Generative AI use and advisory performance among agricultural extension agents in Benin

Harnessing Generative Artificial Intelligence (GenAI) offers promising avenues to enhance agricultural advisory services. Yet, understanding extension agents' engagement with such technologies remains limited. Using the Technology Acceptance Model (TAM) as an analytical lens, this study investigates how agricultural extension agents in Benin interact with GenAI and its impact on their advisory performance. We surveyed 240 extension agents across six districts and applied Partial Least Squares Structural Equation Modeling to examine relationships among perceived usefulness, workload, time pressure, attitudes, behavioral intentions, GenAI use, and performance. Results reveal that GenAI use is positively associated with improved advisory effectiveness. Workload and pressure emerge as key motivators for GenAI use, while perceived usefulness strongly predicts both positive attitudes toward GenAI and perceived ease of use. However, contrary to TAM assumptions, attitude has a negative influence on behavioral intention, a paradoxical engagement pattern implying that while extension agents value GenAI, they hesitate to rely fully on it, reflecting concerns about professional judgment, accountability, and trust in AI outputs. Finally, attitude, pressure, and behavioral intention indirectly affect the performance of extension agents using GenAI. This study contributes to agricultural extension research and AI governance debates by revealing how professional intermediaries navigate tensions between technological promise and institutional responsibility, offering insights for capacity-building and policy frameworks that promote responsible AI integration.