This paper examines worker preferences for AI-enabled changes in task composition and possibly facilitating job characteristics in a discrete choice experiment among workers in a multinational manufacturing firm. The firm implements an AI tool transforming workers’ core task of classifying electronic components in two successive stages: (1) augmenting AI including a decision support feature and (2) autonomous AI automating easy classification tasks. Workers have to choose between jobs that differ in five characteristics: the AI-enabled change in their classification task, breaks, autonomy, training, and a salary increase. The results show that workers prefer the changes induced by augmenting AI over autonomous AI. Workers would forgo a moderate salary increase when their core task is supported by the augmenting AI. Moreover, the perceived usefulness of the augmenting AI negatively moderates the preference for autonomous AI. These results suggest that human-AI complementarity is favorable while simultaneously posing disadvantages for stepwise implementation strategies.