Where AI Exposure Matters: Occupational Structure, Internal Control, and Corporate ESG Performance

While generative artificial intelligence (AI) reshapes job tasks across corporate functions, its organizational consequences for firm sustainability remain insufficiently understood. This study explores how generative-AI susceptibility concentrated in governance-critical occupations relates to corporate ESG (environmental, social and governance) performance. Based on panel data of Chinese A-share listed firms spanning 2010–2024, we construct GovernanceAI (governance-related occupational AI exposure), an occupation-level indicator that reflects the potential susceptibility of generative AI tasks in finance and administrative occupations, but not necessarily the actual use of generative AI in companies. Regression results show that higher GovernanceAI is negatively associated with corporate ESG performance even when conditional on overall workforce-wide occupational AI exposure, with the strongest signal for finance-function positions. Further evidence suggests internal-control quality acts as a plausible mediating channel. This relationship is more pronounced among non-state-owned enterprises and firms with severe financing constraints. Although we employ a Bartik-style instrumental variable for endogeneity diagnosis, IV estimates are specification-sensitive and serve only as supplementary evidence. We conclude that the organizational implications of AI depend not merely on aggregate AI susceptibility but critically on the functional location of AI-driven task shifts within firms.