Agentic AI: Formal AGO framework for business process analysis

A Formal Framework for Declarative Agentic AI in Business Process Analysis was submitted to arXiv on 13 Jun 2026 by Mohammad Azarijafari, Luisa Mich and Michele Missikoff (arXiv:2606.15291). The paper presents a formal framework for Agentic business process analysis built around the AGO methodology and organises its definitions into a Business Process Knowledge Base, or BPKB.

What the paper proposes

The AGO methodology captures three modelling perspectives: who is acting (Agents), why an action is carried out (Goals), and what the relevant entities are (Objects). The authors ground these AGO entity types and their interactions in set theory and mathematical logic. They formalise the entity types, organise all definitions into a BP Knowledge Base, and show how the BPKB supports structured querying, incremental updates, and automatic generation of business process workflows.

The paper emphasises formal precision as a prerequisite for agentic automation of business processes. It positions Agentic AI as enabling autonomous decision-making and dynamic adaptation in Business Process settings, and argues that achieving those capabilities requires BP entities and their interactions to be defined with formal rigor.

How the AGO framework works

The framework starts by defining AGO entity types using set-theoretic and logical constructs. Those formal definitions and the rules that connect them are stored in the BP Knowledge Base. The BPKB acts as the structured repository from which queries can be executed and from which process paths can be derived. The authors highlight three capabilities that arise from this structure: structured querying of the BP domain, incremental updates to the knowledge base as processes or entities change, and automatic generation of BP workflows from the formalised definitions.

The paper frames these capabilities as ensuring soundness and completeness of the derived paths. Organising definitions into the BPKB lets the system derive workflows algorithmically while preserving those formal properties, according to the abstract.

Paper metadata and availability

The submission appears as arXiv:2606.15291 [cs.AI] and carries an arXiv-issued DOI via DataCite pending registration. The authors supplied a PDF and TeX source on the paper page, which also lists links and toggles for associated code, data and media tools and demos.

Why it matters

Agentic AI opens new opportunities for automating Business Process, enabling autonomous decision-making and dynamic adaptation, the authors write. Without formal definitions of the entities and relations involved, those agentic behaviours risk ambiguity or incorrect automation. By providing a set-theoretic, logic-based formalisation and a single BP Knowledge Base, the paper targets the gap between conceptual agent designs and mechanically verifiable process derivation. If adopted, that approach could make it easier to query, update and generate process workflows in environments that need provable properties such as soundness and completeness.

What to watch

Watch the paper's arXiv entry for follow-up versions and for the DataCite DOI registration. Also monitor the linked code and data sections on the paper page for any published tools or demos that implement the AGO definitions and the BP Knowledge Base.