Automated Software Engineering with Generative AI: Code Generation, Testing and Maintenance

Abstract

Generative Artificial Intelligence (GenAI) is rapidly transforming software engineering by enabling machines to generate, analyze, test, document and maintain software with increasing levels of automation. Traditional software development depends heavily on human programmers for requirements interpretation, code implementation, debugging, testing and maintenance. The emergence of large language models (LLMs), code-specialized foundation models and AI-powered development environments has introduced a new paradigm in which artificial intelligence can participate throughout the software development lifecycle. This paper examines the role of Generative AI in automated software engineering, focusing specifically on code generation, software testing and program maintenance. It analyzes the evolution from traditional automated programming techniques to contemporary AI-based code generation systems and discusses how transformer-based models learn programming patterns from large-scale repositories. The paper explores applications including natural-language-to-code generation, code completion, refactoring, bug detection, test-case generation, automated documentation, vulnerability detection and legacy-code modernization. Particular attention is given to the benefits of AI-assisted development, including increased developer productivity, accelerated prototyping, reduced repetitive work and improved access to programming capabilities. At the same time, important challenges remain, including incorrect code generation, software vulnerabilities, hallucinated APIs, inadequate testing, intellectual-property concerns, data privacy, technical debt and overreliance on AI-generated solutions. The paper argues that Generative AI should not be regarded simply as a replacement for software developers but as an intelligent engineering technology that can augment human expertise. Future software engineering environments are likely to integrate generative models with automated testing, formal verification, repository-level reasoning, continuous integration and autonomous software agents. Responsible human oversight, rigorous validation and secure development practices will remain essential for reliable AI-driven software engineering.