should hit.
2. It has a clear finish line. A task that ends when you say so, not one that's open-ended by nature (strategy, "figure out my career," a problem with no single right shape).
3. Getting it wrong costs you five minutes, not a client relationship. Save the first attempt for something reversible.
4. It's genuinely useful, not a toy example. A checkable task doesn't have to be trivial — drafting a real email, summarising a real document, restructuring a real spreadsheet formula all qualify, as long as you can judge the output.

A few examples that fit

• Rewriting a document you've already drafted, so you have the original to compare against.
• Summarising a long document or transcript you've actually read, so you can check what it kept and what it dropped.
• Debugging or explaining a specific error you can reproduce, rather than "review my whole codebase."
• Drafting a reply to a specific email, where you already know what you want to say and are testing whether it says it well.

What to skip

Skip open-ended strategic questions for your very first real task — "what should my business focus on" has no checkable shape yet, and you'll have no way to tell a good answer from a plausible-sounding one. Skip anything where being wrong is expensive and hard to reverse — that's a task for later, once you've built up judgement on smaller ones. And skip picking the task that would impress someone else watching; pick the one that actually teaches you something about how well this works for you.

Guardrails

• A checkable first task isn't a permanent constraint — it's a starting point while you build calibration, not a rule for every task forever.
• Being able to check an answer still means actually checking it — picking a verifiable task doesn't help if you skip the verifying step anyway.
• If your real work is mostly open-ended by nature, that's fine — start with the smallest checkable slice of it you can find, rather than the whole open-ended thing at once.

All 751 AI guides · JulieMango plans from £17/mo