The rapid advancement of artificial intelligence (AI) technologies has introduced the potential for AI- based mentoring or e-coaching through the integration of character simulacra. Organizational mentorship programs are expected to enhance skill development, provide networking opportunities, and offer practical advice on work-life balance. However, despite these benefits, practical challenges persist, such as difficulty acquiring qualified mentors and the time constraints associated with mentoring. Consequently, an alternative mentorship strategy utilizing advanced AI based on large language models such as ChatGPT is gaining attention to address these challenges and enhance the effectiveness of mentorship programs. This study aims to identify the critical attributes of AI-based mentoring systems and empirically analyze how these attributes impact mentee satisfaction through AI mentor efficacy. Additionally, the moderating effect of anthropomorphism on satisfaction was examined. Through conceptual discussions and empirical research, this study identifies essential attribute variables of AI-based mentors and provides insights for future user research in mentoring services.