Compete or cooperate?: Research analysts face choice as AI writes stock reports

A recent release of research reports written by generative AI from a major brokerage has posed a burning question in Korea's financial circle: Will AI kill the research analyst?

An image of an electric stock trading board [GETTY IMAGES]

[NEWS IN FOCUS] A recent release of research reports written by generative AI from a major brokerage has posed a burning question in Korea's financial circle: Will AI kill the research analyst?

Not quite yet, perhaps, but analysts now face a watershed — survive against their machine competitors or thrive with them.

Mirae Asset Securities published a seven-page report on Apple’s latest quarterly performance on May 3. The report dissects every key financial figure and effectively summarizes executive comments.

According to the report, most of the tech heavyweight's key financial indicators missed the AI model's expectations, with product sales, revenue from the Japanese market and iPhone sales "significantly falling below the forecast," whereas service revenue and Mac sales turned out to be better than expected.

The AI analyst blames sluggish iPhone sales for the company's disappointing results in the second fiscal quarter, noting that Apple's stock price has been performing far worse than its major competitors Microsoft and Alphabet.

"The correlation between Apple and the [Nasdaq] market index has grown weak as of late," the report added, concluding that the poor performance has already been reflected in the stock price.

The report certainly lacks a bit of the flair and depth usually added by research analysts, but it still gets the job done — in less than 20 minutes at that, in contrast with the five hours required for human analysts to complete a similar report.

Mirae Asset Securities began releasing AI-written reports on U.S. stocks, including Apple, ExxonMobil and Starbucks, in May. It takes about 5 to 15 minutes on average for the firm’s AI model to draft a report based on regulatory filings on the Securities and Exchange Commission, along with matching graphs and tables.

Of course, the reports are still largely simple in their analysis and cannot provide any buy, sell and hold ratings. Moreover, as the current law prohibits analyst reports from being published by those without certifications, human analysts weighed in to edit and supervise the draft created by AI, though their intervention remained minimal, mostly focused on corrections in translation.

Machine learning and automation tools have been around for quite some time in the financial sector. Many securities firms have been rolling out various AI-powered services, including chatbots, newsletters and brief corporate analyses, but it is still the first time in Korea that full-fledged analyst reports on major companies written by AI have been published.

Mirae Asset plans to refine the model further and significantly increase the number of AI-driven analyst reports.

Other firms in Korea are also ramping up their efforts to incorporate the technology. Korea Investment & Securities, for one, was the first to adopt AI analysis tools to provide research reports focusing on small-cap companies in 2020. The firm has since launched other AI-powered services, including a virtual human analyst that introduces and summarizes research reports for investors.

The latest developments across the industry are enough to raise the question: Would AI be able to replace research analysts or assistants in the long run?

“Of course, there would be some concerns,” acknowledged Kwon Young-bae, head of the AI research support team at Mirae Asset Securities.

Kwon, however, stressed that this may also be “a new opportunity.”

“Whether it be research assistant or analyst, those who would compete against AI in what AI does the best, such as numerical data analysis or quick processing, won’t be able to survive; but those who try to use AI to provide, for example, deeper analysis, may benefit from it.”

Noh Sung-ho, a research fellow at the Korea Capital Market Institute, suggested that “the safest way to go is to utilize [AI tools] to boost existing human research analysts’ productivity.”

Noh stressed that “generative AI is certainly amazing technology, but we haven’t reached a point where people can fully trust AI-written corporate analysis to make investment decisions,” adding that “analyses conducted by trustworthy and responsible individuals are still irreplaceable.”

However, it is crucial to make long-term investments to secure the technology as Korea lags behind market leaders in AI development and utilization, Noh pointed out.

“Managing and developing AI models is costly, not to mention it is hard to evaluate its value and capabilities with tangible profit,” said Noh.

“But [the technology] has already been driving productivity — companies, therefore, need to have a long-term perspective to weigh the benefits against expenses.”

BY SHIN HA-NEE [shin.hanee@joongang.co.kr]