If you're deploying LLM-based decision making that affects lives, you should be the one held responsible for the results. If you don't want to do due diligence on automation, you can screen manually instead.