Abstarct: Abstract Story point estimation in agile project management plays a pivotal role in planning and resource allocation. Traditional estimation methods often suffer from significant inaccuracies due to their reliance on human judgment and lack of transparency. This research presents a BERT-baxsed classification model for story point estimation. In addition to improving accuracy, this model can gain the trust of experts by making its results explainable. To achieve this, textual data from user stories are processed, and their semantic features are extracted using a pre-trained BERT model. These features are then classified into four categories: small, medium, large, and huge. The model's explainability is enhanced by clustering the results using the K-Means algorithm. Experimental results demonstrate that the proposed model achieves an accuracy of 88.71% in classifying story points, outperforming similar methods by effectively identifying complex patterns in task descxriptions. Error analysis reveals that some tasks may be misclassified due to semantic similarities. Specifically, while the model exhibits high accuracy for the large and huge categories, it is more prone to errors when classifying small and medium tasks. By providing a systematic method to assess the reliability of the model's estimations, software development teams can identify and correct inaccurate predictions. This research highlights that employing explainable, BERT-baxsed machine learning models for story point estimation not only increases accuracy but also provides project managers and development teams with a clearer understanding of the model's decision-making process. This, in turn, can lead to improved planning processes and a reduction in costs associated with improper resource allocation.