Formal Foundations of Agentic Business Process Management

Christoph Weinhuber; Giuseppe De Giacomo; Lukas Kirchdorfer; Marco Montali; Timotheus Kampik

arxiv: 2604.17347 · v1 · submitted 2026-04-19 · 💻 cs.AI

Formal Foundations of Agentic Business Process Management

Giuseppe De Giacomo , Timotheus Kampik , Lukas Kirchdorfer , Marco Montali , Christoph Weinhuber This is my paper

Pith reviewed 2026-05-10 06:38 UTC · model grok-4.3

classification 💻 cs.AI

keywords agentic BPMformal foundationsautonomous agentsagent goalsprocess specificationsstrategy guardrailsbusiness process managementmulti-agent systems

0 comments

The pith

Agentic business process management receives formal mathematical foundations by augmenting process specifications with explicit goals for autonomous agents.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that traditional BPM must adapt when execution shifts from centralized control to multiple autonomous agents who make their own decisions. It augments standard process specifications with explicit objectives assigned to each agent, then examines how those agents pursue the objectives under assumptions about others' behavior. This setup lets the enacting organization impose guardrails directly on agent strategies rather than trying to dictate every action. A sympathetic reader would care because it turns potentially unpredictable agentic systems into something that can be formally analyzed and constrained in three standard settings while addressing four core problems.

Core claim

This paper sets up the mathematical foundations of agentic BPM systems in three key settings and analyzes four foundational problems of agentic BPM. The distinguishing feature is that execution is driven by autonomous decision-makers, so the process specification is augmented with explicit objectives assigned to the participating agents. Agents then pursue these goals, at least to the best of their efforts, under suitable assumptions on the behavior of others, by adopting appropriate strategies. Centrally, the organization enacting the process can use these specifications to provide guardrails on the decision-making capabilities of agents at the strategy level.

What carries the argument

Goal-augmented process specifications that assign explicit objectives to autonomous agents and enable organizations to impose strategy-level guardrails.

If this is right

Organizations gain the ability to constrain agent decision-making at the level of strategies rather than individual actions.
Four foundational problems of agentic BPM receive formal treatment across three standard settings.
Process execution remains reliable even when agents act autonomously, provided their objectives are explicitly specified.
Guardrails derived from the specifications can be enforced without assuming complete control over every agent choice.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The framework could be implemented in enterprise software to automatically translate high-level goals into enforceable constraints on deployed AI agents.
It opens a route to verifying compliance in multi-agent business workflows by checking whether observed agent strategies stay inside the formally defined guardrails.
Extensions might combine these foundations with existing workflow engines to simulate and validate agent behaviors before live deployment.
The same approach could apply to regulatory oversight of autonomous systems in other domains where full control is impossible.

Load-bearing premise

Agents cannot be fully controlled, so the process specification must include explicit objectives that agents pursue under assumptions about how others will behave.

What would settle it

A concrete counter-example in which autonomous agents operating in a real business process systematically violate the assumed strategies or fail to respect the guardrails derived from the three formal settings.

Figures

Figures reproduced from arXiv: 2604.17347 by Christoph Weinhuber, Giuseppe De Giacomo, Lukas Kirchdorfer, Marco Montali, Timotheus Kampik.

**Figure 2.** Figure 2: The travel expense process, executed by three agents (indicated by the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

read the original abstract

Just like traditional BPM systems, agentic BPM systems are built around a specification of the process under consideration. Their distinguishing feature, however, is that the execution of the process is driven by multiple autonomous decision-makers, referred to as agents. Since such agents cannot be fully controlled, the process specification is augmented with explicit objectives, or goals, assigned to the participating agents. Agents then pursue these goals, at least to the best of their efforts, under suitable assumptions on the behavior of others, by adopting appropriate strategies. Centrally, the organization enacting the process can use these specifications to provide guardrails on the decision-making capabilities of agents at the strategy level. This paper sets up the mathematical foundations of such systems in three key settings and analyzes four foundational problems of agentic BPM.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper sets up a formal modeling framework for agentic BPM by layering agent goals and strategies onto process specs, but it stays mostly definitional without new theorems or concrete solutions.

read the letter

The core contribution here is taking ordinary process specifications and extending them with explicit objectives assigned to autonomous agents, who then select strategies while making assumptions about how others will behave. They formalize this in three settings and flag four foundational problems, with the organization using the specs mainly to set guardrails at the strategy level rather than direct control. This is a straightforward and consistent way to handle the autonomy issue, and it draws on standard multi-agent ideas without obvious circularity or invented entities that don't fit. It does a reasonable job of organizing the space so that guardrails become part of the model rather than an afterthought. That framing is useful for anyone trying to think systematically about agents inside business processes. The main limitation is that the work reads as setup and problem identification rather than resolution. There is no sign of detailed derivations, proofs, or worked examples that would show the models actually constrain behavior in useful ways or solve the four problems they name. Without that, it is hard to judge how much new leverage the formalization provides over existing BPM or agent formalisms. The math appears to rest on conventional assumptions, so soundness is not obviously broken, but the absence of deeper analysis keeps the advance modest. This is for researchers who work at the intersection of formal methods, BPM, and multi-agent systems. A reader already thinking about agentic processes could pick up the structured vocabulary and the list of open problems. It is not yet something that would change how people build systems. I would send it to peer review so referees can check whether the three settings contain any non-trivial results and whether the four problems are treated with enough rigor to justify the foundations claim.

