This paper presents the results of a systematic mapping review of 53 peer-reviewed publications from 2020 to 2025, which map the integration of Artificial Intelligence (AI) in Agile Software Development (ASD). Key findings reveal broad coverage of machine and deep learning techniques, dominated by Natural Language Processing (NLP) (64%) and generative AI (39.5%), alongside other hybrid methods, such as predictive analytics (20.7%). These studies primarily focus on supporting sprint planning (20 studies), testing (21 studies), and requirements/design (21 studies) through automation and predictive insights. This has led to improved efficiency in 37 studies, representing 70% of studies, and enhanced quality in 47% of them. However, challenges remain, particularly regarding data quality, ethical concerns, and integration complexities. The increase in research since 2023 underscores AI’s potential to enhance agile practices, but it also highlights the need for empirical validation and an ethical approach to address gaps in adoption.