Will Employee–AI Collaboration Enhance Employees’ Proactive Behavior? A Study Based on the Conservation of Resources Theory
Citation. Sun, C., Zhao, X., Guo, B., & Chen, N. (2025). Will employee–AI collaboration enhance employees’ proactive behavior? A study based on the conservation of resources theory. Behavioral Sciences, 15(5), 648. https://doi.org/10.3390/bs15050648
Summary. In a time-lagged survey of employees in Chinese organizations using AI, employee–AI collaboration was associated with greater proactive behavior and lower workload. The authors model workload as a mediator and AI literacy as a moderator: literacy strengthened the collaboration–workload-relief association, while the reported indirect path through workload was statistically significant only in the lower-literacy group. The study frames workload relief as a resource pathway, not evidence that AI collaboration uniformly produces proactivity or reduces work demands in other settings.
Key claims
- In the 412-response, two-wave survey spanning 30 Chinese companies, employee–AI collaboration positively predicted proactive behavior (B = 0.32, p < 0.001), negatively predicted workload (B = −0.22, p < 0.001), and workload negatively predicted proactive behavior (B = −0.28, p < 0.001).
- The reported indirect effect of employee–AI collaboration on proactive behavior through workload was 0.25 (95% CI [0.12, 0.32]) using 5,000 bootstrap resamples.
- AI literacy moderated the association between employee–AI collaboration and workload (interaction B = −0.16, p < 0.01), with stronger workload relief among higher-literacy respondents.
- The conditional indirect effect through workload was significant at lower AI literacy (0.22, 95% CI [0.12, 0.31]) but not higher AI literacy (0.01, 95% CI [−0.06, 0.32]); the paper treats the latter pattern as heterogeneous rather than as evidence of a uniformly stronger proactivity benefit for high-literacy employees.
Evidence & limitations
- Evidence comes from self-report surveys of employees who used AI across manufacturing, finance, retail, and internet-technology organizations in China; a two-week separation between predictor and outcome waves reduces but does not eliminate common-method and causal-inference concerns.
- The design is observational, the measures capture broad perceived employee–AI collaboration rather than named systems or task arrangements, and the data are not publicly available because of privacy or ethical restrictions.
- The article is a peer-reviewed, open-access MDPI journal article under CC BY. Its publisher DOI and metadata were verified through Crossref; the ingested Drive PDF is 16 pages and carries the source-file hash in frontmatter.