AI-Enhanced NLP for Agile Strategy Execution: Leveraging Machine Learning to Automate Backlog Grooming and Sprint Planning at Scale

Abstract

Agile methodologies achieved significant efficiency improvements for software developmentprojects through ongoing delivery practices, iterative work cycles, and collaborative teamwork. The2efficiency of backlog grooming and sprint planning through manual processes deteriorates whenprojects grow. This research develops an NLP and ML-powered AI framework that modifies manual. Agile decision-making processes for backlog evaluation and sprint planning operations. The studyutilisesreal-world Jira data from Kaggle, which contains more than 4.9 million records that combinetextual description elements with structured metadata details. Logistic Regression using TF-IDFvectors analysed issues to identify their type between Bug, Story, and Task, while XGBoost analysedsemantic and structural features to forecast task priorities. The model based on logistic regressionproduced a complete 100% classification precision alongside XGBoost, which generated 86.3% totalaccuracy supported by robust ROC-AUC metrics at each priority level. Experimental tests show thatthe developed framework combines automated Agile decision processes with maintenance ofinterpretability features alongside high scalability needed for enterprise use. The introduction ofexplainable models allows Agile teams to understand decisions while gaining trust to utilise themodel-generated insights. Real-world data enhances both the reliability and the practical relevance ofthe system, which surpasses the synthetic datasets that other studies use. The research delivers anautomated Agile operation solution that can be easily reproduced and lightened for more innovativecontext-aware multilingual human-in-the-loop support systems in large-scale software development.