Impact of AI Adoption on the Sentiment of Customer Service Agents: Evidence from A Field Study

Abstract

We examine how AI affected the sentiment of customer service agents. Collaborating with a meal delivery company, we conducted a field study which exploited variation in AI adoption. We found that AI improved the self-reported sentiment of agents who worked in only the chat channel (especially those with shorter tenure and poorer performance). Yet, AI hurt the sentiment of agents who worked in both the chat and non-chat channels—a negative spillover as a result of AI shifting the reference point. Also, AI increased the likelihood of neutral-to-positive agent sentiment but reduced the likelihood of very positive sentiment, perhaps because AI prevented agents from feeling deeply fulfilled in their jobs. Our results support the use of AI for onboarding new hires and mitigating agent attrition. Managers should anticipate complex psychological consequences (alongside economic benefits) and be cautious of spillovers among employees with a mix of experiences with and without AI.