🤖 AI Summary
This study addresses the escalating complexity of software architectures driven by cloud-native paradigms, microservices, and AI integration by systematically reviewing literature from 2024 to 2025. Focusing on five key dimensions—architectural modeling, quality attributes, self-adaptation mechanisms, AI-assisted decision-making, and architectural evolution—the work employs a thematic synthesis approach to integrate, for the first time, AI-enabled architectural decisions with continuous governance perspectives. It emphasizes the synergy among multi-view modeling, domain-driven decomposition, and runtime observability. The review identifies critical research gaps, including the absence of standardized frameworks for trustworthy AI architectures, insufficient empirical validation, weak integration of security and privacy concerns, and limited investigation into edge and serverless contexts. These insights offer actionable pathways to enhance scalability, maintainability, and software sustainability.
📝 Abstract
Software architecture has evolved considerably in response to the increasing complexity of modern software systems, particularly those based on cloud computing, microservices, artificial intelligence (AI), and distributed computing environments. This literature review synthesizes recent studies published between 2024 and 2025 to examine emerging trends, challenges, and future directions in software design and architecture. The review adopts a thematic synthesis approach to analyse contemporary research across five major areas: architectural modelling and representation, software quality attributes and self-adaptive architectures, architectural evolution and complexity management, artificial intelligence-assisted architectural decision-making, and existing research gaps. The findings indicate that modern software architecture extends beyond traditional structural design to support continuous architectural governance, stakeholder communication, runtime observability, resilience, and intelligent decision support throughout the software lifecycle. Furthermore, the reviewed studies demonstrate that multiple architectural views, continuous monitoring, domain-driven decomposition, and AI-assisted design techniques contribute significantly to improving scalability, maintainability, adaptability, and long-term software sustainability. Despite these advances, several research gaps remain, including limited empirical validation of proposed approaches, insufficient integration of security and privacy into architectural decision-making, inadequate exploration of emerging paradigms such as edge and serverless computing, and the absence of standardized frameworks for trustworthy AI-assisted architecture.