The Difficult Conversation Drafter: Write Any Hard Professional Message Without the Dread

<p>There is a message you have been putting off writing for three days. You know what you need to say. You have rehearsed it in your head. But every time you open a draft, the words come out either too harsh or so cushioned that they have lost their meaning entirely.</p>

<p>This prompt fixes that. Give it the situation, the relationship, your goal, and any tone constraints — and it produces a message that opens with the actual point, provides necessary context without over-justifying, and ends with something concrete.</p>

<h2>The Prompt</h2>

<pre><code>You are helping me write a difficult professional message. Here is the situation I need to address:

SITUATION: [Describe what happened or what you need to communicate — be specific] RELATIONSHIP: [Your relationship to the recipient — colleague, manager, direct report, client, partner, friend] GOAL: [What you need to achieve — deliver critical feedback, set a boundary, share bad news, address conflict, make a request, apologize, end a working relationship] TONE CONSTRAINTS: [Any requirements — keep it brief, stay warm, be firm, preserve the relationship, or similar]

Write a clear, professional message that: 1. Opens by naming the situation directly — no preamble, no "I hope this email finds you well" 2. States the core point plainly in the first 2 sentences 3. Gives necessary context without over-explaining or justifying everything 4. Avoids passive voice that diffuses accountability 5. Ends with one concrete next step, ask, or expectation

Write at a register appropriate for the relationship described. Do not over-apologize or hedge the central message into invisibility.</code></pre>

<h2>Why Most Difficult Conversations Get Written Badly</h2>

<p>When something is uncomfortable to say, we instinctively do one of two things. We either write it bluntly without considering the relationship — which gets the point across but damages trust. Or we hedge so heavily that the recipient finishes reading and is not sure what was actually asked of them.</p>

<p>Both failure modes share a root cause: writing from anxiety rather than purpose. The anxiety about upsetting someone produces the hedging. The frustration that built up before you could say anything produces the bluntness. What you need is a draft that neither of those emotional states would write — something clear, direct, and calibrated to the specific relationship.</p>

<p>That is what this prompt does. By asking you to state the relationship and the goal upfront, it separates your intention from the emotional noise around it. The model then produces a draft from a neutral starting point, which you can adjust for tone before sending.</p>

<h2>What to Fill In</h2>

<p><strong>SITUATION</strong> — Be specific. "My colleague has been missing deadlines on the shared project" is better than "there's a performance issue." "I need to tell a long-term client we can no longer support their pricing tier" is better than "I need to deliver bad news." The more precisely you describe the situation, the more precisely the model can address it.</p>

<p><strong>RELATIONSHIP</strong> — This shapes register and formality. A message to a direct report reads differently than one to a client you have worked with for five years, which reads differently than one to someone you have never met. The model uses this to calibrate how direct vs. warm to be.</p>

<p><strong>GOAL</strong> — What specifically needs to happen after they read this? A behavior change, a decision, a meeting, an acknowledgment? Messages without a defined goal tend to end vaguely. Define the outcome.</p>

<p><strong>TONE CONSTRAINTS</strong> — If you need to preserve the relationship, say so. If brevity matters more than warmth, say that. If you need to be firm because previous soft messages were ignored, that context changes the draft.</p>

<h2>A Real Example</h2>

<p>Suppose you are a team lead who needs to tell a developer that their recent work quality has dropped and you want to understand why before it becomes a formal issue. You would fill in:</p>

<ul> <li><strong>SITUATION:</strong> A developer on my team has shipped three features in a row with significant bugs. Quality was reliable before. I do not know if something has changed personally, if they are overloaded, or if it is a process issue.</li> <li><strong>RELATIONSHIP:</strong> Direct report, good relationship, 18 months on the team.</li> <li><strong>GOAL:</strong> Set up a 1:1 to understand what is happening, without making them feel accused or defensive.</li> <li><strong>TONE CONSTRAINTS:</strong> Warm but honest. Name the problem clearly, but make it clear I am trying to understand, not discipline.</li> </ul>

<p>The resulting draft would open with the pattern directly ("I have noticed something in your recent work I would like to understand better"), name the issue without accusation, and close with a specific ask ("Can we grab 30 minutes this week?").</p>

<h2>What Model to Use</h2>

<p>Claude Opus 4.8 produces the most nuanced results — it handles tone calibration well and rarely produces drafts that are either passive-aggressive or robotically formal. Claude Sonnet 4.6 is faster and good enough for most situations. GPT-4o works well too, especially if you want a slightly more formal register by default.</p>

<p>For messages where the stakes are high — ending a business relationship, addressing a serious performance issue, communicating a major organizational change — use Opus and review the draft carefully before sending. For routine-but-uncomfortable messages (declining requests, giving corrective feedback on a small issue, setting meeting expectations), Sonnet is sufficient.</p>

<h2>One Rule</h2>

<p>Whatever the model produces: read the first two sentences out loud. If someone reading just those two sentences would not know what the message is about, the draft needs to be revised. The core point belongs in the opening, not the third paragraph.</p>