Vague prompt in, vague answer out. Most bad AI answers trace back to a prompt missing one of four things. This is the framework: Role, Task, Format, Constraint (RTFC).
Role — who should the AI act as?
Give it a perspective. “Act as a skeptical editor”, “act as a senior backend engineer”, “act as a first-time customer”. A role narrows the tone and priorities without you having to spell them all out.
Task — what exactly do you want done?
Replace vague verbs with precise ones. Not “help me with this email” — instead “shorten this email to 3 sentences while keeping the ask clear.”
Format — what shape should the output take?
Bullet list, table, JSON, a single paragraph, a script with numbered scenes — say it explicitly. Models default to prose unless told otherwise, and prose is rarely what you actually need for a re-usable output.
Constraint — what are the boundaries?
Length limits, tone, things to avoid, required elements. “Under 100 words”, “no jargon”, “must include a call to action” — constraints are what separate a generic answer from one you can use as-is.
Putting it together
Act as [role]. [Task, stated precisely]. Output as [format]. Constraints: [list].
Example: Act as a skeptical technical reviewer. Review this product description for unsupported claims. Output as a bullet list, one issue per line. Constraints: under 150 words, cite the specific sentence for each issue.
Why this beats “be more specific”
“Be more specific” is true but not actionable. RTFC gives you four concrete slots to fill, and if any one of them is missing, that’s usually exactly where the output falls short.
Leave a Reply