Press Enter to search  ·  Esc to close

The 4-Pillar Framework: why 3-word prompts always fail.

Most professionals type something like "summarise this" or "write an email" and then wonder why Claude gives them something generic and unusable. The problem is not Claude — it is the prompt. Without context, Claude has no choice but to guess.

The 4-Pillar Framework gives Claude everything it needs to produce a precise, professional output every single time. The four pillars are Role, Context, Task, and Format — and they take under 30 seconds to apply.

1. Role
Tell Claude who to be.
"Act as a senior financial analyst..."
2. Context
Tell Claude what the situation is.
"...our Q3 budget came in 18% under target..."
3. Task
Tell Claude exactly what to do.
"...write a 3-bullet executive summary..."
4. Format
Tell Claude how to present the output.
"...in plain English, no jargon, under 100 words."

The rule: if your prompt has fewer than 15 words, you are almost certainly missing at least two of the four pillars. Add them before you hit send.

The 4-Pillar Framework works across every corporate role — but the specific language you use matters. A prompt that works brilliantly for a Finance analyst will feel wrong for an HR manager. Here are three ready-to-use examples you can adapt immediately.

HR
❌ "Write a rejection email"
✅ "Act as an HR Manager. Draft an empathetic, policy-compliant rejection email for an internal candidate. Focus on growth feedback. Under 150 words."
Finance
❌ "Summarise Q3 results"
✅ "✅ "Act as a CFO preparing a board update. Summarise our Q3 results in 3 bullets: performance vs target, key risks, and recommended action.""
Marketing
❌ "Write a LinkedIn post"
✅ "Act as a B2B content strategist. Write a LinkedIn post announcing our new HR software integration. Tone: professional but conversational. Max 150 words. End with a question."

Notice that every good prompt includes all four pillars. Save your best prompts in a Google Doc — they become reusable assets that save hours every week.

Even a well-structured prompt sometimes produces output that is close but not quite right. Instead of starting over or accepting a mediocre result, use the Refine Loop — a single follow-up prompt that forces Claude to self-correct.

The Refine Loop Prompt
Review your last response against my original request. Check: did you follow the role, context, task, and format I specified? If any element is missing or weak, rewrite the response now with those gaps fixed. Do not explain what you changed — just give me the improved version.

This works because Claude is genuinely good at evaluating its own output when asked directly. The key is telling it not to explain — you want the result, not a commentary on the result.

When to use it: any time the output is 80% right but missing something specific. One Refine Loop prompt is almost always faster than rewriting your original prompt from scratch.