When readiness fails: a TOE-TAM mixed-methods analysis of AI-enabled HR digitalization in a resource-constrained...

This study examines barriers to adopting artificial intelligence (AI) in human resource management (HRM)-hereafter AI-enabled human resources (AI-HR)-across Zanzibar's public and private sectors and asks whether the standard predictors of an integrated Technology-Organization-Environment and Technology Acceptance Model (TOE-TAM) framework behave as theorized under mandatory, resource-constrained conditions. An explanatory sequential mixed-methods design combined a survey of HR professionals, information technology (IT) managers, and administrators (N = 137), analyzed with partial least squares structural equation modelling (PLS-SEM), with nine semi-structured key-informant interviews used to explain the quantitative results. The organizational readiness (OR) scale failed internal consistency (Cronbach's α = -0.105) and was therefore excluded before structural testing. Of the six structurally interpretable paths, perceived ease of use strongly predicted perceived usefulness (β = 0.552, p < 0.001), while technological readiness predicted AI-HR adoption readiness only marginally (β = 0.216, p = 0.044; bias-corrected and accelerated [BCa] confidence interval that included zero). Environmental readiness, perceived usefulness, and perceived ease of use did not predict adoption readiness, and adoption readiness did not predict HRM effectiveness. The interviews identified five mechanisms omitted from the reduced model: stalled, partial digitalization; infrastructure and vendor dependence as a binding constraint; capacity deficits and expertise flight; institutional decoupling between formal policy and operational practice; and a cultural preference for human judgement under directive adoption. The findings suggest that AI-HR adoption in resource-constrained small island contexts depends on foundational infrastructure, institutional coherence, and technology sovereignty, while also motivating a sequenced policy agenda and more context-sensitive measurement.