imo, a boss should be able to trust and/or control action--as well as be able to independently validate work--of their superstar staff. a) are ai systems being designed in a way that bosses can do that? b) train bosses (e.g. discard misinfo.) - practical w a wide range of users?