International Journal of Science, Engineering and TechnologyAuthors: Tanishka Bhatia
Abstract: The integration of Reinforcement Learning into Enterprise Information Systems marks a pivotal transition from reactive data management to autonomous, goal-oriented decision-making. While traditional Enterprise Resource Planning and Customer Relationship Management systems have historically functioned as static repositories of business data, the increasing volatility of global markets demands systems capable of real-time adaptation and strategic optimization. This review article systematically examines the application of reinforcement learning techniques across core enterprise domains, including supply chain logistics, dynamic pricing, and human capital management. By framing business processes as Markov Decision Processes, organizations can deploy intelligent agents that learn optimal policies through continuous interaction with operational environments. The analysis highlights the technical shift toward deep reinforcement learning and multi-agent systems, emphasizing the role of digital twins and high-fidelity simulations in bridging the sim-to-real gap. Furthermore, the article addresses critical implementation challenges, such as sample inefficiency, the black-box nature of neural policies, and the necessity of Reinforcement Learning from Human Feedback to ensure alignment with corporate ethics. Ultimately, the synthesis of these findings provides a comprehensive roadmap for transforming information systems into strategic intelligence assets, paving the way for the emergence of the autonomous enterprise.
DOI: https://doi.org/10.5281/zenodo.18152956
