A Planning Approach to Agile Project Management. The JIRA Planner

Currently, planning software development is mostly performed by humans. This task requires reasoning over multiple factors and constraints. Thus, it takes humans time to generate those plans. Also, the resulting plans often do not estimate well how good specific developers will address a task or how long it will take. In this paper, we present an automated approach to generate project plans, assigning human resources to tasks according to previous projects, and taking into account a variety of real constraints. We combine machine learning, to acquire a person’s key skills based on previous development tasks performed, with planning technology in order to provide a unified end-to-end project management tool. This planning application has been developed within a large corporation utilizing historical data from an internal system that stores project tasks. We validate our approach by comparing the plans originally proposed by humans, against those generated by the planning tool. For this comparison, we contribute a set of metrics that assess different properties of the plans, such as the quality of task assignments.