Chapter 2 · Lesson 1
Vague In, Vague Out
Turn an underspecified request into a brief AI can act on without making important guesses.
- 01SituationSee the problem in context
- 02Build the ideaRead and compare examples
- 03Make the callComplete the quick exercise
Lesson work
You ask AI to “make a launch plan.” It returns ten familiar tactics, but none fit your time, audience, or budget.
AI fills gaps with averages
A model cannot see your hidden goal. When audience, outcome, source material, or limits are missing, it fills each gap with a plausible default. The answer can sound polished while solving the wrong version of the task.
Add the details that change the answer
Name the audience and desired outcome first. Then add real constraints, the material AI may use, and the format you need. Skip background that would not change the result. A clear brief is selective, not merely long.
Build the mental model
Prompt quality is often assignment quality. Make the important decisions visible before asking AI to produce the work.
Every gap becomes a guess
If you do not name the audience, AI imagines a broad one. If you do not set a goal, it chooses a common one. If you do not define limits, it optimises for a safe, average answer.
Specific does not mean long
A useful prompt can still be one sentence. The important part is including the few facts that materially change the answer: who it is for, what success means, and what must fit.
Separate inputs from instructions
If AI must work from notes, a report, or customer feedback, label that material clearly and say whether it may add outside knowledge. This prevents source text from being mistaken for a new instruction.
A solo app founder needs a launch plan they can actually complete.
Make me a launch plan for my app.
Create a seven-day launch checklist for a solo iOS app founder. I have 30 minutes per day, no advertising budget, and an existing LinkedIn audience. Give me one concrete action per day in a table.
The improved version fixes five decisions: owner, timeframe, available effort, usable channel, and output format. AI can now prioritise realistic actions instead of listing every common launch tactic.
Why missing context produces an average answer
AI generates likely continuations from patterns learned across large amounts of text. With a broad request such as “make a plan”, many different plans could be reasonable. The response therefore tends toward familiar, widely applicable advice.
Context changes which patterns are relevant. Constraints are especially useful because they remove attractive but impossible options. You are not teaching the model a secret formula; you are reducing ambiguity in the assignment.
Before sending, use a quick test: could two reasonable people read this request and imagine very different deliverables? If so, add the one missing decision that would bring them closer together. If the first answer is still weak, diagnose the missing decision instead of simply asking AI to “make it better”.
Rewrite one request you will actually use
Start with a task from this week. Add an audience, an observable outcome, one real constraint, and the output shape. Read it once and remove any detail that would not change the answer.
Using [approved input], help me [task] for [audience]. The result should help them [outcome]. Keep within [constraint]. Return [format]. If information is missing, [ask / mark it as missing].prompt choice
Which request gives AI enough direction without adding irrelevant detail?
Compare the prompts closely. Look for useful context, constraints, and a clear output.
Lesson complete · Progress saved in this browser.
Do not make a prompt longer by default. Make the decisions that change the answer visible.