What is Agentic BPM?

Agentic BPM (Agentic Business Process Management) extends classic business process management to autonomous actors: processes are no longer executed only by people and hard-coded systems but also by AI agents that make their own decisions within a process frame. Designing, steering, and improving processes must therefore define how much latitude an agent has and how its decisions stay traceable. The concept was set out systematically in a research manifesto by 18 BPM researchers, including Marlon Dumas, Peter Fettke, and Stefanie Rinderle-Ma (March 2026).

Why does this matter?

BPM was developed for flows whose steps are modeled in advance. Agents decide their steps at runtime – so BPM shifts from prescribing to setting frames and reviewing. The manifesto names four capabilities such agents must have: framed autonomy, explainability, conversational actionability, and self-modification.

What does this look like in practice?

In the complaints process, the process model no longer fixes every step but the frame: the goal is a resolution within five days, credit notes above €500 need approval, customer data may only be changed in the CRM. Within that frame the agent decides itself whether to send a replacement, propose a credit note or ask for details — and justifies each choice.

Not to be confused with

Not the same as agentic process automation: APA describes automating individual flows with agents, agentic BPM the entire management cycle from design to improvement. Nor a departure from process models: the model remains, but it defines boundaries and goals rather than every single activity.