BPMN DMN decision table generation based on agentic AI for critical applications

Overview

This study examines the automated generation of decision model and notation (DMN) decision tables from business process model and notation (BPMN) models through agentic artificial intelligence. The research addresses limitations in traditional manual decision table creation, which introduces human error, process inconsistencies, and cognitive biases. The investigation centers on the technical feasibility and performance characteristics of an AI-driven system that autonomously parses BPMN processes, locates decision points, and produces optimized decision tables for critical operational environments.

Methods and approach

A novel agentic AI framework was developed integrating large language models within a coordinated multi-agent architecture. The system employs specialized agents for BPMN process analysis, decision point identification, rule generation, and validation tasks. The framework utilizes a ReAct engine combining reasoning and acting capabilities, augmented with retrieval-augmented generation (RAG) functionality to enhance semantic understanding and decision extraction accuracy. The agents operate within a structured workflow to transform BPMN model specifications into machine-readable DMN decision table representations.

Results

Experimental evaluation in critical application scenarios demonstrated that the system identifies decision table values and criteria not readily apparent through conventional human analysis. The framework successfully disambiguates decision paths of uncertain specification, converting implicit logic into explicit decision rules. The system demonstrated particular efficacy in extracting non-obvious decision criteria and threshold parameters that substantially improve process automation capabilities and operational precision. Generated decision tables exhibit enhanced specificity compared to manually derived equivalents.

Implications

The framework establishes foundational infrastructure for autonomous decision modeling systems capable of adapting dynamically to evolving business requirements within mission-critical operational contexts. Implementation of agentic AI for automated DMN generation enables organizations to reduce cognitive bias and inconsistency in decision management workflows. The approach supports the development of intelligent decision support systems that can autonomously maintain alignment between process models and decision logic as business conditions change.