FLOW-BENCH: Towards Conversational Generation of Enterprise Workflows

TL;DR

This paper addresses enterprise workflow automation by translating natural language instructions into structured BPMN and DMN artifacts using large language models. It introduces FLOW-BENCH, a high-quality NL-to-definition dataset, and FLOW-GEN, an intermediate representation–driven pipeline that converts NL into a Python-like IR suitable for mapping to BPMN/DMN. The approach grounds NL instructions in contextual API documentation and leverages in-context learning to mitigate hallucinations and enhance domain adaptability across eight LLMs of varying sizes. The work provides resources and a practical translation framework to democratize BPA and catalyze further research in the field.

Abstract

Business process automation (BPA) that leverages Large Language Models (LLMs) to convert natural language (NL) instructions into structured business process artifacts is becoming a hot research topic. This paper makes two technical contributions -- (i) FLOW-BENCH, a high quality dataset of paired natural language instructions and structured business process definitions to evaluate NL-based BPA tools, and support bourgeoning research in this area, and (ii) FLOW-GEN, our approach to utilize LLMs to translate natural language into an intermediate representation with Python syntax that facilitates final conversion into widely adopted business process definition languages, such as BPMN and DMN. We bootstrap FLOW-BENCH by demonstrating how it can be used to evaluate the components of FLOW-GEN across eight LLMs of varying sizes. We hope that FLOW-GEN and FLOW-BENCH catalyze further research in BPA making it more accessible to novice and expert users.

Figures (4)

• Figure 4: Simple flow example
• Figure 5: BPMN code corresponding to example in Figure \ref{['fig:flow1']}
• Figure 6: BPMN code corresponding to the prior sequence of Figure \ref{['fig:flowbench_test']}
• Figure 7: BPMN code corresponding to the final output of Figure \ref{['fig:flowbench_test']}