Prompt-craft framework

Let me bust a myth right at the start. There’s a whole industry calling itself “prompt engineering” — selling courses, building consultancies, writing books. Most of it is overcomplicated.

Talking to AI well is not engineering. It’s the same skill as briefing a smart but inexperienced assistant on a job. You give them context, you tell them what you want, you tell them how you want it, and you ask them to come back with questions if anything’s unclear. That’s ninety-five percent of it.

The reason most people get bad answers from AI is the same reason most people get bad answers from new employees: they didn’t explain what they actually wanted.

The basic structure of a good prompt

Every solid prompt has four parts, even if you write them in plain English without thinking about it as a structure:

  1. The role. Who do you want the AI to be? A research assistant? A skeptical advisor? An editor? A friend who knows about cars?
  2. The job. What do you want it to actually do?
  3. The context. What does it need to know to do the job well?
  4. The format. How do you want the answer? A list? A short paragraph? A step-by-step plan? A side-by-side comparison?

Here’s the basic template, written like a normal sentence:

Act as [role]. Help me with [job]. Here’s the context: [the relevant details]. Give me [format]. If anything’s unclear, ask me first.

That last bit — “if anything’s unclear, ask me first” — is the move that elevates most prompts. Without it, AI tools tend to guess. With it, you often get a smart clarifying question that makes the final answer better.

Five prompts every grown man should have in his back pocket

1. The Plain-English Explainer

For when you’re trying to understand a topic without drowning in jargon.

Explain [topic] in plain English for a capable adult who is new to it. Skip jargon. Give me the 80/20 version, what to watch out for, and three things people commonly get wrong about this.

Why it works: it tells the AI exactly who it’s talking to (a capable adult, not a beginner or expert), specifies the depth (80/20), and asks for the common pitfalls — which is often the most useful part of any explanation.

2. The Decision Helper

For when you’re comparing options and want a clear recommendation.

Compare [option A] and [option B] for someone who values reliability, long-term cost, ease of use, and not wasting money. Be direct. Give me a verdict and the situations where you’d change your mind.

Why it works: it makes your priorities explicit, asks for a decision rather than a balanced overview, and creates space for nuance through the “situations where you’d change your mind” line.

3. The Plan Builder

For when you have a goal and need a real path.

Build me a practical step-by-step plan to [accomplish goal]. Include: the first concrete action this week, common mistakes beginners make, what tools or resources I’ll need, and what success looks like at 30, 60, and 90 days.

Why it works: it forces a real plan with specific milestones instead of a generic outline. The 30-60-90 framing is gold because it makes the AI think in time horizons.

4. The Devil’s Advocate

For when you want your idea pressure-tested before you commit to it.

Here’s my plan: [describe plan]. Be a skeptical, experienced advisor. Find weak spots, hidden assumptions, things that could go wrong, and better options I might be missing. Be direct, but practical — not pessimistic for the sake of it.

Why it works: it explicitly asks for the negative case, which AI tools won’t do unless you ask. The “practical, not pessimistic” qualifier keeps the response useful instead of doom-laden.

5. The Email Rewriter

For when you’ve drafted something and you want it to land better.

Rewrite this email so it’s clear, calm, direct, and respectful. Keep my meaning and my voice — don’t make it sound corporate. Trim anything that’s not needed. Here it is: [paste email]

Why it works: “keep my voice” is the magic phrase. Without it, AI tends to flatten your writing into generic professional-speak.

The single technique that improves every prompt

Don’t stop at the first answer.

The biggest mistake new AI users make is treating the tool like a search engine — type a question, take the first result, move on. AI is a conversation, not a query.

After the first response, the move is almost always one of these:

  • “Make it more practical.”
  • “What am I missing?”
  • “Push back on that — what could go wrong?”
  • “Give me three more options I didn’t consider.”
  • “Explain that part in more detail.”

Each of those second-pass prompts takes ten seconds and dramatically improves the quality of what you get. The first answer is rough draft. The second pass is where the work gets useful.

The thing nobody tells you about prompts

Your prompts will get better the more you use AI, without you trying. After a couple weeks, you’ll naturally start including context, specifying format, asking for second passes — because you’ve seen what works and what doesn’t. It becomes a feel.

That’s why the actual answer to “how do I get better at prompts?” isn’t a course or a checklist. It’s using the tool every day for a month and paying attention to what produces good answers.

The five prompts above will give you a head start. The skill comes from the habit.