Retailers sell a multitude of products which are interrelated to each other. However, often the strategies to price products neither consider these dependencies nor changing market condi-tions, resulting in inefficient price setting. The objective of this study is to investigate whether a system of multiple agents, each representing a single product, combined with a machine learn-ing approach can optimize pricing strategies. To achieve this objective, a design science re-search approach is used to implement a multi-agent reinforcement learning (MARL) system that learns a pricing policy for a product cluster and aims on maximizing the cluster’s total profits by optimizing the prices of products dynamically. Six market simulation scenarios with prede-termined market events were used to evaluate the MARL system in comparison to static pricing strategies and single-agent approaches. In all six scenarios, the MARL system leads to increas-ing profits: The daily average profits were 7.8% higher in comparison to the static pricing strat-egy and even 27% higher in comparison with a single-agent approach. The results indicate that retailers can gain a significant competitive advantage by considering product clusters and uti-lize ML algorithms to implement dynamic pricing strategies.