A Systematic Literature Review of the Purpose of Artificial Intelligence Use in Software Architecture Development

Abstract

Artificial Intelligence (AI) is increasingly applied as a computational support capability across software engineering. However, evidence concerning the purposes for which AI is used in software architecture development remains fragmented across task specific studies. This paper reports a systematic literature review that synthesizes AI enabled support as an engineering application involving decision support, architectural artifact processing, and bounded human AI collaboration. The review searched IEEE Xplore, ACM Digital Library, and Scopus, screened 62 publications, and included 11 primary studies for full text evidence synthesis. The extracted evidence was analyzed through thematic synthesis supported by initial coding and systematic evidence mapping. Seven evidence derived purpose categories were identified: architectural knowledge and design rationale recovery or generation; architecture traceability and entity extraction; software architecture description recovery or generation; architecture refactoring, modernization, and decomposition; architecture conformance and governance checking; architecturally significant requirement elicitation; and collaborative architecting decision support. Across the reviewed corpus, AI is concentrated on artifact centered and decision support tasks rather than autonomous orchestration of an end to end software architecture development. AI mechanisms generate, recover, classify, link, transform, or validate architecture related artifacts, while architects and other stakeholders remain responsible for context definition, trade-off analysis, validation, acceptance and accountability. This paper contributes an evidence derived classification providing bounded characterization of AI enabled engineering support and a traceable basis for subsequent software architecture development method design.