A new study conducted by Mercor indicates that current artificial intelligence models can now outperform licensed certified public accountants on structured accounting tasks, beating human professionals in both speed and accuracy. This represents a significant shift from eighteen months prior, when technology of this nature still lagged notably behind human benchmarks.
The Limits of Current Financial Automation
Despite these gains on structured tasks, the technology faces limitations when applied to more complex workflows. On the APEX Benchmark, which presents more demanding challenges, current models fail to complete all required tasks. Consequently, artificial intelligence remains incapable of closing financial books independently without human supervision.
Unanswered Questions and Future Oversight
The announcement did not specify which particular AI models were tested during the Mercor study, nor did it detail the exact metrics used to measure speed and accuracy against human CPAs. Observers will need to watch whether upcoming benchmark iterations address the autonomy gap in financial reporting.
While structured data processing has clearly accelerated, human accountants remain a necessary fixture for complete financial oversight.
Lead image: Mikhail Nilov / Pexels
