AI Crushes a Finance Exam Most Humans Fail. Are Analysts Next?
▼ Summary
– Frontier AI models successfully passed a mock version of the highly difficult Chartered Financial Analyst (CFA) Level III exam.
– OpenAI’s o4-mini and Google’s Gemini 2.5 Flash were the top-performing models, scoring 79.1% and 77.3% respectively, above the 63% passing threshold.
– The exam is particularly challenging because it tests higher-order cognitive skills like analysis and professional judgment through both multiple-choice and essay questions.
– While most models performed similarly on the multiple-choice section, their scores varied significantly on the more complex essay portion, highlighting a difference in reasoning capabilities.
– Despite this advancement, an expert notes that human financial advisors still excel in areas requiring context and intent, such as interpreting body language.
The Chartered Financial Analyst (CFA) Level III exam, a notoriously difficult benchmark for investment professionals, has been successfully passed by several advanced artificial intelligence models. This development signals a significant leap in AI’s ability to handle complex financial reasoning, a domain once considered a uniquely human forte. The exam, which fewer than half of human candidates passed in its most recent sitting, demands more than simple memorization, testing skills like analysis, synthesis, and professional judgment in portfolio management.
A recent study conducted by New York University’s Stern School of Business in collaboration with GoodFin evaluated 23 leading AI models. The research focused on a mock version of the Level III exam, which uniquely combines multiple-choice questions with essay components designed to simulate real-world wealth planning scenarios. While AI had previously conquered the earlier levels of the CFA program, the third level had remained a substantial hurdle due to its emphasis on applied knowledge and nuanced judgment.
The results revealed that a select group of reasoning models, which excel at deconstructing complex problems into manageable steps, achieved passing scores. OpenAI’s o4-mini led the pack with a score of 79.1%, followed closely by Google’s Gemini 2.5 Flash at 77.3%, both comfortably exceeding the 63% passing threshold. Interestingly, while most models performed similarly on the multiple-choice section, their capabilities diverged significantly on the essay portion. This indicates that straightforward tasks have become standardized across AI, but sophisticated reasoning still separates the most advanced systems from the rest.
This achievement raises inevitable questions about the future of financial analysts. A Microsoft report has already flagged personal financial advisors as a role with high exposure to AI automation. However, industry experts suggest that immediate replacement is unlikely. The founder of GoodFin, Anna Joo Fee, points out that machines currently struggle with assessing human context and intent. The nuanced understanding of client body language and subtle cues remains an area where human professionals continue to hold a distinct advantage. The technology is rapidly advancing, but for now, the most effective approach likely involves a collaboration between human expertise and AI’s computational power.
(Source: ZDNET)
