Artificial intelligence (AI) has been gaining traction in business, with many ERP and RPA tools incorporating AI capabilities. Traditionally, optical character recognition (OCR) has been used to digitise accounting tasks by scanning source documents and feeding the raw data into the accounting system, but AI now offers the potential to significantly improve this process. This research investigates the impact of document design on the accuracy of AI-driven accounting data extraction. Using invoices as a representative source document, I examine how visual complexity, colour choices, and background characteristics influence the performance of the Gemini AI platform. While AI demonstrates considerable potential in automating accounting tasks, my research underscores the importance of considering source document design factors that facilitate or hinder the automation of data extraction using AI. This study contributes to the ongoing development of AI in accounting and auditing by highlighting the interplay between document design and AI performance.
Jacob Peng (Thu,) studied this question.