Although there has been a lot of progress in developing Process Mining (PM) algorithms and Artificial Intelligence (AI) techniques in recent years, no effort has been put in developing a common means of mining knowledge-based behavior of Artificial Neural Networks (ANN). In a design-science-oriented way, in this paper, elements of a new kind of AI-PM approach are outlined and demonstrated with ANN. These intend to enable (1) AI engineers to mine an ANN’s inner processes to discover its knowledge-induced behavior, realize conformance checking, e.g. w.r.t. an ANN required behavior, and improve ANN due to enhancement. To illustrate the application of this new approach, a set of novel model views and algorithms are proposed, which are demonstrated on simple example logs. Findings show that AI-PM supports the clarification of ANN behavior: As the ANN’s inner activities and knowledge generation can be mined, its non-transparent black box is unveiled and trustworthiness of ANN is supported.