This study utilises a novel big data set based on online job advertisements – Cedefop’s Skills OVATE – with information on the skills and work activities required by EU employers. The data provide insight into the task profiles of detailed occupations faced with higher automation risk or those relying on alternative digital technologies (robots, computer software, AI). The paper explores suitable machine and deep learning models to test how well a parsimonious set of task indicators can predict occupational automatability. Work activities associated with greater occupational automation risk and robot exposure (e.g. inspecting equipment, performing physical activities), typically concentrated in routine or manual jobs, differ from those prominent in occupations with higher AI exposure (e.g. thinking creatively, evaluating standards).

Creative Commons Attribution 4.0 International (CC BY 4.0)