ChatGPT Beats Humans at Financial Statement Analysis
As accounting professionals, one of our core skills is analyzing financial statements to assess a company's performance and predict future earnings. It's difficult, time-consuming work. But that might be changing.
Researchers at the University of Chicago found that a ChatGPT model, with some fine-tuning, could predict changes in company earnings with over 60% accuracy using only historical financial statements. The model outperformed human analysts, who had just under 53% accuracy.
If you've ever done serious financial statement analysis, you know how much work it is. Gathering historical data, inputting it into spreadsheets, and calculating ratios and KPIs takes a lot of work. It is stunning to see AI handling and doing part of this task better than humans.
Innovations in Financial Analysis
I saw another example of this at last week's Institute of Management Accountants conference. The team at Your Part-Time Controller demoed a new AI-powered tool they built to analyze financial statements for their clients, who are mostly nonprofits. The tool pulls in a client's financials, sends the data securely to OpenAI, and automatically generates detailed narratives and variance analyses to help CFOs and controllers explain the numbers to their non-financially savvy board members. It translates the financials into stories, which is exactly what's needed.
The Future of Financial Analysis
While financial analysts aren't likely to lose their jobs to AI anytime soon, this technology could make them vastly more productive and efficient. Imagine being able to run a first-pass analysis on dozens of companies in the time it used to take to assess just one. Human expertise will still be needed, but AI can significantly enhance our analytical capabilities.
Check out the full conversation in Episode 389 of The Accounting Podcast for more examples of how AI is reshaping our profession.
