Business-to-business software systems are inherently specialized and operationally intricate, which make them crucial to develop accurate personas that reflect real end-user requirements throughout the development lifecycle. As user needs continuously evolve over time, it becomes imperative to establish a data-driven framework capable of persistently updating these personas and promptly integrating those changes into the development process to maintain long-term value delivery. Conventional persona generation techniques typically depend on clustering approaches applied to qualitative and quantitative data—a process that is time-intensive, expensive, and requires considerable domain expertise. This study introduces an automated method for continuous persona generation, extracting user requirements from an ongoing stream of user data. The approach utilizes large language models to interpret qualitative inputs and dynamically generate knowledge graphs, enabling real-time insights into shifting user needs. A case study involving inquiry call logs from a customer support center was conducted to validate the method. Results demonstrated that the proposed approach outperformed a traditional clustering-based baseline in approximately 86% of the cases in a question–answering task aimed at evaluating the structuring and retrieval of user requirements. Furthermore, clear insights into user pain points and requiring improvements were provided, reinforcing the effectiveness and practical utility of continuous, data-driven persona generation.

Although there is a vast body of literature related to methods for detecting change-points (CPs), less attention has been paid to assessing the statistical reliability of the detected CPs. In this paper, we introduce a novel method to perform statistical inference on the significance of the CPs, estimated by a Dynamic Programming (DP)-based optimal CP detection algorithm. Our main idea is to employ a Selective Inference (SI) approach---a new statistical inference framework that has recently received a lot of attention---to compute exact (non-asymptotic) valid p-values for the detected optimal CPs. Although it is well-known that SI has low statistical power because of over-conditioning, we address this drawback by introducing a novel method called parametric DP, which enables SI to be conducted with the minimum amount of conditioning, leading to high statistical power. We conduct experiments on both synthetic and real-world datasets, through which we offer evidence that our proposed method is more powerful than existing methods, has decent performance in terms of computational efficiency, and provides good results in many practical applications.