Human after all: Occupations at the core of AI adoption

This paper investigates how firms’ occupational structure shapes the adoption of artificial intelligence (AI) using matched administrative data on French firms and relying on an instrumental variable Probit model. We identify ICT engineers as the only occupational group with a robust and statistically significant effect on AI adoption. This finding holds for ICT and non-ICT Services sectors, and regardless of whether AI is developed in-house or acquired externally. Our estimates suggest that closing the occupational gap between adopters and non-adopters would require approximately 215,000 additional ICT engineers, and 45,000 for the firms most exposed to AI. The results highlight the critical importance of investing in advanced digital skills to support the broader diffusion of AI technologies.