KRIVET

This approach goes beyond a binary judgement of whether an occupation can be automated. Its advantage is that it can transparently quantify in which specific domains of a job AI has reached human level and where a gap still remains. Owing to this advantage, the measure can be used not only to measure AI exposure but also for various forms of in-depth analysis. For example, it enables job redesign analysis, applying explicit technology development scenarios to project how particular tasks and occupations may change. From a macroeconomic perspective, it can serve as a key input for modelling productivity gains and structural change across industries resulting from AI adoption. It also provides essential policy evidence for designing curriculum reform and worker retraining by identifying the areas in which distinctly human abilities are becoming more important.

The following chapters examine step by step how this measure was constructed. Chapter II reviews the process by which experts assessed the ability requirements of a small sample of occupations to establish a reference standard, followed by the procedure for extending this to all occupations using LLMs. Chapter III presents the resulting mapping as an analysis of AI exposure by occupation and capability domain. Chapter IV applies future AI capability development scenarios to examine how exposure may change. Finally, Chapter V summarises the findings and draws implications.