Trust‑but‑verify in 90 seconds
Lightweight checks for facts, numbers, and assumptions—without slowing you down.
When AI outputs feel generic, it’s usually not an “AI problem”—it’s an input problem. This short brief makes your request specific enough to be useful, without turning prompting into a full-time job.
Fill these in before you prompt. You’ll get fewer “fluffy” responses and more usable drafts.
1) Goal: What are we trying to produce (and what action will it enable)?
2) Audience: Who is this for, and what do they already know?
3) Context: The key facts, background, and constraints the model must use.
4) Guardrails: What to avoid (tone, claims, policy, sensitive info, length).
5) Definition of done: Format + bullets + examples + what “good” looks like.
These five lines reduce ambiguity. That’s how you get outputs that match your intent, your audience, and your real constraints.
Replace brackets with your details, then paste the whole thing into your AI tool.
Goal: Draft my weekly project update for leadership.
Audience: Directors; they want clarity, risks, and next steps. No jargon.
Context: Project = [name]. Wins: [1-2 bullets]. Blockers: [bullets]. Metrics: [numbers]. Timeline: [dates].
Guardrails: No claims we can’t support. Keep it concise. Use my tone: direct, calm, no hype.
Definition of done: 120–160 words, 3 sections (Wins / Risks / Next), bullets, include 1 ask.
If you have an older doc you like, paste a short excerpt and say: “Match this voice.” Even 3–5 sentences helps.
Keep your outputs accurate and your inputs safe.
Lightweight checks for facts, numbers, and assumptions—without slowing you down.
Privacy-first guidance plus safer alternatives and redaction habits.