The quality of an AI result often depends on the quality of the decision made before the first prompt.
A vague request such as “make this better” leaves the tool to guess what better means. It may make the work longer, smoother or more impressive while missing the person who must use it and the result it must produce.
You do not need an enormous prompt. You need a short brief that removes the important guesses.
The seven-part AI brief
- End result: What must exist when the task is finished?
- Audience: Who will read, use or decide from it?
- Purpose: What should that person understand, feel or do next?
- Source material: Which facts, files and examples should the AI use?
- Format: What shape, length and structure should the result have?
- Good enough: What must be true before the work can stop?
- Boundaries: What should the AI leave out, avoid or refuse to invent?
1Name the finished result
“Help me with my website” is not a finished result. “Write a standalone article that explains how to brief AI to a cold reader” is.
Name the thing you expect to receive: a decision memo, a lesson plan, a comparison table, a first draft, a list of verified errors or a page ready to publish. This turns a conversation into a task.
2Tell it who the work is for
The same subject needs different language for a new customer, a colleague, a child, a lawyer or somebody who has never heard of you.
Do not merely say “make it simple.” Say who the reader is, what they already know and why they are here. That gives the AI a real standard for deciding what needs explanation.
3State what the work must do
Content is not useful because it exists. It is useful because it helps somebody understand a problem, make a decision or complete an action.
Finish this sentence: “After using this, the reader should be able to…” If you cannot finish the sentence, the AI cannot reliably choose what belongs in the answer.
4Give it the sources that matter
If the task depends on your experience, a company policy, customer language, a document or a set of facts, provide those materials. Ask the AI to distinguish what came from the sources from what it is inferring.
Examples are more useful than piles of adjectives. One real article can show what “clear and human” means more precisely than ten instructions about tone. Give only the examples that define the current task. An entire archive can create as much confusion as no context at all.
5Choose the format before the answer arrives
Specify the useful form: five bullets for a meeting, a 900-word article, a two-column comparison, a three-minute script or a plain-language email.
Format is not decoration. It determines whether the result can be used. If a busy reader needs to make one decision, a long report may be a worse answer even when every sentence is intelligent.
6Define “good enough” while you are still clear-headed
Without a finish line, every AI answer creates another possible improvement. Decide in advance what the work must contain, what it must not contain and how accurate or polished it needs to be.
Good enough may mean every factual claim is sourced, every instruction is usable without prior context and the final piece fits on one page. Once those conditions are met, stop. More output is not automatically more value.
7Ask for one complete job
Do not ask the tool to solve the article, marketing plan, product, website and next six months in one conversation. Ask for one result you can inspect and use.
If the answer misses, diagnose the miss. Was the audience wrong? Was a source missing? Was the format unsuitable? Give one focused correction. If the second result still misses, stop prompting and repair the brief or choose a different method.
A complete example
- Weak request
- Write something about AI productivity. Make it engaging.
- Clear brief
- Create a 900-word standalone article for freelancers who use AI but keep working longer. Help them leave each AI session with one finished piece of work. Use the attached notes as the source. Open with the problem, explain a five-step method, include one worked example and end with a checklist. Do not invent research or promise that AI will save a specific number of hours. It is good enough when a new reader can use the method without knowing me or reading another page.
Keep the final judgement human
A good brief improves the odds of useful work. It does not transfer responsibility to the tool.
You still decide whether a claim is true, whether an example is fair, whether the advice fits the situation and whether the final work is worth putting in front of another person. AI can help assemble and challenge the work. It cannot carry the consequences of publishing it.
Before you prompt, check six things
- I can name the finished result in one sentence.
- I know who will use it and what they need from it.
- I supplied the relevant facts, files and examples.
- I chose a format that fits the real use.
- I defined what good enough means.
- I know which decisions remain mine.
Next
Give the whole AI session a stopping point
A good brief defines the work. The AI Sanity System adds the time limit, revision limit and end-of-day boundary.
Use the AI Sanity System →Want help applying this to your work?
Private coaching is available for people who want a clearer, more deliberate way to use AI and online tools without letting the work consume the day.
Explore private coaching →
Michael has spent much of his working life online, and much of the rest of it trying to stop technology from taking over. He writes about what helped, what failed and how to keep the useful parts of digital life.