An AI exposure estimate is not a jobs forecast

Studies of artificial-intelligence “exposure” often map system capabilities to tasks listed inside occupations. A high score may mean many tasks could change. It does not say whether firms will adopt the tool, whether law permits it, whether customers accept it, or whether saved time becomes fewer workers, more output or different duties.

Headlines frequently turn that conditional map into a countdown to job loss. The correction cuts both ways: exposure is not proof of unemployment, but neither is it proof that workers benefit. A plausible future includes jobs redesigned around new tools; another concentrates bargaining power and removes entry-level tasks. Institutions, prices and workplace choices decide between them, so dated scenario ranges are more honest than a single number.