This study develops and evaluates a Retrieval-Augmented Generation Large Language Model (RAG-LLM) system to enhance investment decision-making among financial consumers in South Korea. By integrating Large Language Models with Retrieval-Augmented Generation techniques, the system provides personalized, data-driven investment advice tailored to individual risk profiles inferred from demographic and financial characteristics. Utilizing publicly available financial data and consumer behavior literature, the model retrieves relevant information and generates recommendations regarding asset selections. The system's performance is assessed using portfolio metrics like expected returns, risk levels, Sharpe ratios, and utility based on constructed mean-variance optimal portfolios, and compared against naive random selection and traditional LLM-based systems. Preliminary results indicate that the RAG-LLM significantly outperforms baseline models, leading to higher Sharpe ratios and utility with reduced risk. This approach enhances financial decision-making, particularly benefiting financially marginalized groups who lack access to traditional advisory services. The research underscores the potential of AI-driven solutions in promoting financial inclusion, reducing disparities in investment outcomes, and contributing to a more equitable financial ecosystem.