ChatGPT prompting for work: the simple formula

Here’s what this looks like in real life: you paste a message and ask “reply to this,” and the AI responds with something that sounds polite… but not like you. Or it misses the one policy constraint that actually matters.

The fix is surprisingly small: give the model the same brief you’d give a teammate. That’s it.

One nice side effect: when you use this formula consistently, you build a personal library of prompts that feel like your voice. Over time you’ll edit less and trust the drafts more.

A good “after” prompt usually includes one of these: an audience, a constraint, or a required format. Notice how the examples below force the output to be usable.

Example 1 — Email reply
Before: “Reply to this customer.”
After:
Role: Customer support rep.
Goal: Reply to a customer upset about a delayed delivery.
Context: Order shipped 2 days late. We can offer 10% off next order. We can’t refund shipping.
Rules: Empathetic, no blame, no legal language, under 120 words.
Output: Email with subject line + 2 short paragraphs + one clear next step.

Example 2 — Weekly update
Goal: Draft my weekly update for leadership.
Audience: Directors; they want risks and next steps.
Context: Wins: [bullets]. Blockers: [bullets]. Metrics: [numbers]. Timeline: [dates].
Rules: Direct tone, no hype, keep it concise.
Output: 120–160 words with 3 sections (Wins / Risks / Next) and 1 ask.

Tip: if you only change one thing, change the output format. “Give me a table” or “give me three options” will instantly make results feel less random.

1) Email drafts
Goal: Draft an email reply.
Audience: [customer/colleague].
Context: [facts].
Rules: Tone = friendly professional. Under 120 words. Avoid promises we can’t keep.
Output: Subject line + 2 short paragraphs + 1 clear next step.

2) Summaries → action
Goal: Summarize these notes.
Context: [paste notes].
Rules: Don’t invent facts. Mark unknowns as questions.
Output: Summary + Decisions + Action items (owner, deadline) in a table.

3) Brainstorming that doesn’t waste time
Goal: Generate ideas for [topic].
Context: Audience = [who], constraints = [budget/time].
Rules: No generic ideas. Each idea must include a concrete example.
Output: 15 ideas, then rank top 5 by impact/effort with 1-sentence rationale each.

4) Reports
Goal: Turn these bullets into a first-draft report.
Context: [bullets].
Rules: Use plain language, avoid speculation, keep it skimmable.
Output: 1 page: Summary, Findings, Recommendations, Risks, Next steps.

5) Customer replies (safer)
Goal: Draft a support reply.
Context: [sanitized issue + timeline].
Rules: Be empathetic. Don’t request sensitive info. Don’t cite policies you can’t quote.
Output: Reply + 3 clarifying questions + escalation criteria.

Follow-up A (clarify):
Ask up to 5 clarifying questions before you answer. Then provide the draft.

Follow-up B (quality control):
Point out anything that is vague, risky, or missing. Suggest safer wording.