Predictive business process monitoring deals with predicting a process’s future behavior or the value of process-related performance indicators based on process event data. A variety of prototypical predictive business process monitoring techniques has been proposed by researchers in order to help process participants performing business processes better. In practical settings, these techniques have a low predictive quality that is often close to random, so that predictive business process monitoring applications are rare in practice. The inclusion of process-context data has been discussed as a way to improve the predictive quality. Existing approaches have considered only structured data as context. In this paper, we argue that process-related unstructured documents are also a promising source for extracting process-context data. Accordingly, this research-in-progress paper outlines a design-science research process for creating a predictive business process monitoring technique that utilizes context data from process-related documents to predict a process instance’s next activity more accurately.