When do Managers Rely on AI? Cognitive Exhaustion and AI Advice in Innovation Selection Processes

Artificial intelligence (AI) is transforming decision-making in various organizational settings by supporting human decision-makers. However, although several AI-augmented decision-making applications occur in multi-decision settings, current literature treats AI reliance in those settings as purely influenced by single-decision level variables, fully neglecting factors specific to the multi-decision setting. Yet, understanding such factors is crucial, as reliance in single decisions might be contingent upon the overall cognitive experience with the task and the AI tool. In particular, the AI-augmented decision-making literature has overlooked two cognitive factors specific to multi-decision settings – i.e., cognitive load stemming from the task-level workload and cognitive dissonance from a task-level disagreement with AI in the multi-decision session. We develop a Multi-Decision Augmentation Theory (MDAT) of AI reliance in multi-decision settings and theorize that a higher task-level workload increases AI reliance, whereas experiencing higher task-level disagreement with AI during the multi-decision session leads to reduced AI reliance in single decisions. We test our theoretical framework in two multi-decision innovation selection contexts by conducting a lab experiment with 53 young business professionals and a field experiment with 88 managers in one firm and further contextualize our results with 20 interviews. We explore the implications of these findings for theories of AI reliance in augmented decision-making and organizational practice.