Nothing can prevent human error. But systems can be set up in such a way that limits the risk of that human error causing failure. If the agent has to actually

Nothing can prevent human error. But systems can be set up in such a way that limits the risk of that human error causing failure. If the agent has to actually enter the answers before being able to issue a temporary password, then (assuming the author is correct) this particular problem wouldn't have happened.