Building a system that ingests agencies' historical Meta/Google Ads campaigns, tags them with psychological DNA (hook types, angles, personas), then predicts if new campaigns will hit targets before spending.

Tech: n8n + Claude API + pattern matching logic. Learns ONLY from agency's own data (not market benchmarks).

Output: "78% success probability. Risk: Intro too long. Fix: Hard CTA (+54% CVR based on Campaign X)."

Questions for agency owners: 1. Real problem or solving what doesn't matter? 2. Would you pay $297-497/mo? 3. Creative fatigue kills this? (Do 6-month-old patterns still work?)

Not selling - validating