The AI demo still needs a person with judgment

What the source reports

Ethan Mollick writes about demonstrations he made with AI. His point, as captured in today’s Beyond Prompting digest, is that the demos still needed a person with subject knowledge. That person had to spot weak choices and reject them. A quick, polished result can hide the judgment behind the work. The digest suggests asking for two safe draft versions and writing down why one is worth keeping. This is a practical way to preserve the human choice, not proof that AI always improves the final work. Mollick’s examples are not a controlled study. The limit matters: a finished artifact alone cannot show whether the underlying claim is right or whether the steps can be checked.