Human + AI in Accounting: Early Evidence from the Field

The question of how artificial intelligence (AI) will reshape accounting is critical to the economy and has far-reaching implications for practitioners, regulators, and academics. The advent of generative AI stands to revolutionize accounting processes and their resultant outcomes. According to a recent report by Gartner (2024), the adoption rate of generative AI in accounting was 5% in 2023, but is expected to increase rapidly in the coming years. The introduction of Generative AI to accounting has two main potential areas of significant consequence. First, Generative AI can significantly transform financial reporting quality by automating routine tasks, accelerating complex data analysis, and enhancing reporting accuracy. Second, as a result, Generative AI can significantly transform accounting careers both on the intensive margin (e.g. a shift in type of tasks) and on the extensive margin (e.g. a shift in labor demand and supply). The goal of our research project is to evaluate these two intertwined topics in tandem through proprietary field data obtained from our Partner Firm. The specific research questions in this project include: 1) Whether and to what extent does Generative AI affect financial reporting quality, 2) What is the quantifiable efficiency gain in accounting processes assisted by Generative AI, 3) What is the impact of Generative AI on the development and career satisfaction of accountants, and 4) Whether Generative AI produces accounting information that enhances firm operations.
An important addition to the paper is that we are conducting a novel framed field experiment. We plan on recruiting approximately 300 accountants to accomplish a series of accounting tasks (e.g. categorize transactions, accounts reconciliation, etc.) using accounting software that may be assisted by AI. We use this opportunity to corroborate many findings that we glean in the proprietary field data from our Partner Firm. The value of this test is two fold: 1) it resolves some element of the endogeneity problem (i.e. the accounting firms in our study are selecting to use AI; whereas the recruited participants in the framed field experiment has treatment v. control) 2) it allows us to understand the underlying mechanisms of outcomes pertaining to productivity, errors, interaction, etc.