Research

We analyze a hiring interview in which the firm chooses the quality of an AI that drafts each answer, and candidates decide how carefully to verify it. Because a correct draft passes whoever reviews it, AI quality determines how often candidates must catch errors, and hence how visible supervision skill is. Informativeness about supervision is zero without AI and, once AI is present, strictly declines as AI quality rises; informativeness about AI-operation skill peaks at an intermediate quality. In a minimal hiring problem, weaker interview AI yields strictly more productive hires whenever the role weights supervision enough.