Task distribution and human resource management using reinforcement learning

The process of assigning tasks in large companies is a costly expenditure of human resources. Usually, many people are employed to distribute tasks as best as possible among the people involved in the projects. While there are software applications that support this effort, they are limited, and the people who make the decisions about where to send the various tasks considering load balancing, evaluating the capabilities of the possible solvers and many other factors are still handled manually. In this paper, we propose a solution using reinforcement learning to train an automatic agent capable of managing the process itself, thus reducing human effort and cost. Our method first attempts to learn from existing datasets and then improve itself in an unsupervised manner. The results are promising and validate our original idea that using an automated agent to address the observed gap can be a valuable addition to existing task management applications.