The practice: After Claude produces an output, add: "Before I review this, read it as the [PROSPECT TITLE] receiving it and identify the single sentence most likely to make them stop reading."
Why it works: This forces Claude to evaluate output from the reader's perspective rather than the writer's. It surfaces the weakest part before you send it.
Effective self-review prompts:
• "Read this as a skeptical CFO who receives 50 cold emails a day. What would make them delete it?"
• "Identify the sentence that sounds most like it was written by AI and rewrite it"
• "Rate this email 1-10 for reply likelihood, then explain the rating"
• "What's the weakest line in this email and why?"
The efficiency gain: One self-review prompt catches the most obvious problems in 10 seconds. You can jump directly to fixing what Claude flags instead of reading the entire output looking for issues.