AI readiness is often treated as a technical skill, as though one could study for it, take a test, and be done. It is a pitfall particularly in education – what does it even mean to teach residents to be ready for the AI age?
Researchers at UT Austin asked 523 early-career professionals to do client-like work with an AI agent, then compared their results with an AI-only baseline. About half were “AI Amplifiers”: they did better than AI alone. A quarter were “Delegators”: their work was about as good as the AI’s. The rest, “Apprentices,” did worse, even though they were not necessarily less capable.
Yes – about a quarter of the early-career professional appeared to produce worse outputs when using AI compared to than just having it done by AI alone. (The reality is probably that the researchers were better prompt engineers than the apprentices).
Strong users framed the problem, used real domain knowledge, checked the output, and refined it over several rounds. AI was used as a collaborator that needed direction and oversight.
This feels familiar in resident education. An algorithm can produce a useful first pass. Safe clinical value still depends on whether the clinician asks the right question, spots what does not fit, and knows what to do next.
The lesson is not to teach longer prompts. It is to teach judgment in the workflow: frame, check, refine, decide.
AI can raise the floor. People still determine the ceiling.
Source: McCombs School of Business, University of Texas at Austin.