Stakeholder engagement with AI service interactions
研究将具身对话代理视为利益相关者，通过代理概念区分其伙伴与仆人角色，发现伙伴角色结合人类外观和情感对话风格最能提升客户感知价值、使用意愿和支付意愿，并揭示过度依赖AI的风险。
Abstract Recent advancements in artificial intelligence (AI) have ushered in a wave of AI innovations in the form of embodied conversational agents. These stakeholders offer new ways to engage customers in the co‐creation of services but still face significant customer skepticism. To address this challenge, we frame interactions between customers and embodied conversational agents through the lens of stakeholder engagement and apply the concept of proxy agency from social cognitive theory. This framework allows us to identify two primary stakeholder roles for embodied conversational agents: partner and servant. We conceptualize how these roles inform optimal design for embodied conversational agents and shape a two‐stage value‐by‐proxy process, comprising proxy efficacy and outcome expectancy. Additionally, we uncover tensions within this process due to over‐reliance on AI, as well as significant outcomes that extend beyond the immediate interaction. Our study, using a custom‐developed embodied conversational agent with a sample of 596 U.S.‐based respondents, reveals that positioning an embodied conversational agent in a partner role, combined with a human (vs. robot) appearance and emotional (vs. functional) conversation style, has the strongest positive impact on perceived value‐by‐proxy, usage and advice implementation intentions, and willingness to pay. We also observe an inverted U‐shaped moderation by reliance in the relationship between proxy efficacy and outcome expectancy, signaling the potential risks of over‐reliance on AI. Furthermore, we provide qualitative insights into why some customers avoid engaging with embodied conversational agents. Overall, we offer a nuanced perspective on embodied conversational agents as active stakeholders within organizational systems, advancing both theoretical understanding and practical applications of this rapidly evolving technology.