ReflectAI: Design and Evaluation of an AI Coach to support Public Servants’ Self-Reflection

Näscher, Hans; Schaffner, Jan; Koddebusch, Michael

Abstract

Digital transformation in public administration requires the development of digital government competences (DGCs) among public servants. Self-reflection, a critical personality-related DGC, fosters adaptability and continuous learning but remains challenging to cultivate through traditional training.
Employing a design science research approach, we developed ReflectAI, an LLM-based AI coach designed to support public servants in developing self-reflection competence. The system was evaluated through a two-week user study with seven public servants.
Our findings revealed three key benefits: increased awareness of self-reflection opportunities, improved thought structure, and valuable conversation documentation. Implementation challenges included time constraints, technical infrastructure limitations, and organizational support.
Our study contributes to research by demonstrating the applicability of AI coaching to personality-related DGC development, thus extending the body of knowledge of both competence development and AI coaching. 
With regard to practice, we highlight the importance of structured self-reflection processes and the required organizational conditions necessary for successful adoption. 
The resource-efficient approach to developing self-reflection competence facilitates this adoption through an innovative application of AI in public organizations.

Keywords
AI Coaching; Digital Government; Self-Reflection; Design Science Research