I used to toss a vague idea into an AI and hope for genius. It spat back polite oatmeal.
Hard truth: AI amplifies your thinking. If your idea is fuzzy, the output will be politely fuzzy.
Here’s what actually works for me.
Step 0: Write the messy sentence.
– One line only. No polishing. Example: “Most meetings are status theater.”
Step 1: Give the idea a spine.
– Point, Proof, Payoff.
– What I believe, why it matters, what someone can try.
Step 2: Use AI like a ruthless assistant, not a ghostwriter.
– Prompts I paste:
– “List 5 angles on: ‘Most meetings are status theater.’ Keep it practical.”
– “Give me 3 counterarguments.”
– “Draft a 150–200 word outline in my tone: candid, short paragraphs, plain language, no buzzwords, no emojis.”
Step 3: Put myself back in.
– One personal moment. One opinion that could be wrong. This is what makes it human.
Step 4: De-robotify.
– Find-and-replace these: leverage→use, utilize→use, synergy→fit, paradigm→approach.
– Read it out loud. If you feel like a LinkedIn brochure, cut more.
Step 5: Add one fact.
– A stat, a quote, or a tiny example. Then link it or name the source.
Step 6: Land it.
– One sentence takeaway. No TED Talk ending. Just the point.
Mini example:
– Seed: “Most meetings are status theater.”
– Spine: Point (we perform, don’t decide), Proof (calendar data or your Tuesday), Payoff (try a “no agenda, no meeting” rule for a week).
Tools that help:
– Perplexity or Elicit for quick sources.
– Hemingway to remove fluff.
– Voice Memos + Whisper to brain-dump ideas.
– A saved prompt in Notes so I don’t reinvent it each time.
Uncomfortable note: If you won’t spend 5 minutes clarifying your spine, AI can’t save the post. It’ll just format your confusion in crisp paragraphs.
My quiet rule: write like I talk, then let AI tighten the bolts. Not the other way around.
