Shumiao Ouyang

with Yue Guo, Siliang Tong, Tao Chen, and Subodha Kumar

​​

Information Systems Research, R&R

​

Abstract: As the demand for personalized investment advice grows, more fintech platforms aim to develop conversational robo-advisors based on generative AI chatbots (GAICs) to surpass the pre-written script limitations of traditional AI chatbots. However, developing GAICs is time-consuming and costly, with little evidence of their substantial business value as financial advisors. This issue has ignited a debate about the ability of GAICs to provide human-like responses, leading more institutions to question their effectiveness in developing anthropomorphic AI agents, a key concern in information systems (IS) practice and literature. Until now, the business value of GAICs as financial advisors has been underexplored in the literature. To fill this gap, we collaborated with Ant Fortune, China’s leading fintech platform, to study the effects of Zhi Xiaobao, their newly launched GAIC. First, contrary to the belief that AI fintech platforms mainly help novices by broadening access to financial advice, our findings reveal that experienced investors, not novices, derive greater value from GAICs. This counterintuitive result highlights the nuanced reality that people with higher financial literacy and practical experience are more adept at understanding AI-generated content (AIGC), contributing a novel insight to IS literature on user engagement with interactive AI. Next, despite Zhi Xiaobao's superiority in several dimensions, we find it interesting that the traditional emphasis on quantitative metrics, such as content length, typically applied to user-generated content (UGC), proves inadequate to evaluate AIGC. This discrepancy highlights the risk of cognitive overload with AIGC and suggests the need to recalibrate quality assessment criteria. Such recalibration would ensure accurate content quality evaluation with prevailing AIGC, which is vital for IS research. Lastly, challenging the prevailing assumption that AI universally improves investment decisions, we discovered that GAICs particularly benefit risk-averse individuals by mitigating precipitate investment behaviors and promoting judicious and assertive actions. This finding diversifies the understanding of the impact of AI on investment behavior in the IS literature and highlights the ability of GAICs to provide personalized advice. Our research informs policymakers on the use of GAICs, underscoring the need for financial education, thus bridging IS practice with literature to optimize the strategic application of GAICs.