CFA, CPA Exams Prove Valuable Skills—Regardless of AI’s Progress

Accounting, finance, and law face important questions in the age of advancing artificial intelligence capabilities. What does passing the CPA, CFA, or bar exam signal to the profession? Does it show technical proficiency or minimum competency to be a certified public accountant, chartered financial analyst, or licensed attorney? Or does it indicate skills that can’t be replicated by AI?

Those questions are becoming crucial as large language models such as ChatGPT, Gemini, and Claude have improved their ability to pass white-collar certification exams since the initial public launch of ChatGPT 3.5 in 2022.

In recent years, LLMs have gone from failing a simulated CPA exam and bar exam to passing both by comfortable margins. And LLMs now can pass all three parts of the grueling CFA exam in minutes. Over that same period, CPA and CFA candidates’ exam performance declined, despite staying within historical norms.

The LLMs are performing better while human candidates are performing worse. The underlying point of the exams will decide the impact this trend will have on accounting and finance professionals.

If the exams indicate technical proficiency, certified professionals may be at risk of near-term job losses. The CFA certification typically requires thousands of study hours over years, while the CPA exam requires hundreds of study hours—both only to achieve pass rates approximating a coin flip.

And those hours spent training and studying aren’t replicable or scalable. Unlike the training of LLMs, one employee’s certification and training over the certification period can’t be passed along to the next hire or existing analyst. They must undergo the same certification at the same time commitment subject to the same passing statistics.

By contrast, LLMs only need to be trained once, and they’ll pass the CFA exam or CPA exam every time and retain that proficiency into perpetuity. Firms could then deploy teams of managers and directors with AI agents or more novice and less credentialed staff with trained LLMs to do the work once done by a CFA or CPA.

However, if the exams merely show minimum competency, AI’s passing performance is to be expected and the small percentage points change in pass rates among human candidates are likely just statistical noise, even if the LLM won’t struggle to pass again.

Much of what the CFA and CPA exams test are basic rules, structures, and formulas—which plays to LLMs’ strengths. Even the ability of current LLMs to pass the third part of the CFA exam, with its essay components, isn’t surprising, as the writing and research skills of the models have evolved with each subsequent update or version release.

Regardless of the signal inferred by the exams, passing all three parts of the CFA exam indicates more than knowledge. At a time when AI is giving all professionals access to similar information, the credentials still serve as a differentiator.

They demonstrates perseverance, critical thinking, time management, and a willingness to learn—skills employers value despite AI’s presence. In fact, they may become even more valuable because they can’t be obtained by entering a prompt into LLMs.

For instance, Accenture announced last month that staff who can’t be retrained to use AI will be fired. I suspect other global accounting and consulting firms will follow suit. A good indicator of which employees and job applicants can be retrained to use AI is those who spent hundreds of hours studying to successfully pass the CFA or CPA exam and have earned the credential.

The AI tools in accounting and finance may have similar foundational knowledge as the employees, but the employees also will possess the skills to adapt to evolving AI models. And they will have proof that they know how to assess the foundational knowledge the AI tools presume to have.

It’s difficult to discern what is wrong with AI output if one doesn’t first know what is right. By passing an exam where over half of CFA candidates fail, those with the CFA will only stand out further.

The CPA and CFA exams are rigorous, multipart, difficult, and time-consuming. They aren’t for everyone with a finance or accounting degree. As AI handles more of the work traditionally done by entry-level staff, the certifications will set employees apart from their competition and the AI tools themselves.

The noise of discerning competency in accounting and finance is only going to get worse. College grades are inflated. Written measurements tools and even videos can be gamed with AI. So, what can set someone apart? Those three capital letters behind their name.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law, Bloomberg Tax, and Bloomberg Government, or its owners.

Author Information

Jack Castonguay is a CPA, associate professor of accounting at Hofstra University, and vice president of content development at KnowFully Learning Group.

Write for Us: Author Guidelines