For complex tasks — deal analysis, win/loss post-mortems, objection strategy — Claude produces better outputs when it thinks through the problem before giving you the answer.
Add one of these phrases to your prompt before the task instruction:
When Claude reasons out loud, you can see its assumptions and correct them before they infect the final output. If Claude assumes your deal stalled because of budget, but you know it stalled because the champion went quiet, you can intervene at the reasoning stage and get a much more targeted answer.
Especially valuable for: deal diagnosis, MEDDIC gap analysis, competitive positioning decisions, and any time you're asking Claude to make a judgment call about a complex situation.
For simple writing tasks — cold email drafts, LinkedIn messages, follow-ups — asking Claude to think out loud before writing just adds steps and length. Use it for analysis, not for writing.