Mercor Study: AI Outpaces Accountants in Speed
Alex da Cruz
Alex da Cruz is a full-stack developer based in São Paulo, Brazil. He works with React, TypeScript and automation, and uses AI daily to solve real problems in code and operations — not as a demo. He has run an e-commerce operation end to end, and now builds and maintains the automation pipeline behind this blog. He writes about what he actually tests.
According to a study by Mercor cited by The Decoder, modern AI models now surpass licensed CPAs in speed and precision on structured bookkeeping duties. Just eighteen months ago, top models scored well below the average accountant benchmark of 37 percent. Today, they complete those exact tasks almost flawlessly.
How do models perform on complex tasks?
On the broader APEX Accounting Benchmark, which spans 160 tasks across 10 simulated companies, performance drops significantly. Claude Opus 5.5 leads the leaderboard by meeting 61.8 percent of grading criteria, followed by Fable 5.1 at 61.0 percent and GPT-6 Astra at 57.9 percent. These figures show that no current model solves even 60 percent of the full test without errors.
What changes for finance teams on Monday morning?
AI excels at hunting down details and following strict instructions, making it ideal for repetitive data sorting. Yet, the study omits crucial aspects of financial roles, such as client communication, team collaboration, and years of contextual judgment. Accountants cannot be replaced yet, but finance workflows will shift toward review and supervision rather than manual entry.
Sources
Frequently asked questions
- Can AI replace accountants completely?
- No. While AI outperforms humans in structured bookkeeping tasks, it cannot close the books independently or handle tasks requiring contextual judgment.
- Which model leads the APEX Accounting Benchmark?
- Claude Opus 5.5 currently leads the benchmark with 61.8 percent of grading criteria met, according to the Mercor study.
Comments
0 comments
Be the first to comment.
Continue Lendo
Meta Muse AI Agent: Workflows, Pricing, and Security Controls
Meta enters autonomous AI execution with Muse. Here is how its API connectors, pricing tiers, and virtual security controls function.
Suno Speech: how to generate voiceovers with AI music
Suno has launched a Speech feature in beta, allowing creators to generate synthetic voiceovers and background music simultaneously.
Microsoft MAI Models: Voice Agents with 100ms Latency
Microsoft released new transcription and voice models with 100ms latency and voice cloning capabilities for AI agents.