1.1Anatomy of a good prompt (RCTF)
The essentials in 30 seconds
- Four blocks, four sentences: Role + Context + Task + Format — that is the gap between a generic answer and a near-final deliverable.
- Format is the highest-ROI block: always specify length, structure and tone (“subject + 120 words, warm tone”).
- Precise beats long: everything that changes the answer, nothing more — a precise task verb (compare, list, critique) beats “tell me about”.
Threaded mini-case — Amine, project manager in Algiers
Amine, a project manager in a services company in Algiers, must announce a schedule change to his stakeholders. His first prompt “write me an announcement about the project delay” returns a generic, unusable text. He rebuilds his request in four blocks: role (project manager announcing a milestone), context (delayed delivery, internal recipients and client), task (write the announcement email) and format (subject + 120 words, professional and reassuring tone, one clear instruction in bold). The answer becomes a near-final deliverable.
This same case returns in the lesson, the diagram, the script and the 48h challenge — and continues in 1.2, 1.3 and 1.4.
1. Four blocks
The difference between a generic answer and a near-final deliverable lies in the request. A professional prompt is written in four blocks: the role (“You are the internal communications lead”), the context (situation, audience, constraints), the precise task, and the expected format (length, structure, tone).
Four sentences are enough — but all four. Amine proves it on his schedule-change announcement: without the four blocks, the text stays generic; with them, it can be sent after a short proofread.
2. Why each block works
The role calibrates the level of language, the context eliminates false assumptions, the task with a precise verb drives the action, the format avoids the 800-word essay when you wanted 5 bullets.
Amine understands what his first attempt was missing: without context, the AI assumed a random recipient and tone; without format, it delivered a useless length. Each added block removes one assumption.
3. The reusable template
Build yourself a reusable personal template: 30 seconds of framing saves 10 minutes of rework. Role: … / Context: … / Task: … / Format: … — a simple file you fill in for each request.
Amine saves his template and reuses it for every project document: announcements, follow-ups, meeting notes. Framing becomes a reflex, not an effort.
🔎 Concrete example — 🔎 Concrete example — From lazy prompt to pro prompt
The same need, two results. Test both prompts as they are.
Lazy: “Write me an announcement about the project delay” → generic.
Pro: “You are a project manager [role]. Our delivery planned for the 15th is postponed to the 30th; stakeholders must reorganise their acceptance testing [context]. Write the announcement email [task]. Format: subject + 120 words, professional and reassuring tone, one clear instruction in bold [format].”Reading: the RCTF version specifies recipient, situation and form — the answer becomes actionable and correctly toned.
Lesson: keep your RCTF template at hand and fill it in in 30 seconds.
Section vocabulary
- Role
- The hat given to the AI: it calibrates vocabulary, level and angle of the answer.
Simple example: “You are a project manager announcing a schedule change” in Amine's prompt.
What it is not: magical expertise added to the model, which keeps its limits.
Reflex: one hat per request, adapted to the recipient.
- Context
- The situation and the audience: the richer it is, the fewer assumptions the AI invents.
Simple example: The postponed delivery and the stakeholders Amine describes before his task.
What it is not: ten lines of filler that dilute without changing the answer.
Reflex: everything that changes the answer, nothing more.
- Task
- The requested action with a precise verb: write, compare, list, critique, rephrase.
Simple example: “Write the announcement email to stakeholders” rather than “tell me about the delay”.
What it is not: a vague verb like “tell me about” that produces a lukewarm encyclopedia.
Reflex: one precise action verb per prompt.
- Format
- The output form: length, structure, tone — the most often forgotten block, the highest-ROI one.
Simple example: “Subject + 120 words, professional tone, one instruction in bold” as required by Amine.
What it is not: letting the AI choose: an 800-word essay when you wanted 5 bullets.
Reflex: always length + structure + tone.
In practice — In practice — The same need, from lazy prompt to pro prompt
- Lazy version: “write me an announcement for the summer closure” — note the generic answer.
- RCTF version: “You are the office manager of a 12-person accounting firm [role]. We are closed from 4 to 18 August; clients must anticipate their document requests [context]. Write the announcement email to clients [task]. Format: subject + 120 words, warm and professional tone, one clear instruction in bold [format].”
- Compare both answers line by line: precision, tone, actionability.
- Save your template in a personal file: Role: … / Context: … / Task: … / Format: … — you will now fill it in in 30 seconds.
Ready-to-say script — explain RCTF to a colleague (30 s)
“For a good prompt, give four things: who you want the AI to be, the situation and the recipient, the precise action, and the form — length, structure, tone. Example: project manager, delivery postponed to the 30th, write the announcement, subject + 120 words. Thirty seconds of framing, ten minutes of rework saved.”
Say it in pairs; show your filled-in template on one of Amine's real cases.
Copy-paste model — RCTF template
ROLE: You are _______________
CONTEXT: _______________ (situation, recipient, constraints)
TASK: _______________ (precise verb: write, compare, list, critique)
FORMAT: _______________ (length + structure + tone)
SENSITIVE DATA: anonymised [ ] (see level 5.1)Beginner in a hurry
Remember just one thing this week: always add the FORMAT (length, structure, tone) to your current prompts — it is the highest-ROI block on its own.
Project manager in Algiers
Amine keeps his template in a file shared with his team: same blocks, same habits — project follow-up prompts become consistent.
Sensitive request
If the context contains internal data: anonymise before prompting (fictitious names and figures) — the SHAPE of the context is enough to frame the answer, see 5.1.
Already long prompt
A long block of text that does not work? Remove the filler and check the four blocks one by one: it is almost always the format or the task verb that is missing.
✓ Best practices
- The 4 blocks every time.
- Format: length, structure, tone.
- Reusable template at hand.
✗ Common pitfalls
- Forgetting the format.
- Vague verb (“tell me about”).
- Confusing long with precise.
Key takeaways
- Role + Context + Task + Format: the 4 blocks of a professional prompt.
- Format is the highest-ROI block: always specify length, structure and tone.
- A precise task verb (compare, critique, list) beats “tell me about”.
- Keep your template at hand: 30 seconds of framing saves 10 minutes of rework.
48h challenge — apply tomorrow
Take one real email to write and run it through the RCTF template (model above): compare with your usual prompt. 15 minutes is enough.
- I can name the 4 RCTF blocks and the effect of each.
- I know why format is the highest-ROI block.
- I compared lazy vs RCTF and saved my template.
Frequently asked questions
Is a long prompt always better?
No: a PRECISE prompt is better. Ten lines of useful context help; ten lines of filler dilute. The right measure: everything that changes the answer, nothing more.
Does the role really work or is it a placebo?
It measurably steers the register and priorities of the answer (try “explain VAT” as a primary-school teacher, then as a tax specialist). It confers no magical expertise: the model's limits remain.
What are the 4 blocks of a good prompt?