How to Write Better AI Prompts: The Lean Framework
A simple 4-part framework for writing AI prompts that get consistent, professional results — no jargon, no prompt-engineering PhD required.
Most “prompt engineering” advice online is either too abstract (“be specific!”) or too much (a 12-part template you'll never actually use). This is the version I actually use — 4 parts, on one sticky note.
The Lean prompt framework: R.G.C.F.
- Role — who is the AI pretending to be?
- Goal — what should be true when it's done?
- Context — what does it need to know that it can't guess?
- Format — what should the output look like?
See it in action
Weak prompt
Write me a LinkedIn post about our new feature.
Same request, with R.G.C.F.
Role: You're a B2B copywriter who writes like a calm senior product person. Goal: Make senior operations leaders click through to read the launch post. Context: The feature is [X]. It saves them ~2 hours per week on [Y]. Our voice avoids hype and emoji. Format: 3 short paragraphs, ≤ 120 words total. Hook, insight, single CTA link.
Same model. Same feature. Wildly different output.
The 3 mistakes that kill 80% of prompts
- No role → the AI defaults to a bland assistant voice.
- Fuzzy goal (“make it good”) → you'll get generic. Say what success looks like.
- No format → you'll get a wall of text you have to reformat by hand.
One-line upgrade
Add this to any prompt that isn't working: “Before you answer, ask me 3 questions if anything is unclear.” It cuts back-and-forth by half.
Save your best prompts
The people getting the most out of AI aren't better writers. They just stopped starting from scratch. Keep a doc, a Notion page, or a spreadsheet with your best R.G.C.F. prompts, one per row. In a month you'll have a library that's more valuable than any $99 prompt pack — because it's tuned to you.
Want a printable one-pager version of this framework plus 20 ready-made prompts? Grab the free PDF on the homepage.
Get the free Agentic Workflow Starter Kit
Five ready-to-run agent briefs with guardrails and acceptance criteria built in.
Get the KitKeep reading
Claude Bioweapon Flags, Distillation Accusations, and ChatGPT Comes for Junior Bankers: Sept 11 AI Roundup
Anthropic blocks suspected bioweapon misuse and renews distillation claims, Instacart ships Clementine, Chesky says surf the wave, and OpenAI targets junior bankers.
Frontier AINvidia Joins the AI Model War: Nemotron Is More Than a Free Model
Nvidia now supplies the chips and open AI models. What Nemotron means for OpenAI, Anthropic, enterprise costs, and model choice.