Turning AI Capability into Performance: How AI Understanding and AI Skills Shape Employee Productivity through Employee–AI Collaboration in a Chinese Smart Hospital

Abstract

Hospital investment in artificial intelligence (AI) rarely produces productivity gains by itself. Drawing on the technology acceptance model and person–environment fit theory, this study asks whether two capability-oriented employee conditions, understanding of AI and practical AI skills, improve employee productivity directly, or whether their value depends on being converted into day-to-day collaboration with AI systems. Survey data were collected from 694 employees of a tertiary public hospital in China. The measurement model was validated in SmartPLS 4, and the structural hypotheses were tested with path analysis on composite scores, using Monte Carlo confidence intervals for the indirect effects. AI understanding (β = 0.434) and AI skills (β = 0.312) both predicted employee–AI collaboration, and collaboration was the strongest predictor of productivity (β = 0.426, p < .001). The direct capability–productivity paths were positive but comparatively small (understanding: β = 0.169, p < .001; skills: β = 0.073, p = .041), while the indirect paths through collaboration were significant in each case and carried 52% and 64% of the respective total effects. The pattern is consistent with complementary partial mediation: knowing what AI can do, and being able to operate it, pay off chiefly when they are enacted in collaborative work practice. The findings move the explanatory weight from acceptance attitudes to enacted collaboration, and they imply that hospital training and system design should be evaluated by whether they change how employees work with AI, not merely what employees know about it.