A recent study by Mercor reveals that current AI models outperform licensed CPAs in both speed and accuracy when handling structured accounting tasks. According to The Decoder, this marks significant progress compared to eighteen months ago, when AI lagged far behind human accountants.
Despite these advancements, The Decoder reports that no AI model can yet fully complete all tasks on the APEX Benchmark independently. Human supervision remains essential to close the books and ensure accuracy in final accounting processes.
For Japanese markets, where precision and regulatory compliance are critical, the integration of AI into accounting workflows could enhance efficiency but will still rely heavily on expert oversight to meet stringent standards.
