No Hard Feelings: The Protective Power of AI Empathy During Service Interaction Failures

Artificial intelligence (AI) agents such as voice assistants (VAs) are increasingly integrated into customer service, especially in high-stakes industries like banking. Yet, service failures remain common, harming customers’ experience and triggering misbehavior. This study investigates whether AI empathy improves customer satisfaction and reduces verbal aggression in AI-led interactions, examining the underlying mechanism and the role of service failure. We employed a 2 (AI Empathy: empathic vs. non-empathic) × 2 (Service-Interaction Failure: present vs. absent) between-subjects design in which participants interacted with a VA to resolve a service-related issue. Our results show AI empathy enhances customer satisfaction and reduces verbal aggression. A sequential psychological mechanism explains these effects: AI empathy boosts perceived mind in the AI agent, which in turn enhances its social presence. Notably, these effects are amplified when service failures occur, where AI empathy plays a protective role against their disruptive effect. Prosodic and textual sentiment analyses during service failures further showed greater satisfaction, relief, and positivity in empathic conditions. This work also introduces a methodological paradigm shift in studying human–AI interactions by using dynamic, prompt-based, publicly available AI voice technology that enables ecologically valid customer-service scenarios. These findings guide organizations in designing AI-empathy features for customer support.