Skipping the Trial: An Empirical Investigation of Automation Bias in Robo-Advisor Adoption

The adoption of AI-powered robo-advisors represents a significant shift in wealth management, yet it raises concerns about the uncritical delegation of financial decisions. While traditional technology adoption models focus on static beliefs leading to adoption, they overlook the crucial behavioral process of initial trust calibration. This study introduces and investigates “trialing behavior”—the active, user-led evaluation of a robo-advisor before delegation—as a critical step that, when skipped, constitutes a behavioral manifestation of automation bias. Analyzing the transaction records of 1,862 investors on a leading financing platform, we first use fuzzy c-means clustering to identify that approximately 30% of investors skip any trialing. We then develop and test hypotheses that examine how key user, task, and system characteristics suppress rational evaluation. Our results reveal a nuanced picture: while general investment experience promotes trialing, this effect is attenuated by a history of negative outcomes (e.g., high default rates). Furthermore, larger investment amounts and longer robo-advisor operational tenure significantly suppress trialing. These findings indicate that in high-stakes decisions, cognitive overload may overwhelm accountability pressures, while system longevity appears to serve as a heuristic for trust, potentially fostering complacency. This research contributes to technology adoption literature by conceptualizing trialing as a key behavioral mechanism and enriches automation bias theory by resolving theoretical tensions and introducing novel antecedents. Our findings offer critical insights for designing more ethical robo-advisors and investor protection frameworks.