5 papers

Business Process

Agentic BPM: goal-driven process design, process mining, workflow graphs, and governing non-deterministic work.

Papers
5
Read time
1.4h
Avg length
16m
01
Business Process

Agent Behavior Mining: Generative AI Agent Governance in Business Processes

Reformat every LLM agent's tool call, hand-off, and chain-of-thought into a standard XES event log — the file format process-mining tools have read for 15 years — so you can point ordinary process-mining software at a swarm of non-deterministic agents and get back the same discovery/conformance/cost dashboards you'd get from a deterministic workflow.

HOANG VU, MAXIMILIAN KÖRNER, ADRIAN REBMANN, GABRIEL KEVORKIAN, MICHAEL PERSCHEID, GREGOR BERG, TIMOTHEUS KAMPIK (SAP RESEARCH / UMEÅ UNIVERSITY) · 2026
Read
16 min
Source
arXiv:2606.20669
02
Business Process

Agentic Business Process Management Systems

Business Process Management is shifting from *automation* (systems that execute pre-scripted rules) to *autonomy* (agents that sense, reason, and act beyond the script) — and this paper sketches the five-layer architecture, the autonomy spectrum, and the agent-orchestration patterns a BPM platform needs to actually support that shift.

MARLON DUMAS, FREDRIK MILANI, DAVID CHAPELA-CAMPA (UNIVERSITY OF TARTU, ESTONIA) · AI4BPM WORKSHOP 2025 KEYNOTE
Read
13 min
Source
arXiv:2601.18833
03
Business Process

Agentic Business Process Management: A Research Manifesto

Stop treating LLM agents as fancy task-runners bolted onto a workflow diagram — treat the agents themselves as the process, and build a "frame" around each one so its autonomy stays pointed at the organization's goals instead of its own.

CALVANESE, DE GIACOMO, DUMAS, FOURNIER, KAMPIK, ET AL. (18 AUTHORS, 15 INSTITUTIONS INCL. OXFORD, IBM RESEARCH, SAP, META) · DAGSTUHL SEMINAR 25192 / PMAI@ECAI 2025
Read
20 min
Source
arXiv:2603.18916
04
Business Process

An Agentic AI for a New Paradigm in Business Process Development

Instead of hard-wiring a business process as a fixed sequence of tasks, describe it as a set of *goals*, the *information objects* that prove each goal is met, and the autonomous *agents* that produce those objects — and let the runnable workflow emerge from which objects are ready.

MOHAMMAD AZARIJAFARI, LUISA MICH, MICHELE MISSIKOFF (UNIVERSITY OF TRENTO / CNR-IASI) · ITAL-IA 2025
Read
14 min
Source
arXiv:2507.21823
05
Business Process

Business World Model

Give an AI agent a live, structured map of your company — the entities it runs on, the ML models that predict how those entities move, and the short list of levers you can actually pull — so it can simulate several campaigns and their side effects before you run one, instead of just executing the campaign you already typed out.

CECIL PANG, HIROKI SAYAMA · ARXIV PREPRINT, 2026
Read
18 min
Source
arXiv:2606.10044