Does AI do more harm than good? Assessing innovativeness and complaining intentions for successful and failed Mechanical and Feeling AI services

Shopping on amazon or booking holidays online are just two examples of AI-enabled services. Although of great practical importance, literature on AI-enabled services is scarce. Especially, no study exists that assesses the effect of the different AI levels of a service on its innovativeness. Moreover, empirical evidence on customer complaining behavior as a reaction to failed AI services is also missing. Using an online experiment (n=437), our paper strives to close this research gap. Our results show that customers will perceive a Feeling AI service (high degree of AI) as more innovative than a Mechanical AI service (low degree of AI). Moreover, customers using a failed Feeling AI service will complain more than customers using a failed Mechanical AI service. Finally, customers getting an AI Service Recovery will complain less than customers getting a Human Recovery. Our results highlight the importance of AI-level when creating and managing AI services