Abstract:

Generative AI and agent-based systems challenge the traditional BPM view of processes as predefined flows executed by humans and information systems. AI agents can interpret context, select actions, recommend decisions, and influence process paths, making business processes partially autonomous rather than merely automated.
This paper argues that BPM therefore needs principles for governing agent autonomy. We propose the Agentic Process Governance Layer as a research direction for defining what agents may do independently, when they must escalate to humans, and what traces of their actions should be recorded. This shifts BPM from asking “what else can be automated?” to asking how organizations can responsibly manage processes that begin to reason, decide, and adapt on their own.