Institutional investors are missing opportunities to use artificial intelligence for effective decision-making, according to Ruslan Goyenko, an associate professor of finance at McGill University, speaking during a session at the Canadian Investment Review’s 2024 Risk Management Conference.
He urged investment leaders to push beyond traditional fundamentals and open their minds to the use of AI. As well, he highlighted the lack of training around AI tools. “There’s no program that prepares people on the real intersection of data science and finance.”
Goyenko and his colleagues embarked on a research project to find out how an AI tool would fare in producing accurate investment analyst forecasts. The tool wasn’t designed to predict results, he said, but to offer a roadmap of the fundamentals for an industry or specific companies.
Read: Institutional investors considering AI need responsible rulebook: report
The study’s AI tool was built with publicly available financial filings from the Securities and Exchange Commission’s electronic data gathering, analysis and retrieval system dating back to 1993. Specifically, the tool was tasked with closely reviewing the risk and management discussion and analysis sections of these financial filings.
The goal in collecting all of this data, said Goyenko, was so the tool could learn and predict the financial performance and earnings of a company during the next performing quarter. On average, the researchers found companies that beat analyst forecasts went on to see higher returns while underperforming ones saw negative returns.
The AI tool does the work of 1,000 analysts with a singular objective, with the additional ability to learn dynamically from a substantive data pool. As a technology tool, it improves on other digital attempts at scanning for keywords within financial disclosures, he noted, because it’s capable of understanding context. In addition, the researchers made sure to retrain and fine tune the AI’s language models on financial targets.
The study found the AI tool outperformed the average analyst forecast, said Goyenko. In particular, it performed well in portfolios where the analysts disagree the most. “The target here [isn’t] predicting returns; returns are incredibly hard to predict [because] they’re driven by many things. What we’re looking at is predicting fundamentals.”
In March 2020, for example, the AI tool pointed to buy shares of communications software firm Zoom because the company was spending resources to expand its sales personnel and to increase its client base. Initially, the public markets reacted negatively and returns for the company were negative, he said. However, Zoom went on to beat analyst expectations.
Read more coverage of the 2024 Risk Management Conference.