In administrative processes, such as financial or governmental processes, humans typically do most of the work and must be allocated to tasks in an efficient manner. This allocation is made complicated by the different authorizations and the varying effectiveness of people for tasks. Moreover, administrative processes operate under substantial uncertainty, as the customer's journey through the process typically is uncertain upon their arrival.
To help solve this problem, this talk presents a novel model for resource allocation in administrative processes and delineates its differences from existing resource allocation models, highlighting the computational complexity that is inherent to the model. The presentation proceeds to show several solution methods tailored to address the challenges posed by the model. Specifically, it discusses heuristics, periodic stochastic programming, and deep reinforcement learning. The talk specifically addresses the challenges that are encountered when implementing these solution methods, some of which remain unresolved. By doing so the presentation aims to shed light on promising avenues for future research in this domain.