Referee Report

0 major / 2 minor

Summary. The paper claims to establish mathematical foundations for agentic Business Process Management (BPM) systems, which extend traditional BPM by having processes executed by multiple autonomous agents. Process specifications are augmented with explicit objectives assigned to agents; agents pursue these goals via strategies under assumptions about others' behavior. Organizations can impose guardrails at the strategy level. The work defines these foundations in three key settings and analyzes four foundational problems of agentic BPM.

Significance. If the formalizations prove rigorous and the analyses of the four problems are substantive, the paper could offer a useful definitional bridge between BPM and multi-agent systems, clarifying how autonomy and guardrails can be modeled without full control over agents. This aligns with standard game-theoretic and strategic modeling approaches and may support future work on verifiable agentic processes.

minor comments (2)

The abstract and high-level description do not specify the three key settings or the four foundational problems; including brief characterizations or section references in the introduction would improve clarity for readers.
Notation for agent objectives, strategies, and guardrails should be introduced with explicit definitions and examples early in the manuscript to aid comprehension of the subsequent formal development.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of the paper and for recommending minor revision. The assessment that the work provides a useful definitional bridge between BPM and multi-agent systems is appreciated. No specific major comments were listed in the report, so we have no point-by-point responses to provide at this stage. We remain ready to make any minor adjustments required for the revised version.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper establishes mathematical foundations for agentic BPM by defining process specifications augmented with explicit agent objectives and strategies pursued under assumptions about others' behavior. This construction draws from standard multi-agent system formalisms and BPM concepts without any derivation that reduces to its own inputs by construction. No self-definitional loops, fitted parameters renamed as predictions, load-bearing self-citations, uniqueness theorems imported from prior author work, or smuggled ansatzes are present in the provided abstract and high-level description. The work is positioned as a definitional setup and analysis of four foundational problems across three settings, remaining self-contained against external benchmarks in formal methods.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The ledger is populated based on the abstract since the full manuscript text was not accessible for detailed analysis.

axioms (1)

domain assumption Agents pursue assigned goals to the best of their efforts under suitable assumptions on the behavior of others.
This is stated as the basis for how agents operate in the system.

invented entities (1)

agentic BPM systems no independent evidence
purpose: To model business processes driven by autonomous agents with goals and guardrails.
Introduced as the core concept distinguishing from traditional BPM.

pith-pipeline@v0.9.0 · 5433 in / 1337 out tokens · 57074 ms · 2026-05-10T06:38:29.094689+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

[1]

van der Aalst, W.M.P., Lohmann, N., La Rosa, M.: Ensuring correctness during process configuration via partner synthesis. Inf. Syst.37(6), 574–592 (2012)

work page 2012
[2]

In: ICAPS

Aminof, B., De Giacomo, G., Murano, A., Rubin, S.: Planning under LTL envi- ronment specifications. In: ICAPS. pp. 31–39 (2019)

work page 2019
[3]

Aminof, B., De Giacomo, G., Perelli, G., Rubin, S.: LTL Synthesis Under Multi- Agent Environment Assumptions. In: KR. pp. 719–728 (2025)

work page 2025
[4]

In: IJCAI

Aminof, B., De Giacomo, G., Rubin, S.: Best-effort synthesis: Doing your best is not harder than giving up. In: IJCAI. pp. 1766–1772 (2021)

work page 2021
[5]

MIT Press (2008)

Baier, C., Katoen, J.: Principles of Model Checking. MIT Press (2008)

work page 2008
[6]

In: BPM Workshops

Barros, A., Dumas, M., Oaks, P.: Standards for web service choreography and orchestration: Status and perspectives. In: BPM Workshops. pp. 61–74 (2005)

work page 2005
[7]

In: STACS

Berwanger, D.: Admissibility in infinite games. In: STACS. pp. 188–199 (2007)

work page 2007
[8]

In: ECAI Workshops (2025)

Calvanese, D., De Giacomo, G., Kampik, T., Lesperance, Y., Marrella, A., Matta, A.: Autonomy in business process execution: why we need first-class abstractions for goals and normative frames. In: ECAI Workshops (2025)

work page 2025
[9]

