Where people in the job disagree with the index

The index rates what current AI systems can produce. Under every task on every job page, people who do that job can answer a simpler question: does AI already do this in your job? This page lists the tasks where their answers and our ratings are furthest apart.

The two measure different things, so a gap is not an error on either side. Where people report more AI use than the rating, adoption may be running ahead of what the rubric credits, or the rating may be too cautious. Where they report less, the work may be possible but not yet adopted, which is exactly the difference between exposure and displacement. We read this page before every quarterly re-rating.

A task appears once it has 10 answers. “People say AI does it” counts “already” as one and “partly” as a half. Answers are self reported and not a sample of anyone in particular.