Investors are increasingly adopting artificial intelligence tools in their workflows, but the technology—at least for now—works best hand in hand with human input.
“Advances in software and processing power are enabling the dynamic combination of different factors, allowing for more sophisticated portfolio construction and signal extraction,” says Paul Kenney, senior vice president at Syntax Data. “Crucially, AI can help distinguish meaningful data from noise.”
In a 2024 paper in the Journal of Financial Economics, “From Man vs Machine to Man + Machine: The art and AI of stock analyses,” researchers found that AI tools could outperform humans in 54.5% of stock predictions, although humans were better than AI at understanding the institutional backgrounds of firms and industries, something with which AI struggled. The best outcomes will happen when AI and humans work together, the paper argued.
“We found that when firms have huge amounts of information—tax documents, press releases, etc.—machines did better at processes and synthesizing information,” University of Maryland researcher Sean Cao wrote in the paper. “But when a firm has a lot of intangible information—like a strong team, a lot of R&D, lots of knowledge capital—experienced human analysts are better at forecasting in those cases.”
For more stories like this, sign up for the CIO Alert newsletter. ?
By itself, AI cannot uncover in data a hidden pattern that somehow generates alpha, notes Stas Melnikov, head of quantitative research and risk data solutions at SAS.
“If that was the case, then many other market participants would be able to do the same, and that alpha would no longer exist,” Melnikov said via email. “You cannot just prompt ChatGPT to make an investment strategy that maximizes the Sharpe ratio. As an extrapolation of all the knowledge used to train it and available via the internet, the information used to generate that signal has already been uncovered. If there was any alpha before, it is gone now.”
Still, AI tools are already an important part of investment research through informing its process.
Assisting With Research, Operational Tasks
Asset managers, banks and institutional investors have been using AI tools to automate significant portions of their operations, such as note-taking, summarizing earnings calls and drafting investment memos.
“AI can be used as a junior analyst to help with investment research. As the research universe [can] be broadened for a given portfolio, the opportunity set—and potential alpha associated with it—increase,” Melnikov said.
While AI is being used to inform the investment process, human oversight is still required when final decisions are made, and there are many aspects of investment research that cannot be replaced by a chatbot.
“AI still cannot replace the ‘boots on the ground’ research—talking to company management, suppliers, competitors, employees and more—at the heart of a fundamental investment approach,” Melnikov said.
A paper from BlackRock noted that while AI can help navigate thematic trends—something the paper noted can harness alpha—a portfolio manager’s expertise is engrained in every part of the process.
“The involvement of human experts will remain crucial to any application of AI in investment risk management,” the paper stated.
Others agree about the need for human involvement.
“I have one fund adviser that I work with [that has] an investment strategy that significantly leverages AI for purposes of selecting the investments,” says Karen Aspinall, the financial services practice area chair and a partner in Practus LLP. “But it is not a situation where the AI just runs and selects the investments, and that’s the end of it. There is still that human oversight on it at the end of the process.”
“I can say, across all the emails I got from different researchers and sell side [analysts] over the last week … I’ll get a pretty decent response that covers all the outliers,” says Max Gokhman, deputy CIO of Franklin Templeton Investment Solutions. “That helps me, instead of reading all those emails myself, I can quickly figure out who has the opinions that are most contrarian or most interesting to me for a particular asset class. That’s your really basic efficiency savings: It doesn’t generate alpha, but it gives you more time, and more time for a human should mean they have more time to generate alpha themselves.
Subject to Data Quality
Bad data quality can undermine the effectiveness of AI tools. Inaccurate, incomplete and biased data can lead to flawed insights in the research process.
“Bad quality data results in noise and false signals, which hurt alpha,” Melnikov said. “Theoretically, AI can be used to clean the data and improve its quality. If such efforts are successful, alpha will be generated. However, AI for data quality is a double-edged sword, as it can erroneously classify outliers and anomalies as errors whereas those may contain an important signal. At this stage, data cleaning cannot be fully outsourced to AI.”
Automation and machine learning can accelerate analysis and uncover patterns, but these benefits are only as reliable as their underlying inputs, reinforcing the need for human oversight over AI.
“While AI can reduce costs and broaden access, concerns about data quality and over-automation underline the importance of maintaining human oversight, especially in complex or ethically sensitive decisions,” says Markus Alin, CEO of Sharpfin, a Stockholm-based software provider for wealth management.
Tags: Artificial Intelligence, Asset Management, Technology
#Institutional #Investing