Four named failure modes.
Each one maps to a specific broken ingredient in the Professional Recipe. Diagnose the ingredient, not the output. Fix the recipe, not the plate.
The Generic Output Trap
Missing — Examples (#4)The output is clean but lifeless. Reads like stock copy. Doesn't sound like your business.
The station was never shown the house standard. It’s doing its best — from nothing.
Add three to five real work samples to the recipe. Watch the next plate sharpen in one sitting.
The Process Cage
Missing — Output Over Process (#5)You gave a sharp agent a paint-by-numbers kit, and it stopped using its judgment.
You over-specified the steps instead of the destination. The agent is following instructions, not cooking.
Describe what “plated well” looks like. Let the station figure out the path.
The Open Door Problem
Missing — Guardrails (#3)The station sent out a wrong price. Or made a promise you can’t keep. Or said something you’d never say.
You never told the station what it cannot do. Nothing stopped it.
Write the safety stops into the recipe. What’s off-menu? What needs the Chef’s eyes? Hard-code it.
The Silent Critic
Broken — Feedback Loop (#7)Your team is quietly rewriting bad output every day. They’ve given up on reporting it.
The Feedback Loop is broken. The recipe never gets updated — just the individual plates.
Build a thirty-second path from “miss” to recipe update. Fix it once, fix it everywhere.
The cultural shift.
Don't rewrite the output. Update the recipe. This is the hardest mindset change for a team that's been quietly fixing AI output all year. The thirty seconds you spend updating the prompt saves thirty minutes a week forever. Fix it once. Fix it everywhere.
