Government is measuring how fast its AI works, not whether it actually worked

Speed is not the same as a correct answer, and most agencies still cannot show, end to end, that a given case was decided correctly