From Business Events to Auditable Decisions: Ontology-Governed Graph Simulation for Enterprise AI

TL;DR

This paper introduces LOM-action, an ontology-driven simulation framework for enterprise AI that ensures decisions are grounded, auditable, and more accurate than existing large language model-based systems.

Contribution

It presents a novel event-driven ontology simulation approach that improves decision trustworthiness and auditability in enterprise AI systems.

Findings

Achieves 93.82% accuracy and 98.74% F1 on benchmark datasets.

Outperforms frontier baselines with a four-fold F1 advantage.

Demonstrates that ontology-guided simulation is key to trustworthy enterprise decisions.

Abstract

Existing LLM-based agent systems share a common architectural failure: they answer from the unrestricted knowledge space without first simulating how active business scenarios reshape that space for the event at hand -- producing decisions that are fluent but ungrounded and carrying no audit trail. We present LOM-action, which equips enterprise AI with \emph{event-driven ontology simulation}: business events trigger scenario conditions encoded in the enterprise ontology~(EO), which drive deterministic graph mutations in an isolated sandbox, evolving a working copy of the subgraph into the scenario-valid simulation graph $G_{\text{sim}}$; all decisions are derived exclusively from this evolved graph. The core pipeline is \emph{event $\to$ simulation $\to$ decision}, realized through a dual-mode architecture -- \emph{skill mode} and \emph{reasoning mode}. Every decision produces a fully traceable audit log. LOM-action achieves 93.82% accuracy and 98.74% tool-chain F1 against frontier baselines Doubao-1.8 and DeepSeek-V3.2, which reach only 24--36% F1 despite 80% accuracy -- exposing the \emph{illusive accuracy} phenomenon. The four-fold F1 advantage confirms that ontology-governed, event-driven simulation, not model scale, is the architectural prerequisite for trustworthy enterprise decision intelligence.