Process automation is a key subfield of business process management. Recent advances in AI research promise to yield a new type of intelligent process automation that can support high-variability, flexible, knowledge-intensive processes previously hard to enhance with process automation. However, primarily proposed, subsymbolic deep learning approaches fail to reliably consider the complex knowledge inherent to these processes and provide adequate explanations for their decisions. Neuro-symbolic reasoning approaches based on knowledge graphs promise to address these challenges by allowing to holistically encode complex domain knowledge and to perform explainable reasoning thereupon. In this vision paper, we investigate the potential of knowledge graphs for intelligent process automation. Using tangible examples, we show how they can be used to enable explainable, knowledge-aware process automation, integrating a wide range of process knowledge. We show that such knowledge-aware process automation can contribute to addressing two current challenges of the BPM community: the automation of knowledge-intensive processes and the design of AI-augmented business process management systems. Finally, we discuss avenues for future research.