This study aims to present design principles for a next-generation learning management system (LMS) that reflects edutech trends and the higher education environment from 2025 to 2030, specifically for designing an agentic AI LMS based on generative AI. To achieve this, a review of prior research on Learning Management Systems (LMS), Large Language Models (LLM), Retrieval-Augmented Generation (RAG), LTI Advantage, and Learner Experience (LX) confirmed that the core requirements for next-generation LMS converge on personalized learning support, real-time feedback, learning analytics and prediction, and instructor task automation. Accordingly, we demonstrated an RAG-based architecture mitigating LLM hallucination and timeliness issues, an LTI Advantage-based hub structure for securely integrating external AI tools, and an intent-based LX design prototype minimizing cognitive load. Simulation evaluation showed that applying RAG v2 reduced the hallucination rate from 37.0% to 8.0% and improved the evidence provision rate to 97.2%. Furthermore, after LTI Advantage integration, the data loss rate decreased to 0.4%, and the interoperability index was evaluated at 0.89. The application of intent-based LX resulted in perceived usefulness of 4.62/5 and perceived ease of use of 4.48/5, confirming its effectiveness in improving the learning experience. In summary, the agentic LMS combining RAG - LTI Advantage - intent-based LX presents a practical design alternative capable of transforming existing management-centric LMSs into learner-centered, active learning environments.
Jung et al. (Sat,) studied this question.