LLM-Powered Enterprise Applications: A Systematic Review of Cloud Computing, Payment Gateway, and Risk Management Architectures

Abstract

Large language models (LLMs) are increasingly embedded within enterprise technology stacks, extending beyond conversational interfaces into core operational systems including cloud infrastructure, payment processing, and risk management. This systematic review synthesizes current architectural approaches to LLM-powered enterprise applications across three interdependent layers: the cloud-computing infrastructure required to deploy LLMs reliably at production scale, payment-gateway applications in which LLMs are increasingly used for fraud detection and transaction-risk analysis (Cao et al., 2024), and the risk-management and compliance architecture required to govern LLM-driven enterprise decisions. The review draws on applied research addressing zero-downtime AI model updates in real-time inference systems (Sannidhanam, 2023a), AI-driven failure prediction and automated recovery in self-healing distributed systems (Sannidhanam, 2023b), event-driven and event-streaming architectures for high-volume transaction and claims processing (Event Streaming Architectures for High-Volume Transaction Processing, 2022; Sannidhanam, 2021), and enterprise risk-governance architecture, including configurable workflow-based risk management (Basireddy, 2022) and audit-ready compliance platforms specifically designed for LLM-regulated enterprises (Basireddy, 2023). The review finds that reliable enterprise LLM deployment depends on a layered architecture in which real-time, event-driven data infrastructure, resilient and self-healing cloud operations, and auditable governance controls are treated as co-equal design requirements alongside model performance, and it identifies payment-gateway fraud detection as one of the most mature and rapidly growing applied use cases for enterprise LLM deployment. The review concludes by identifying gaps in the integration of these three architectural layers and outlining priorities for future research on production-grade, governed LLM enterprise systems.