Quality-Driven Architectural Evaluation for AI-Enabled Cloud-Native Systems: An Evidence-Grounded Framework for Scalability, Resilience, Performance, Cost Efficiency, and Sustainability

This preprint presents an evidence-grounded framework for evaluating AI-enabled cloud-native systems across five critical quality dimensions: scalability, resilience, performance efficiency, cost efficiency, and sustainability. The paper synthesizes ISO/IEC 25010, ATAM, cloud-native architecture guidance, and recent empirical studies to support a practical five-step architecture review process. It shows how scenario-based evaluation and evidence-backed trade-off analysis can improve architectural decision-making in modern distributed systems. An illustrative fraud-analytics case demonstrates how the framework can be applied in regulated, latency-sensitive, and cost-conscious AI-enabled environments. The contribution of the paper is a structured and auditable approach for comparing architectural options such as centralized microservices, event-driven microservices, serverless-centric systems, and edge-cloud hybrids.