The great divide: An empirical study of the gap between AI-driven and traditional product backlog prioritization

Product backlog prioritization is a crucial step in development of software, commonly performed by product managers (PM), product owners (PO), or other product professionals. Traditional prioritization methods used in practice have different challenges, like biases or a lack of efficiency. Many alternative methods are developed in academia, but remain largely unused in industry practice. This could be because of a lack of exposure and science communication, technology challenges, or inertia. This research addresses the existing gap through a hybrid approach: systematic literature review (SLR) and survey. The SLR uncovers a wealth of prioritization methods and classifies them into nine methodological groups. The survey of 307 product professionals, with 287 qualifying for final analysis based on recent prioritization experience, shows important trends related to satisfaction and use levels of traditional methods, as well as AI/ML-based methods. Conventional methods such as MoSCoW, RICE, and WSJF are most common, but satisfaction levels vary. Only 7.3% respondents frequently use AI and ML methods; however, 63.4% are open to trying them in the future, which shows potential for future adoption. This research calls for a better alignment between scholarly innovation and industry practice, and its findings provide an empirical basis for understanding backlog prioritization trends.