Designing AI to elicit positive word-of-mouth in service recovery: The role of stress, anthropomorphism, and personal resources
Byron Keating, Rory Mulcahy, Aimee Riedel, Amanda Beatson, Kate Letheren
Abstract
Service organizations are increasingly deploying generative AI (GenAI) chatbots to handle service failures, yet there is a critical gap in understanding how anthropomorphic AI design can improve service recovery outcomes. This study addresses that gap by investigating whether making AI agents more human-like can mitigate customers’ stress during service recovery and foster positive word-of-mouth (PWOM). Grounded in the Transactional Model of Stress and Coping , we propose that anthropomorphic cues in AI interactions reduce customers’ stress appraisals of service failures. A multi-study experimental design was employed, including a pilot study and three scenario-based experiments that manipulated AI anthropomorphism and service failure severity. The results show that anthropomorphized AI significantly lowers customer stress levels and, in turn, increases PWOM, with stress appraisals mediating the relationship between AI anthropomorphism and positive word-of-mouth. Notably, these benefits emerged mainly for low-severity service failures, and the stress-reduction effect of an anthropomorphic AI agent was most pronounced for customers with limited personal coping resources. These findings provide actionable insights for service managers and AI designers: incorporating human-like warmth and competence into AI service agents can enhance recovery experiences by alleviating customer stress, thereby encouraging PWOM and improving overall service recovery effectiveness.