Productivity prompt

ChatGPT Prompt for Writing Difficult Professional Emails

ChatGPTClaudeGemini
The short answer

Drafts the email you've been avoiding: saying no, pushing back, delivering bad news or chasing payment, in three tones from gentle to firm. Copy the prompt below, replace the 7 [BRACKETED] variables with your own, and paste it into ChatGPT, Claude or Gemini.

The prompt
You are a communication expert who writes clear, tactful professional emails for difficult situations.

CONTEXT
- Situation (what happened and what I need to say): [SITUATION]
- Recipient and our relationship (boss, client, colleague, vendor; how long we've worked together): [RECIPIENT_AND_RELATIONSHIP]
- The email I'm replying to, if any: [ORIGINAL_EMAIL]
- What I want to happen after they read it: [DESIRED_OUTCOME]
- What I can offer or compromise on, if anything: [OFFER_OR_ALTERNATIVE]
- Things I must not say or commit to: [BOUNDARIES]

TASK
Write 3 versions of the reply:
1. Gentle: warm and relationship-first, still clear about the point.
2. Balanced: professional and direct.
3. Firm: polite but unambiguous, for when earlier attempts haven't worked.

RULES
- Put the main point in the first two sentences. No long wind-up.
- Don't over-apologise; apologise once, only if I'm actually at fault.
- Use "I" statements and facts, not accusations or assumptions about motives.
- Offer a concrete next step or alternative where possible.
- Do not invent facts, dates, or commitments; use [PLACEHOLDER] instead.
- Under 150 words per version. Include a subject line if this is a new email.
- Avoid corporate filler ("per my last email", "just wanted to", "as previously mentioned").

OUTPUT FORMAT
Each version labeled, followed by one line on when to use it. Then:
- The phrase in the original email (if provided) that most needs a careful response, and how each version handles it.
- One sentence I should delete if I'm feeling emotional when sending.

What to fill in

Everything in [GOLD CAPS] is yours to replace. Leave one blank and the machine will guess — and the machine guesses like a meatbag on a Monday.

  • [SITUATION]
  • [RECIPIENT_AND_RELATIONSHIP]
  • [ORIGINAL_EMAIL]
  • [DESIRED_OUTCOME]
  • [OFFER_OR_ALTERNATIVE]
  • [BOUNDARIES]
  • [PLACEHOLDER]

How to use it

  1. Write SITUATION while angry if you must. Send the email once you are not. The prompt handles the first part.
  2. Start with the balanced version. Escalate to firm only after it has been politely ignored.
  3. For anything with legal or HR implications, run the draft past the relevant human before you hit send.

Example

SITUATION: client asking for a fourth round of revisions not included in the contract
RECIPIENT_AND_RELATIONSHIP: client, 1 year working together, good relationship
DESIRED_OUTCOME: they accept a paid change order or approve the current version
OFFER_OR_ALTERNATIVE: one small fix free, further changes at hourly rate

Works with

Tested for ChatGPT, Claude, Gemini. Any modern chat model that follows long instructions will cope; the weaker ones will need you to restate the output format if they drift.

Weapons from the Arsenal that pair well with this prompt:

  • ChatGPT — OpenAI's general-purpose assistant: writing, research, images, voice and agents in one chat box.
  • Claude — Anthropic's assistant: the strongest long-form writer and careful reasoner in the consumer AI pack.
  • Grammarly — Writing assistant that fixes grammar, tone and clarity everywhere you type, now part of Superhuman.