How to Write Better AI Prompts: The RCT Framework

Ask most people why AI gave them a mediocre answer and they blame the AI. Usually it was the prompt. "Write me a marketing email" is not an instruction. It is a wish.

The fix is not a secret list of magic words. It is structure. The one we teach is RCT: Role, Context, Task. Get those three right and the quality jumps immediately.

Why most prompts fail

A vague prompt forces the model to guess. Guess your audience, your tone, your goal, your constraints. It fills every gap with the blandest, most average option, because average is the safest guess. That is why so much AI output reads like nobody in particular wrote it for nobody in particular.

You are not trying to get an answer. You are trying to get your answer. That takes specifics.

R is for Role

Tell the model who to be. "You are a B2B email copywriter who writes short, direct emails for busy finance directors." A role sets tone, vocabulary, and assumptions in one line. The difference between a generic reply and a sharp one often comes down to this alone.

C is for Context

Give it the situation. Who is the audience, what do they already know, what is the goal, what must it avoid. "We sell HRD Corp claimable AI courses. The reader is an HR manager who has heard of AI but has not booked a course. Keep it under 120 words. Do not use the word 'revolutionary'."

Context is where people get lazy, and it is the part that matters most. The model cannot read your mind or your last quarter's numbers. Tell it.

T is for Task

State exactly what you want it to produce. Not "help with an email" but "write three subject lines and one 100-word email body with a single call to action." A precise task gives you something you can use, not something you have to interpret.

Put together

Role: You are a B2B copywriter who writes plain, direct emails.

Context: We sell HRD Corp claimable AI courses in Malaysia. The reader is an HR manager, mildly interested, not yet booked. Under 120 words. Avoid hype words.

Task: Write three subject lines and one email body with one clear call to action.

Run that and compare it to "write me a marketing email." It is not close.

The real skill is iteration

The first answer is a draft, not a verdict. This is the part almost everyone skips, and it is the part that matters most.

Here is what changed. The newer reasoning models already work through problems step by step on their own, so the old trick of typing "think step by step" barely moves the needle anymore. What still separates people who get great output from people who get average output is not a magic phrase. It is that they go back and forth.

Treat the model like a junior writer you are coaching, not a vending machine you hit once. That means:

• Say what is wrong, specifically. Not "make it better" but "the second paragraph is too formal, and the call to action is buried."
• Change one thing at a time. Fix the tone, look, then fix the length. Piling five edits into one instruction gets you mush.
• Show it what good looks like. Paste an example of the style you want. One good example beats a paragraph of description.
• Build on the best version. When a draft is close, tell it to keep that and adjust from there, so you do not lose the good bits.

Most people run one prompt, get a so-so answer, and either use it or give up. The people pulling ahead run three or four turns in the time it takes to complain. The tool is fast. The bottleneck is whether you bother to push back.

None of this is hard. It just takes a few turns more than most people are willing to spend, which is exactly why the people who do it pull ahead.

If you want your whole team writing prompts like this instead of a couple of enthusiasts, that is what our Mastering AI Prompting workshop is for. It is the know-how to get consistently more useful outputs, and the investment is only a half-day course. It is HRD Corp claimable too.