Source-linked AI summary
Agentic Business Process Management: A Research Manifesto
Diego Calvanese, Angelo Casciani, Giuseppe De Giacomo, Marlon Dumas, Fabiana Fournier, Timotheus Kampik, Emanuele La Malfa, Lior Limonad, Andrea Marrella, Andreas Metzger, Marco Montali, Daniel Amyot, Peter Fettke, Artem Polyvyanyy, Stefanie Rinderle-Ma, Sebastian Sardiña, Niek Tax, Barbara Weber
TL;DR
Organizations seek autonomous software systems, but agentic systems do not inherently ensure conformity with organizational processes, constraints, regulations, and goals. This manifesto introduces Agentic Business Process Management as a process-aware paradigm, develops its core abstractions and capabilities, and identifies open implementation and research challenges.
Problem
Agentic systems can host interoperating agents without realizing shared process awareness, leaving a need for a holistic approach to aligning autonomous agents with organizational processes and constraints.
Method
The manifesto develops APM around framed agency, defines process and normative frames, and synthesizes capabilities and research challenges through academic, industry, and Dagstuhl Seminar discussions.
Results
APM is articulated through four capabilities—framed autonomy, explainability, conversational actionability, and self-modification—and a conceptual architecture for governing autonomous agents in business processes.
Takeaways & Limitations
APM offers a roadmap for bridging BPM, AI, and multi-agent systems while emphasizing that LLMs alone are insufficient for truly intelligent BPM systems.
Takeaways & Limitations
The manifesto leaves implementation details open for process awareness, framing, framed knowledge and goals, explainability, conversational actionability, and self-modification.
Abstract
from arXiv · showhide
This paper presents a manifesto that articulates the conceptual foundations of Agentic Business Process Management (APM), an extension of Business Process Management (BPM) for governing autonomous agents executing processes in organizations. From a management perspective, APM represents a paradigm shift from the traditional process view of the business process, driven by the realization of process awareness and an agent-oriented abstraction, where software and human agents act as primary functional entities that perceive, reason, and act within explicit process frames. This perspective marks a shift from traditional, automation-oriented BPM toward systems in which autonomy is constrained, aligned, and made operational through process awareness. We introduce the core abstractions and architectural elements required to realize APM systems and elaborate on four key capabilities that such APM agents must support: framed autonomy, explainability, conversational actionability, and self-modification. These capabilities jointly ensure that agents' goals are aligned with organizational goals and that agents behave in a framed yet proactive manner in pursuing those goals. We discuss the extent to which the capabilities can be realized and identify research challenges whose resolution requires further advances in BPM, AI, and multi-agent systems. The manifesto thus serves as a roadmap for bridging these communities and for guiding the development of APM systems in practice.
1. Introduction and Motivation
The manifesto introduces Agentic Business Process Management (APM) to address the gap between autonomous-agent governance and the established organizational perspective of Business Process Management. It frames autonomy through process awareness so agents can pursue organizational goals within constraints.
- Increasingly autonomous LLM-based agents create business risks through compliance violations, social-norm violations, adverse outcomes, and technically uncertain behavior.
- APM addresses the lack of a holistic BPM perspective for governing autonomous agents as first-class participants in organizational processes.
- The manifesto introduces framed agency as the core of APM and develops a joint position from BPM, autonomous-agent, and industry perspectives.
- APM agents must support framed autonomy, explainability, conversational actionability, and self-modification.
2. Agentic Business Process Management Systems
An APM system is a socio-technical, agent-centric system in which human and technical agents perceive, reason, and act within process-aware frames. Its architecture distinguishes system-level coordination from agent-level control, knowledge, explanation, and self-modification.
- Fundamental concepts: APM treats agents as primary functional entities that sense, reason, and act to deliver software-system functionality.
- Architecture: Agents operate through continuous perception, reasoning, and action, using tools such as sensors, actuators, messaging, and services.
- Fundamental concepts: Process awareness requires agents’ inner workings to conform to organizational processes, operational constraints, regulations, and goals.
- Fundamental concepts: An APM system is a socio-technical system jointly realized by agents, some of which are at least partially process-aware.
- Fundamental concepts: An APM agent is an autonomous primary execution entity whose autonomy is framed to produce process-aware behavior aimed at process goals.
