Build one now
Six fields that force the clarifications people usually skip.
Read the guide →You do not need a course. You need to stop leaving out the five things that actually decide whether the answer is usable.
A language model cannot ask your colleague, open your files or read the meeting you just left. When a brief leaves out the goal, the reader and the constraints, the model does the only thing it can: it produces the statistically safe, middle-of-the-road version. That is what generic AI prose actually is. It is not laziness on the model's part, it is missing input.
So the fix is rarely a cleverer phrasing. It is supplying what was missing.
A verb and a deliverable. “Draft a one-page decision memo recommending whether we renew the contract.” Not “help me with the contract.”
What the model cannot know and would otherwise invent. Situation, audience, prior decisions, the actual data.
Length, tone, what must be included, and explicitly what it must not do. “Do not invent figures I have not given you” prevents a whole class of failure.
How the result will be used. A table, an email, a memo with a recommendation, a JSON object with named fields.
What a good answer looks like, so you can judge it. Cite sources, flag assumptions, mark anything uncertain.
Here is the difference in practice. The second version is not longer for the sake of it. Every added sentence removes a decision the model would otherwise have guessed.
Write an email of at most 120 words to a business customer waiting on a delivery.
Context: the order has slipped from 28 to 31 July. This is our fault, a warehouse error. The customer has ordered from us for three years.
Constraints: apologise specifically without blaming the carrier. Confirm the new date. Invite them to reply if that date does not work. Do not offer a discount, and do not add any detail I have not given you.
Format: subject line, then the body. No bullet points.Six fields that force the clarifications people usually skip.
Read the guide →Interactive before-and-after on three realistic tasks.
Read the guide →The reasoning behind each of the five parts.
Read the guide →