Risk it, if you dare: A quantitative analysis of the effects of perceived risk facets on the adoption of generative artificial intelligence in knowledge work

The adoption of generative artificial intelligence (GenAI) is influenced by perceived risks, which can impact perceptions of performance, effort, and emotions. Understanding how different facets of risk drive functional and non-functional perceptions and lead to actual use would be crucial to improving its acceptance and adoption in knowledge work. By assessing performance, privacy, psychological, social, and time risk, and incorporating them into the artificial intelligence device use acceptance (AIDUA) model as part of a sequential adoption process, this study highlights the shaping effect of GenAI on risk perceptions and its broader influence on socio-technological interactions. Using structural equation modeling with three data samples from different countries ( n = 1,450), we identify a negative effect of performance risk and a positive effect of time risk on perceived performance. Performance, time, and psychological risk increase effort perceptions, underscoring the multi-faceted dynamics of effort formation. Performance and psychological risk directly and negatively influence emotions, emphasizing the need to consider non-functional antecedents in adoption. In contrast, privacy and social risk lack salience for perceived performance, perceived effort, and emotions. While not statistically significant, the high awareness of privacy risk highlights the relevance of privacy-calculus discussions in the context of responsible GenAI integration within corporations and society. Moreover, the time-risk-performance paradox opens avenues for human-AI collaboration, whereby learning and augmentation may support adoption. Thus, by specifying the types and impacts of risk, this study offers guidance for addressing risk-related socio-technological interactions and calls for risk mitigation endeavors in knowledge work. • Risk facets shape GenAI adoption via perceived performance, effort, and emotions. • Performance risk reduces perceived performance and emotions, yet increases effort. • Psychological risk increases effort perceptions and negatively influences emotions. • Time risk boosts perceived performance, despite adding to perceived effort. • High privacy risk awareness contrasts with its limited effect on GenAI adoption.