- Architecture: The agent reasoning module maintains mental and intentional models while supporting self-modification and explanation.
3. Envisioned Capabilities of an APM system
APM organizes autonomous behavior through frames and equips agents with explainability, conversational actionability, and self-modification. Together, these capabilities support bounded decision-making, coordination, governance, and adaptation within business processes.
- Framing: Frames define evolving rules, restrictions, and regulations that bound agents’ autonomous decisions at agent, process, and organizational levels.
- Framing: Normative frames specify what process behavior is allowed, while operational frames specify strategies or actions for accomplishing process objectives.
- Framing: Local frame operationalization can limit an agent’s process context and capabilities, creating a security and privacy boundary if the agent is compromised.
- Explainability: Explainability requires agents to produce framing-consistent, contextual, causally grounded, and interpretable explanations.
- Explainability: Explainability supports autonomy by enabling agents to resolve behavioral misalignment and reducing human intervention through understandable behavior.
- Conversational actionability: Conversational actionability combines interaction and enactment so agents can coordinate, receive instructions, and report to human and other agents.
- Conversational actionability: Selecting an effective mix of tools for functions such as time-to-completion prediction remains a research challenge.
- Self-modification: Self-modification enables agents to adapt and evolve over time toward individual and collective process goals.
4. Challenges
APM faces research challenges in framing agent autonomy, making agents explainable and actionable, enabling safe self-modification, modeling uncertainty, and addressing security and legacy-system integration.
- Framed Autonomy: A pragmatic, robust notion of an agent is needed so practitioners can understand and frame agents’ autonomy in business process execution.The notion must remain intuitively understandable to business process management practitioners while supporting process awareness.
- Framed Autonomy: Frames must be operationalized on real-world symbolic data so runtime systems can enforce compliance, explain normative requirements, and guard execution to legal actions.Formal logic with strict transition semantics and guarded execution is identified as one possible foundation.
- Explainability: APM must develop explanation mechanisms that capture changing preferences, contextual agent behavior, acquired knowledge, causal relationships, and actionable corrective options.Explanations should support autonomous action by indicating how the explanandum’s state could be altered without necessarily escalating externally.
- Conversational Actionability: Conversational actionability requires agents to communicate with human principals and other agents through natural or domain-specific languages while supporting process-related interaction.The paper identifies conversation with human principals and other agents as a distinct challenge for APM systems.
- Self-Modification: Self-modifying agents require governance, confidence-based human handoff, internal evaluation of adaptations, coordination across linked processes, and continuous learning from experience.Proposed directions include runtime monitoring, anomaly detection, performance baselines, causal reasoning, process mining, and reinforcement learning.
- Self-Modification: APM must model both aleatoric and epistemic uncertainty using probabilistic, Bayesian, fuzzy, quantitative, and qualitative approaches, including decisions about human deferral.The paper highlights hybrid methods and reliability estimates for proactive adaptation when data is sparse or ambiguous.
- Cross-cutting and Complementary Challenges: Further challenges include automating legacy BPM onboarding and protecting agentic workflows against prompt injection, compromised agents, memory poisoning, unauthorized actions, and privacy leakage.Security proposals include technical and semantic guardrails plus constrained architectural patterns such as action selection and plan-then-execute.
5. Conclusions and Outlook
The manifesto positions APM as an interdisciplinary bridge for developing BPM systems in which autonomous software agents and humans collaborate within symbolic frames. It outlines research priorities spanning formal foundations, explainability, conversational actionability, self-adaptation, and reproducible evaluation.
- APM treats intelligent business process execution as collaboration between autonomous software agents and humans.
- APM agents require symbolic frames, while LLMs alone are insufficient for truly intelligent BPM systems.
- The manifesto bridges BPM, autonomous agents, and machine learning while responding to the growing role of LLM-based agents in BPM.
- Future research: Future research should establish formal abstractions for frames and goals, including algorithms for frame-compliant and performant operational specifications.
- Future research: APM research should develop technically sound, dynamically interpretable explanations that are actionable at both execution and process-management levels.
- Future research: Real-world deployment of concurrently evolving, process-aware, self-adaptive agents requires governance and systematic assessment of adaptation consequences.