The responsible worker in the age of generative AI: How human-AI co-creation shapes experienced responsibility for work outcomes

The use of generative AI in organizations becomes most apparent when it produces errors. Implementing such systems leads to structural realignments of agency between AI and humans, which, if poorly designed, can result in human workers disengaging from responsibility for work outcomes. This study investigates how collaboration with generative AI systems should be designed to foster human responsibility and reduce errors in AI-augmented work. Adopting a co-creation perspective and drawing on role theory, the research examines how different co-creation designs—human-first versus AI-first and high- versus low-performing AI—shape workers' perceptions of responsibility and their interactions with the generative AI system. An online lab experiment reveals two competing effects: while a human-first approach enhances workers' responsibility, it simultaneously reduces the intensity of co-creation participation when engaging with generative AI; conversely, a high-performing AI diffuses responsibility yet encourages more intense interactions. Building on these findings, the study also provides empirical evidence for the downstream consequences of responsibility, showing that workers' experienced responsibility mitigates errors in work outcomes. These insights advance theory on human-AI co-creation and offer practical design implications for fostering effective and responsible collaboration in AI-augmented work settings.