INTEGRATION OF SMALL LANGUAGE MODELS FOR ACCOUNTING AUTOMATION IN AN ERP

Abstract

This thesis explores the application of Small Language Models (SLMs) for accounting automation, namely in invoice data extraction and account reconciliation. The purpose of this thesis is to evaluate the feasibility of using SLMs for these tasks and to implement a prototype tool that is integrated into an ERP system. This thesis follows the design science research methodology to develop and evaluate the SLM-based tools. This thesis compares several different SLMs with different prompts to find the most reliable and accurate model and prompt. Testing is conducted with synthesized exploratory data. Results show that certain SLMs are able to achieve these functionalities with decent accuracy and inference time. This suggests that SLMs are a viable and efficient solution for accounting automation in ERP systems.