In: AAAI

De Giacomo, G., De Masellis, R., Montali, M.: Reasoning on LTL on finite traces: Insensitivity to infiniteness. In: AAAI. pp. 1027–1033. AAAI Press (2014)

work page 2014
[10]

In: IJCAI

De Giacomo, G., Vardi, M.Y.: Linear Temporal Logic and Linear Dynamic Logic on finite traces. In: IJCAI. pp. 854–860 (2013)

work page 2013
[11]

In: IJCAI

De Giacomo, G., Vardi, M.Y.: Synthesis for LTL and LDL on finite traces. In: IJCAI. pp. 1558–1564 (2015)

work page 2015
[12]

In: Process Mining Handbook, LNBIP, vol

Di Ciccio, C., Montali, M.: Declarative process specifications: Reasoning, discovery, monitoring. In: Process Mining Handbook, LNBIP, vol. 448 (2022)

work page 2022
[13]

In: CIAA

Duret-Lutz, A., Zhu, S., Piterman, N., De Giacomo, G., Vardi, M.Y.: Engineering an ltlf synthesis tool. In: CIAA. pp. 129–147 (2025)

work page 2025
[14]

In: BPM (Forum)

Geatti, L., Montali, M., Rivkin, A.: Foundations of collaborative declare. In: BPM (Forum). pp. 55–72 (2023)

work page 2023
[15]

Haarmann, S., Weske, M.: Correlating data objects in fragment-based case man- agement. In: BIS. LNBIP, vol. 389, pp. 197–209. Springer (2020)

work page 2020
[16]

In: ToPNoC VII

van Hee, K.M., Sidorova, N., van der Werf, J.M.: Business process modeling using petri nets. In: ToPNoC VII. pp. 116–161 (2013)

work page 2013
[17]

Heindel, T., Weber, I.: Incentive alignment of business processes. In: BPM. pp. 93–110 (2020)

work page 2020
[18]

In: FORTE

Hildebrandt, T.T., Slaats, T., L´ opez, H.A., Debois, S., Carbone, M.: Declarative choreographies and liveness. In: FORTE. pp. 129–147 (2019) Formal Foundations of Agentic BPM 17

work page 2019
[19]

Massuthe, P., Serebrenik, A., Sidorova, N., Wolf, K.: Can I find a partner? unde- cidability of partner existence for open nets. Inf. Process. Lett.108(6), 374–378 (2008)

work page 2008
[20]

Montali, M., Torroni, P., Zannone, N., Mello, P., Bryl, V.: Engineering and verify- ing agent-oriented requirements augmented by business constraints withB-tropos. Auton. Agents Multi Agent Syst.23(2), 193–223 (2011)

work page 2011
[21]

Ruiz, M., Costal, D., Espa˜ na, S., Franch, X., Pastor, O.: Gobis: An integrated framework to analyse the goal and business process perspectives in information systems. Inf. Syst.53, 330–345 (2015)

work page 2015
[22]

Tour, A., Polyvyanyy, A., Kalenkova, A.A., Senderovich, A.: Agent miner: An algorithm for discovering agent systems from event data. In: BPM. pp. 284–302 (2023)

work page 2023
[23]

In: Responsible BPM Forum

Vu, H., Klievtsova, N., Leopold, H., Rinderle-Ma, S., Kampik, T.: Agentic business process management: Practitioner perspectives on agent governance in business processes. In: Responsible BPM Forum. pp. 29–43 (2026)

work page 2026
[24]

Springer (2024)

Weske, M.: Business Process Management: Concepts, Languages, Architectures. Springer (2024)

work page 2024
[25]

John Wiley & Sons, Chichester, UK, 2 edn

Wooldridge, M.: An Introduction to MultiAgent Systems. John Wiley & Sons, Chichester, UK, 2 edn. (2009)

work page 2009

[1] [1]

van der Aalst, W.M.P., Lohmann, N., La Rosa, M.: Ensuring correctness during process configuration via partner synthesis. Inf. Syst.37(6), 574–592 (2012)

work page 2012

[2] [2]

In: ICAPS

Aminof, B., De Giacomo, G., Murano, A., Rubin, S.: Planning under LTL envi- ronment specifications. In: ICAPS. pp. 31–39 (2019)

work page 2019

[3] [3]

Aminof, B., De Giacomo, G., Perelli, G., Rubin, S.: LTL Synthesis Under Multi- Agent Environment Assumptions. In: KR. pp. 719–728 (2025)

work page 2025

[4] [4]

In: IJCAI

Aminof, B., De Giacomo, G., Rubin, S.: Best-effort synthesis: Doing your best is not harder than giving up. In: IJCAI. pp. 1766–1772 (2021)

work page 2021

[5] [5]

