Akhmedov Shakhzod Uktam Ugli School of Business Administration

Abstract

In the age of advanced artificial intelligence, AI agents are increasingly taking over human agents in consumer service tasks, understanding how their design shapes consumer response is very critical. This study explores whether AI agent appearance (human-like versus machine-like) affects users' interaction in airline service contexts and examines how trust and fairness influence these effects under different service outcomes (success vs. failure). A 2 × 2 between-subjects online experiment (N = 240) manipulated AI agent appearance and service performance (success vs. failure) in a simulated mobile chat interface, with participants reporting trust, fairness and behavioural intentions. Results showed that service outcome was the strongest predictor of consumer responses, while AI agent appearance had no main effect in service success condition while being important if service fails. Under failure conditions, however, human-like AI agents were liked and recommended more than machine-like AI agents. Findings from this research suggest that in high-stakes situations, functional performance and perceptions of fairness take precedence over visual design, whereas human-like attributes may only mitigate negative reactions in an instance of service failure, providing practical insights for the design of effective AI service agents. Keywords: AI agent, anthropomorphism, trust, fairness, service condition, consumer response