Leveraging AI for Continuous Quality Assurance in Agile Software Development Cycles

Abstract

In the era of Agile software development, where rapid releases and continuous integration are essential, ensuring consistent software quality becomes increasingly complex. This research explores the integration of Artificial Intelligence (AI) into Continuous Quality Assurance (CQA) within Agile Software Development Cycles. Using the PROMISE dataset comprising 10,885 entries of real-world software metrics and defect labels, we implemented and evaluated AI models for real-time defect prediction, complexity analysis, and risk assessment. The Naive Bayes classifier achieved an accuracy of 98.16%, with high precision and recall across both defective and non-defective classes. Linear Regression, applied for defect-proneness estimation, yielded a low RMSE of 0.25, indicating strong predictive performance and effectively predicting defect-prone modules. As a result, our approach led to a 15.48% reduction in post-release bugs and an 80.71% decrease in manual testing time, significantly improving sprint-level feedback and delivery quality. Compared to manual testing, the AI-driven approach significantly improved defect detection rates and reduced testing time, supporting faster and more reliable software delivery. These results validate that AI integration in Agile environments not only automates and accelerates the quality declaration development but also sustains software reliability in all iterative development cycles.