Retrieval-Augmented Generation Agents for Decision-Support Systems: A Scoping Review

Recent advances in artificial intelligence have significantly expanded the capabilities of large language models across knowledge-intensive tasks. However, standalone large language models often exhibit limitations related to factual grounding, contextual relevance, and reliability when deployed in real-world decision environments. These challenges become particularly critical in decision-support systems where outputs must be traceable and aligned with domain-specific evidence. Retrieval-Augmented Generation (RAG) has emerged as a promising architectural paradigm that integrates external information retrieval with generative language models, allowing systems to retrieve relevant documents before producing responses. This hybrid architecture enhances contextual grounding, reduces hallucination risks, and improves the reliability of AI-assisted decision processes. Despite the growing adoption of RAG-based architectures, the literature remains fragmented across application domains, architectural designs, and evaluation strategies. Therefore, this study aims to systematically map the current research landscape of RAG-enabled intelligent agents applied to decision-support systems through a scoping review methodolog