Prompting is a brief. And you already know how to brief.
tier I · foundations · 8 min · interactive article ✦
Today's good news: the most rewarding AI skill for a manager isn't technical. It's a skill you've been practicing for years with your teams, your vendors, your agencies — the art of the brief. Prompting is nothing else. And like any brief, there's a method.
Why your prompts disappoint
“Summarize this document for me.” Result: a generic summary, too long, with no angle. The problem isn't the model — it's the brief. You'd never hand a consultant a mission in one sentence; the LLM deserves the same framing. See for yourself:
“Summarize this steering-committee report.”
→ Result: 800 words, neutral tone, no angle, secondary points on the same level as the decisions. Unusable as is.
“Act as a senior PMO. Here is the steering-committee report [pasted]. My executive committee wants a 1-page brief: 3 wins, 2 alerts, 1 decision to make. Factual tone, no project jargon.”
→ Result: exactly the page expected, ready to review. Same model, same document — only the brief changed.
The structure that delivers: role, context, task, format
the RCTF structure — the 4 building blocks of a prompt that delivers
Role — “Act as an experienced CFO”: you calibrate the tone, the level of rigor and the vocabulary.
Context — your situation, your constraints, your audience: “this document will be read by the executive committee, who didn't follow the project.” It's the block 90% of users forget.
Task — precise and bounded: “identify the top 3 risks and propose one action per risk.”
Format — “a 3-column table: risk, impact, action. 200 words maximum.” Say it, or you'll get a wall of text.
Two next-level techniques
The example (few-shot): show an example of the expected result — “here's the format of a good one-pager, produce the next one on this model.” The model imitates remarkably well. Iteration: the first result is a working draft, not a final copy. “Shorter. More direct. Add the Q3 figures.” The best users don't write better prompts on the first try — they iterate faster.
test yourself — just like in the path
You need to ask the AI for an analysis of your sales pipeline for the executive committee. Which prompt starts off right?
✓ Role, context, bounded task, imposed format: all four blocks are there. It's exactly what you assemble, piece by piece, in the level 3 exercise.
Blocks are missing — role? format? scope? Try again.
Going deeper
Should you say “please” to an AI?
No measurable effect on quality — but no harm either. What matters: clarity, structure and context. Save your energy for the “context” block, that's the one that pays.
The 3 mistakes that ruin a prompt
1) The open question with no scope (“tell me about our market”); 2) stacking 6 requests into one sentence — break it into steps; 3) the missing output format, which dooms you to a wall of text to rework.
The reflex to install
Before sending an important prompt, reread it as if you were handing it to a brilliant new hire who knows nothing about your company. What they'd be missing to succeed is what your prompt is missing.
Frequently asked questions
What is a good prompt?
A good prompt gives a role, a context, a precise task and an output format. It's a structured brief — like a clear instruction to a colleague — not a magic spell.
Are there magic words for AI?
No. “Incantations” are folklore. What truly changes the result is the clarity of the brief and, often, an example of the expected output.
Why does the AI give me generic answers?
Most often because the prompt lacks context. Add your precise situation and an example of the result you want: the answer immediately gets targeted.
Will prompting become useless?
No. Clearly stating what you expect remains a durable skill, whatever the interface: it's framing, not a technical trick.
This notion is a full level of the IAPLC path.
In the course, you practice it on YOUR context: lessons generated for your job, interactive exercises, gamified progression.