Abstract

Rapid technological change has been reshaping firms’ skill demand, and the diffusion of generative artificial intelligence (genAI) is expected to accelerate this process. This paper examines how generative AI affects firms’ demand for apprenticeship training, which marks a key entry channel into employment for young workers in many occupations in Germany. Using German apprenticeship vacancy data from 2019 to 2024, we exploit the introduction of ChatGPT-3.5 in November 2022 as a common shock and estimate event-study models that leverage cross-occupational variation in AI exposure. We find that apprenticeship demand in highly AI-exposed occupations declines significantly compared to less exposed occupations, with vacancies falling by an average of 3.5\% after November 2022. This effect is driven by a reallocation toward less AI-exposed occupational groups rather than within occupational group changes and is concentrated among medium-sized establishments and cognitively intensive occupations. In contrast, overall full-time employment in highly AI-exposed occupations remains stable, while the share of apprenticeship-trained workers falls and the share of university-educated workers rises. These results suggest that genAI alters firms’ training and hiring strategies, with implications for the future composition of required skill levels.