The Application of AI Customer Service

The widespread adoption of AI-powered customer service systems has transformed how users engage with digital platforms. However, understanding how users form satisfaction and trust toward these systems remains a major challenge, especially when emotional responses and platform-level design factors interact in complex ways. To address this, the authors propose a Multi-Level Structural Equation Modeling (ML-SEM) framework that captures both individual-level perceptions (e.g., flow, trust, empathic accuracy) and organization-level design variables (e.g., interface quality, anthropomorphic and empathic features). The model simultaneously estimates direct and mediated effects, including cross-level pathways linking chatbot design to satisfaction. Using two large-scale, multi-domain datasets (Bitext and Twitter), they evaluate ML-SEM against six recent baseline models. Results show that ML-SEM outperforms all baselines, achieving the lowest RMSE (0.243) and highest adjusted R2 (0.691) in satisfaction prediction, with statistically significant improvements (p < 0.01).