Expertise Locator Systems (ELS) are recognized as key tools in knowledge management programs. However, they face significant challenges, such as maintaining expert profiles, knowledge representation, and evaluating their effectiveness. This research conducted a systematic literature review to explore how Large Language Models (LLMs) from generative artificial intelligence can contribute to overcoming these challenges and enhance the implementation of ELS. The findings highlight the importance of prompts, augmented information retrieval, and knowledge graphs in the development and application of LLMs. Based on these insights, we propose an architecture for ELS supported by LLMs and present a prototype to assess the feasibility of this approach. Preliminary results indicate a substantial potential of LLMs to foster the development of a new generation of knowledge-based systems.