- Why vague prompts get vague answers — and the single change that fixes this
- The 3-part formula (Role + Task + Context) that works for almost any request
- Before-and-after prompt examples you can copy and adapt immediately
- The follow-up technique most people don't know about that gets dramatically better results
The garbage in, garbage out problem
Here's the most common experience someone has when they first try AI: They open ChatGPT, type something like "help me with my resume," get a generic three-paragraph response that barely applies to their situation, and close the tab thinking "I don't see what all the fuss is about."
What went wrong? Not the AI. The prompt.
AI tools are powerful but not psychic. They respond to exactly what you give them. A vague request gets a vague answer. A specific, context-rich request gets a specific, genuinely useful answer. The gap between those two experiences is enormous — and completely within your control.
Think of prompting like briefing a very capable contractor. The more clearly you explain the job, the materials, and the outcome you want, the better the result. Show up with "I want something done to my kitchen" and you'll get confused looks.
The good news: this is a learnable skill, and the core of it fits in a single formula.
The 3-part formula: Role, Task, Context
Almost every effective prompt includes three elements, even when the person writing it doesn't realize it. Once you see the pattern, you can't unsee it.
Before and after: see the difference
Let's look at three common scenarios and see what happens when you apply the formula.
Scenario 1: Writing help
Scenario 2: Research and explanation
Scenario 3: Everyday life help
The follow-up is where the magic happens
Here's the thing most new AI users miss: the first response is rarely the best one. AI conversation is iterative. You don't have to accept the first answer — you can push it, redirect it, and refine it until you get exactly what you need.
Think of it less like a vending machine (put in request, get one answer) and more like a conversation with a thoughtful collaborator. Some of the most useful things you can say after a first response:
"That's too formal — make it sound more casual."
"Give me three alternative versions."
"Explain the third point in more detail."
"Pretend I know nothing about this — start over."
"What am I missing that I should consider?"
Iteration is not a sign that the AI failed — it's how the tool is supposed to work. Professional AI users rarely use first-draft responses directly. They treat the first answer as a starting point.
A few more things that change results dramatically
Ask for a specific format
If you want bullet points, say so. If you want a table, ask for a table. If you want it in under 100 words, specify that. AI defaults to flowing paragraphs unless you tell it otherwise. Specifying format is one of the easiest improvements you can make.
Tell it what you don't want
Negative constraints are powerful. "Don't use bullet points." "Don't use corporate jargon." "Don't suggest anything that costs money." "Don't give me medical advice — just help me form better questions for my doctor." Constraints narrow the output toward what you actually need.
Give it examples
If you want something written in a particular style, show it an example. "Here's an email I wrote last month — match this tone." Or "Here's the kind of explanation I find helpful — apply that style to this topic." AI is very good at matching patterns when you provide them.
- "You are a plain-English financial advisor. Explain what a credit score is and give me 3 specific things I can do this month to improve mine. I'm starting from scratch with no credit history."
- "You are a thoughtful friend who is good at writing. Help me write a thank-you note to my coworker who covered for me while I was sick last week. Keep it genuine, not corporate. Under 80 words."
- "You are a doctor speaking to a patient who is anxious about their health. Explain what high blood pressure means, what causes it, and what questions I should ask my doctor at my next appointment. Don't diagnose me — just help me understand the topic."