Impact of Generative AI on Agile Software and Product Teams: A Structured Review and Human-in-the-Loop Adoption Framework

Abstract

Generative artificial intelligence (GenAI) is increasingly embedded in software engineering and product-development workflows, particularly through large language models that can generate code, tests, documentation, summaries, design alternatives, and natural-language explanations. This paper examines how such capabilities interact with Agile ways of working, where value is delivered through short feedback cycles, cross-functional collaboration, empirical learning, and shared ownership. A structured review of research and practitioner evidence is used to organize GenAI's effects across product discovery, requirements, backlog refinement, sprint planning, implementation, testing, review, release, and continuous improvement. Evidence indicates that AI assistance can accelerate selected development tasks and reduce friction in activities such as code generation and comprehension, while practitioner surveys also report perceived benefits in code quality, learning, and customer alignment. At the same time, faster generation can shift bottlenecks toward validation, integration, security review, architectural consistency, and decision quality. The paper therefore treats productivity as a multidimensional construct rather than a simple measure of output volume. A Human-in-the-Loop GenAI Agile Adoption Framework is proposed around five recurring controls: framing, generation, verification, decision, and learning. The framework connects product, engineering, quality, security, and governance responsibilities and introduces measurable adoption dimensions spanning delivery flow, quality, developer experience, product outcomes, and risk. The resulting model positions GenAI as an augmentation layer within Agile systems rather than as a substitute for human accountability.