Monitoring service operations in digital banking platform ecosystems through theory-guided LLM analytics: Insights into service quality, operational risk, and customer experience

Digital banking platforms operate as complex service ecosystems involving banks, technology providers, payment infrastructures, and regulators, creating challenges for monitoring service performance, operational risk, and accountability. Understanding how customers evaluate these dimensions remains important to digital service management and technology-adoption research. This study develops a theory-guided large language model–structural equation modeling (LLM–SEM) framework that transforms unstructured customer reviews into theoretically grounded indicators for behavioral analysis. Drawing on UTAUT-3, perceived risk theory, and attribution theory, the framework uses structured chain-of-thought prompting with LLaMA 3.1-8B to extract indicators of service quality, operational risk, customer experience, behavioral outcomes, and actor attribution from customer reviews of leading smart banking applications in Saudi Arabia. The resulting indicators are examined using PLS-SEM and attribution-oriented text mining. Performance expectancy, trust, and hedonic motivation exhibited strong associations with usage experience, which was strongly associated with behavioral intention; behavioral intention, in turn, was strongly associated with recommendation intention. System performance and service delay risks emerged as salient operational-risk correlates of behavioral intention, whereas financial risk showed a limited association. Customers predominantly associated positive and negative service outcomes with the focal platform rather than with banks or regulators. This pattern supports the concept of platform-centered accountability, whereby operational responsibility may be distributed across ecosystem actors while perceived accountability is concentrated on the primary service interface, positioning the platform in customer narratives as the central ecosystem orchestrator and focal point of accountability. Multi-inference analysis, combined model–prompt analysis using GPT-5.6 Sol, and survey-based validation provide complementary evidence of methodological stability and the robustness of the findings. The study contributes to service operations, platform ecosystem, technology-adoption, and AI-assisted behavioral research by showing how theory-guided LLMs can complement established measurement approaches as scalable semantic measurement instruments for transforming customer-generated content into operational and behavioral insights. The findings also underscore the importance of service operations monitoring, ecosystem governance, and perceived platform accountability within FinTech ecosystems.