The generative AI Trust Loop: mechanisms of human-AI collaboration and organizational learning in Taiwanese SMEs

Generative artificial intelligence (GenAI) is now embedded in many organisational decisions, yet firms vary widely in how they govern trust and adapt after adoption. We address this puzzle by developing the Generative AI Trust Loop – a conceptual model that extends the Technology–Organization–Environment (TOE) framework with Dynamic Capabilities Theory. The recursive feedback at the core of the loop is offered as a theoretical proposition; cross-sectional PLS-SEM can test the structural pieces, not the full cycle. Our evidence draws on two strands: survey data from 400 Taiwanese SMEs analyzed through PLS-SEM, and six in-depth case studies that probe the mechanisms behind the coefficients. Across both, explainability, AI reliability, and human supervision relate jointly to trust calibration, which mediates their links to organisational learning. Digital culture, treated as an organisational moderator, strengthens that calibration–learning link. The broader argument is conceptual. When trust calibration is read as an ongoing governance process rather than a one-time antecedent, it becomes a plausible micro-foundation for the meta-learning capability that lets SMEs co-evolve with GenAI rather than merely adopt it. We close with practical human-in-the-loop governance guidance for SMEs, while leaving the cyclical learning logic for longitudinal work.