ASSESSING LLM’S THROUGH SCENARIO-BASED SOFT SKILLS TASKS FOR MANAGERS: EXPLORATORY STUDIES

Purpose: The aim of this study is to analyze the training potential of large language models (LLMs) in developing managerial soft skills. By potential, we refer to the model’s demonstrated understanding of the key elements involved in resolving conflict scenarios. Need for the study: Soft skills remain a critical component of managerial effectiveness, yet traditional training methods often lack scalability and accessibility. With the rise of advanced language models capable of human-like interaction, there is an opportunity to explore their applicability in leadership development. This study addresses a gap in current literature by evaluating the practical use of LLMs for supporting the growth of interpersonal and emotional intelligence among managers. Methodology: The study employed an experimental design based on ten conflict scenarios representing typical managerial challenges. Each scenario consisted of a short background narrative, a description of the employee’s personality, and a task for the manager. The scenarios were processed by three large language models—GPT-4o, Gemini 2.0 Flash, and Claude 3.5 Haiku. The model responses were evaluated by the authors and supervisory LLM model across seven categories. Findings: All tested models demonstrated a solid understanding of conflict dynamics and offered appropriate managerial responses. The supervisory distinctly favored Claude 3.5 Haiku, which it rated notably higher than the other models. Practical Implications: The study confirms that large language models can potentially support managerial development by offering interactive, on-demand, and risk-free simulations. A solid understanding of the dynamics underlying conflict scenarios may indicate the validity of using them in simulation-based training.