Why Instructions Matter More Than the Model
Why an AI model answers exactly what you typed rather than what you meant, and how to close that gap with one worked example.
Sloppy input, sloppy output
An AI model doesn't fill in gaps the way a coworker who knows your business would. It answers the words you actually gave it. Leave out the audience, the format, the constraints, and it guesses, usually toward the most generic version of your request.
Same underlying question, two very different results:
A return policy typically outlines the timeframe customers have to return items, any conditions (like original packaging or receipts), refund methods, and exceptions for certain product categories. It's meant to build trust while protecting the business from misuse...
You can return your boots within 30 days of delivery for a full refund, as long as they're unworn and still have the original tags attached. Once we receive them, refunds are processed to your original payment method within 5–7 business days. To start your return, just click "Start a Return" in your order confirmation email and print the prepaid label.
The second prompt wins because it names the topic, the audience, the tone, and the shape of the output. The model isn't smarter here; it just has less room to guess.
Think of a recent AI answer that missed the mark. Rewrite the prompt using this shape:
"Explain [specific topic] in [specific style] for [specific audience] in [specific format]."
Notice how much of the ambiguity disappears once those four blanks are filled in: that's the whole trick this chapter is teaching.