Who builds this, and why

Every few months another headline announces a percentage of jobs that AI will eliminate. Almost none of these numbers can be checked, because almost none of them say what was measured, when it was measured, or what would prove them wrong.

This index takes the opposite approach. It scores the smallest unit of work there is, a single standardised work activity, against a published rubric, with a stated capability reference date, and it publishes its own inter-rater reliability. If a score is wrong you can point at exactly which of six ratings is wrong and say so.

The Task Exposure Index is published by A.I.T. Multiverse Consulting Ltd, which is accountable for the methodology, the ratings and every figure on this site.

The rating process is documented in full: a calibration set rated first, the remaining activities rated independently against the same written brief, and the agreement between raters measured and published rather than asserted. Where the index relies on judgment, and it does in five of its six dimensions, that judgment is written down, scored, and open to challenge activity by activity.

This site carries advertising and, on some pages, affiliate links to training and software products. Affiliate relationships have no influence on any score. Ratings are assigned against the rubric before any commercial consideration, the rating files are published, and no advertiser sees a score before it is public. The advertising and affiliate disclosure sets out exactly how this is kept apart.

No AI company, staffing firm, training provider or government body funds this index.

Corrections are issued as numbered errata against the release they affect. Previous releases are never silently edited. They stay published exactly as they were, because a quarterly series whose past can be rewritten is worthless.

Scores may be quoted and reproduced with attribution to the Task Exposure Index and the release identifier. Please cite the release, since the numbers change every quarter by design.