The Impact of Artificial Intelligence Adoption on Employee Performance: The Mediating Role of Digital Skills

Abstract

Organisations are adopting artificial intelligence (AI) rapidly, yet evidence on whether and how adoption improves individual employee performance remains fragmented. Drawing on the technology acceptance tradition, task–technology fit, human capital theory and the job demands–resources model, this study develops and tests a model in which digital skills mediate the relationship between AI adoption and employee performance. The study follows a transparent, simulation-based design. A measurement instrument (17 reflective items) and a population model with a priori specified effects were defined. Survey-type data for 350 employees were generated by Monte Carlo simulation and analysed with partial least squares structural equation modelling (PLS-SEM), consistent PLS (PLSc) and 5,000-sample bootstrapping. Three further Monte Carlo experiments (1,000 to 8,400 replications) examined parameter recovery, statistical power for the indirect effect, and the sensitivity of estimates to common method variance. The measurement model met all conventional criteria (α = .85–.88, composite reliability = .89–.91, AVE = .58–.67, HTMT ≤ .46). AI adoption was positively related to digital skills (β = .412, p < .001) and digital skills to employee performance (β = .321, p < .001). The indirect effect was significant (β = .132, 95 % CI [.088, .191]) and accounted for 51 % of the total effect, indicating complementary partial mediation. In the Monte Carlo experiments, standard PLS-SEM under-estimated the a and b paths by 12–13 % and the indirect effect by 23 %, whereas PLSc was virtually unbiased. Detecting a small indirect effect with 80 % power required more than 500 respondents. A method factor with loadings of 0.30 inflated the indirect effect by 31 %, yet the first factor still explained only 45 % of the variance on average, so Harman's single-factor test would not have flagged the bias. The study clarifies the mechanism through which AI adoption may translate into performance, namely the development of employees' digital skills. It also provides validated design guidance, including estimator choice, sample size and method-bias control, for empirical research on AI in the workplace. Because the data are simulated, the estimates illustrate the model's implications and the behaviour of the analytical procedure; they are not empirical evidence. The model must be confirmed with field data.