Generative AI in Agile Software Development: A Comprehensive Survey

Abstract

Generative artificial intelligence (GenAI) adoption is reforming the software development life cycle (SDLC) practices, and Agile is no exception to this trend. The use of generative AI tools, huge language models, and domain-specific copilots is on the rise in automating coding, refactoring, testing, documentation, and requirements engineering tasks. Although initial data suggest benefits in productivity, quality, and time-to-market, GenAI also poses challenges related to technical dependability, integration, ethical challenges, and organizational maturity. However, despite this rapid evolution, currently lack systematic insight into how generative AI aligns with Agile values and practices, resulting in a fragmented research terrain. This survey critically reviews the studies related to generative AI in Agile-based SDLC. The contribution is a structured overview of applied ML across SDLC stages, Agile practices, and types of generative models through an easily comprehensible taxonomy. Second, by investigating such studies in terms of research methods, application areas, evaluation measures, and tools, it further synthesizes methodological findings. Faster code generation, automated testing and documentation, better prioritization of the backlog, higher developer productivity, and cost savings are contrasted against hallucination, scalability, bias, and adoption challenges. The survey also highlights opportunities and future research scope, including human–AI collaboration, benchmark standards, interpretable and trustworthy AI, domain-specific frameworks, and others. This paper advances the understanding of the implications of generative AI for Agile SDLC, from both the opportunities and challenges perspectives, and lays the foundational ground for future research and practice.

Ähnliche Arbeiten