Interference Timing of GenAI Sales Agents in Virtual Reality

Abstract This study investigates how interference timing of artificial sales agents influences user experience in embodied virtual commerce environments. Specifically, we examine how different timing strategies affect consumer perceptions for generative artificial intelligence (GenAI)-driven agent interference in a virtual reality (VR) shopping scenario. Using a controlled experiment with 100 participants in a VR showroom for 3D printers, we compared two interference conditions: the agent either enters the room and offers assistance immediately or delayed after the consumer has examined all products once. The sales agent, implemented through an integrated artificial intelligence (AI) pipeline combining speech-to-text, a large language model, and text-to-speech capabilities, assisted participants in their decision-making process. Employing a Bayesian research methodology, our findings reveal that the timing of initial agent interference credibly influences user experience, primarily mediated through the consumer’s first impression of the agent’s warmth. The immediate interference approach demonstrated superior outcomes. These results provide valuable insights for the design of AI-driven sales environments, demonstrating both the technical feasibility of implementing multistage AI pipelines in virtual commerce and the importance of carefully considered interaction timing. Our findings contribute to the body of knowledge on digital service system transformation in commercial settings and offer practical guidelines for developing more effective virtual shopping experiences.