Dual effects of AI on academic work performance: A job demands-resources theory systematic review

BackgroundArtificial intelligence (AI) is rapidly reshaping academic work, yet its occupational health implications for academicians remain under-examined. The Job Demands-Resources (JD-R) framework offers a dual lens for understanding AI as both a performance-enhancing resource and a source of job demands, but no systematic review has applied this framework across the full spectrum of academic roles.ObjectiveThis systematic review examines how AI impacts academicians' performance across four core domains-teaching, research, community engagement, and professional development-using the JD-R framework to understand AI's dual nature as both a resource and a demand.MethodsFollowing PRISMA guidelines and the five-stage framework of Khan et al.,¹ we systematically searched Scopus and Web of Science databases for empirical studies published between 2021 and 2025. After screening 868 records, 30 peer-reviewed journal articles met the inclusion criteria. A two-phase coding approach (deductive and inductive) was employed, with intercoder reliability (Cohen's Kappa = 0.91) indicating strong agreement. Thematic analysis was guided by the JD-R framework.ResultsTeaching performance dominates the literature (70%), reflecting AI's immediate impact on instructional practices. AI functions as a job resource through enhanced personalized learning, assessment automation, and administrative efficiency when supported by self-efficacy, AI literacy, and institutional resources. However, AI creates significant job demands including academic integrity management, assessment redesign, continuous upskilling, and ethical dilemmas. Professional development (27%) exhibits the most severe demand-resource imbalance, with faculty experiencing skill obsolescence anxiety, technostress, and imposter syndrome. Research performance (20%) shows a relatively balanced profile, while community engagement (7%) remains critically understudied. Identity-threatening demands-academic integrity in teaching, authorship ambiguity in research, and skill obsolescence in professional development-generate disproportionate psychological strain requiring specialized buffers.ConclusionsAI's impact on academic work is inherently dualistic, with outcomes determined by the balance between available resources and demand intensity. Psychological resources, particularly self-efficacy, emerge as the most powerful enablers mediating the effectiveness of institutional and knowledge resources. The review extends JD-R theory by identifying identity-threatening demands as a distinct demand category and proposing a Resource Cascade Model wherein distal organizational resources foster proximal psychological resources. Institutions must adopt systemic, autonomy-supportive interventions addressing the full spectrum of AI's impact on academic work.