Design of Task Allocation and Decision-Making Styles for AI Agents Based on LLM

This study introduces an innovative framework that aligns LLM-based AI agents with enterprise needs by integrating dual-process theory and resource-based task complexity models. Utilizing large language models and Retrieval-Augmented Generation, our system simulates diverse task and decisionmaking scenarios through a Strategy Generator and Evaluator. The results reveal that decision styles-Rational/Intuitive and Participative/Autonomous-vary in performance according to task demands, with effective knowledge management further boosting outcomes. This framework advances efficient AI integration in enterprise settings, enhancing both decision flexibility and execution efficiency.