Abstract
A recent research manifesto has introduced the vision of AI-Augmented BPM (ABPM), where BPM systems are infused with AI to continuously adapt and improve a set of business processes with respect to one or more performance indicators. In the ABPM lifecycle, process modelling is lifted to the more general notion of process framing, which aims at capturing the boundaries within which the executions of one or more processes of interest should be confined. In this paper, I argue in favour of constraint-based declarative process specifications for process framing. I provide a list of key features that are needed towards this goal, and show how they are matched by research milestones recently obtained in this setting. In particular, I discuss how to deal with deviations, uncertainty, and object-centric processes.