Not necessarily. If yor process accounts for "all models make mistakes", you are probably limiting how autonomous the model can be, and then there's not much tangible difference between models beyond that threshold.