MIT Press (2008)

Baier, C., Katoen, J.: Principles of Model Checking. MIT Press (2008)

work page 2008

[6] [6]

In: BPM Workshops

Barros, A., Dumas, M., Oaks, P.: Standards for web service choreography and orchestration: Status and perspectives. In: BPM Workshops. pp. 61–74 (2005)

work page 2005

[7] [7]

In: STACS

Berwanger, D.: Admissibility in infinite games. In: STACS. pp. 188–199 (2007)

work page 2007

[8] [8]

In: ECAI Workshops (2025)

Calvanese, D., De Giacomo, G., Kampik, T., Lesperance, Y., Marrella, A., Matta, A.: Autonomy in business process execution: why we need first-class abstractions for goals and normative frames. In: ECAI Workshops (2025)

work page 2025

[9] [9]

In: AAAI

De Giacomo, G., De Masellis, R., Montali, M.: Reasoning on LTL on finite traces: Insensitivity to infiniteness. In: AAAI. pp. 1027–1033. AAAI Press (2014)

work page 2014

[10] [10]

In: IJCAI

De Giacomo, G., Vardi, M.Y.: Linear Temporal Logic and Linear Dynamic Logic on finite traces. In: IJCAI. pp. 854–860 (2013)

work page 2013

[11] [11]

In: IJCAI

De Giacomo, G., Vardi, M.Y.: Synthesis for LTL and LDL on finite traces. In: IJCAI. pp. 1558–1564 (2015)

work page 2015

[12] [12]

In: Process Mining Handbook, LNBIP, vol

Di Ciccio, C., Montali, M.: Declarative process specifications: Reasoning, discovery, monitoring. In: Process Mining Handbook, LNBIP, vol. 448 (2022)

work page 2022

[13] [13]

In: CIAA

Duret-Lutz, A., Zhu, S., Piterman, N., De Giacomo, G., Vardi, M.Y.: Engineering an ltlf synthesis tool. In: CIAA. pp. 129–147 (2025)

work page 2025

[14] [14]

In: BPM (Forum)

Geatti, L., Montali, M., Rivkin, A.: Foundations of collaborative declare. In: BPM (Forum). pp. 55–72 (2023)

work page 2023

[15] [15]

Haarmann, S., Weske, M.: Correlating data objects in fragment-based case man- agement. In: BIS. LNBIP, vol. 389, pp. 197–209. Springer (2020)

work page 2020

[16] [16]

In: ToPNoC VII

van Hee, K.M., Sidorova, N., van der Werf, J.M.: Business process modeling using petri nets. In: ToPNoC VII. pp. 116–161 (2013)

work page 2013

[17] [17]

Heindel, T., Weber, I.: Incentive alignment of business processes. In: BPM. pp. 93–110 (2020)

work page 2020

[18] [18]

In: FORTE

Hildebrandt, T.T., Slaats, T., L´ opez, H.A., Debois, S., Carbone, M.: Declarative choreographies and liveness. In: FORTE. pp. 129–147 (2019) Formal Foundations of Agentic BPM 17

work page 2019

[19] [19]

Massuthe, P., Serebrenik, A., Sidorova, N., Wolf, K.: Can I find a partner? unde- cidability of partner existence for open nets. Inf. Process. Lett.108(6), 374–378 (2008)

work page 2008

[20] [20]

Montali, M., Torroni, P., Zannone, N., Mello, P., Bryl, V.: Engineering and verify- ing agent-oriented requirements augmented by business constraints withB-tropos. Auton. Agents Multi Agent Syst.23(2), 193–223 (2011)

work page 2011

[21] [21]

Ruiz, M., Costal, D., Espa˜ na, S., Franch, X., Pastor, O.: Gobis: An integrated framework to analyse the goal and business process perspectives in information systems. Inf. Syst.53, 330–345 (2015)

work page 2015

[22] [22]

Tour, A., Polyvyanyy, A., Kalenkova, A.A., Senderovich, A.: Agent miner: An algorithm for discovering agent systems from event data. In: BPM. pp. 284–302 (2023)

work page 2023

[23] [23]

In: Responsible BPM Forum

Vu, H., Klievtsova, N., Leopold, H., Rinderle-Ma, S., Kampik, T.: Agentic business process management: Practitioner perspectives on agent governance in business processes. In: Responsible BPM Forum. pp. 29–43 (2026)

work page 2026

[24] [24]

Springer (2024)

Weske, M.: Business Process Management: Concepts, Languages, Architectures. Springer (2024)

work page 2024

[25] [25]

John Wiley & Sons, Chichester, UK, 2 edn

Wooldridge, M.: An Introduction to MultiAgent Systems. John Wiley & Sons, Chichester, UK, 2 edn. (2009)

work page 2009