This study proposes an AI-enhanced multi-criteria evaluation framework for the selection of national-level track and field coaches, integrating expert knowledge with computational intelligence. Using the Modified Delphi Method, DEMATEL, and Analytic Network Process (ANP), five key dimensions and twenty evaluation criteria were established through literature review and expert interviews. Natural Language Processing (NLP) was applied to extract and validate initial criteria from expert narratives, while causal relationships and relative weights were determined via expert surveys and DEMATEL analysis. ANP was employed to compute global priorities and rank criteria. To further strengthen the model's applicability, AI-based decision support mechanisms were embedded to simulate selection scenarios and provide predictive insights. Results indicate that “coaching experience,” “scientific training application,” and “continuous professional development” are among the most critical selection criteria. This hybrid framework not only improves transparency and objectivity in the coach selection process but also serves as a strategic tool for talent identification in elite sports systems.