Artificial Intelligence-Driven Software Engineering: Improving Software Quality, Development Productivity, and Intelligent Automation — A Comprehensive Review

Artificial intelligence (AI), particularly machine learning (ML), deeplearning, natural language processing (NLP), and large language models(LLMs), is changing how software is specified, designed, implemented, tested, maintained, and governed. This paper presents a comprehensivenarrative-informed systematic review of AI-driven software engineering(AI-SE), emphasizing software quality, developer productivity, andintelligent automation. It synthesizes established AI-for-software- engineering research with emerging LLM-based workflows and identifieswhere evidence is promising, mixed, or insuf icient. The review organizesapplications across the software development life cycle (SDLC):requirements engineering, architecture and design, code generation andreview, testing and verification, defect prediction, security analysis, maintenance, project management, and DevOps. It also develops a human- centered reference architecture and a risk-aware adoption framework. Across the literature, AI systems can accelerate bounded tasks such as codecompletion, test generation, documentation, search, and classification;however, benefits are conditional on task complexity, repository context, developer expertise, tool integration, and verification practices. Generatedartifacts can contain functional defects, insecure patterns, outdateddependencies, licensing uncertainty, and plausible but unsupportedexplanations. Therefore, productivity should be measured as end-to-enddelivery of accepted, maintainable, secure work—not typing speed orsuggestion acceptance. Quality should be evaluated through layeredevidence: executable tests, static and dynamic analysis, security checks, code review, traceability, operational outcomes, and human judgment. Thereview proposes research questions, a reproducible search and screeningprotocol, thematic synthesis, and a measurement agenda. It concludes thatAI-SE is best understood as socio-technical augmentation rather thanautonomous replacement: organizations should introduce AI with explicitboundaries, human accountability, data governance, provenance, continuous evaluation, and feedback from production. The paper providesa foundation for researchers and practitioners designing reliable, equitable, and measurable AI-enabled engineering environmen