Commerzbank is hoping that computers will soon be able to do at least as good a job writing its equity research reports as the armies of junior analysts that the big banks are no doubt looking to trim thanks to expensive MifidII regulations and restrictions that have cut funding costs for research departments.
Even as its captured the attention of bank executives, automated and computerized equity analysis has, for the most part, been a disaster over the last couple of years. While some larger firms may use algorithms and some automation to crank out macro economic reports, and while computers may be getting better at scraping and reporting data (without actually analyzing it), performing equity analysis requires a deeper look behind the numbers and its simply not a task optimized for automation.
However, we are apparently at that stage in the cycle where cutting costs becomes far more important then being productive or effective, particularly since MiFid II is forcing a race to bottom as investment banks seek out deep cuts in their research departments, driven by a drop in revenue that has accompanied being forced to charge a separate, optional, rate for research instead of bundling those costs with trading fees.
One of these competitors, Germany's second largest bank has decided that the time has come to automate some of its equity analysis, and according to the Financial Times "Commerzbank is experimenting with artificial intelligence technology that automatically generates sports reports to see if it can write basic analyst notes, as Mifid II forces banks across the world to trim research costs."
The German bank is working on the project with Retresco, a content automation company in which it invested two years ago through its fintech incubator unit. The project is still at an early stage and could take years to produce reports that banks would be happy to send to their clients, but the notion of AI replacing human research analysts is already attracting attention from senior bankers.
"There’s definitely work that can be done, parts of the [research] process that can be enhanced by algos and AI tools," the head of one investment bank told the Financial Times, describing earnings reports as something that "should be robo-written."
Research into AI and automation solutions that can lessen the burden of data-intensive research will likely soon be a theme across the big banks, as they scramble to reduce one of their biggest cost-centers in a time of declining revenues.
The Europe head of another investment bank said research was an area that was rife for automation over time, while analysts at several other banks said their managers were experimenting with AI and automation applications.
Banks are under fierce pressure to cut the costs of producing research on stocks and bonds following the implementation in January of European investor protections known as Mifid II. The measures force investors to pay for research explicitly instead of bundling its costs into trading commissions. Some firms say their implied research revenue has fallen by as much as 30 per cent as a result.
Possibly ignoring the fact that almost everybody (in the U.S.) reports in Non-GAAP numbers now and that any and all addbacks to earnings generally need to be looked at and analyzed on their own, a Commerzbank executive is confident that the venture would ultimately be successful.
Michael Spitz, head of Commerzbank’s R&D unit, Mainincubator, said the area showed promise because “equity research reports reviewing quarterly earnings are structured in similar ways” and the source documents are often prepared under common reporting standards. "That makes it easier for a machine learning program to extract and contextualise relevant data, which can be then framed in a report using natural language processing tools." Retresco’s original business uses similar technology to write soccer reports in Germany, in other words if it works for sports it should work for the market.
Mr Spitz said the technology was already advanced enough to provide around 75 per cent of what a human equity analyst would when writing an immediate report on quarterly earnings. “If it is related to much more abstract cases, we feel that we are not there yet — that we can or maybe will ever replace the quality of a researcher,” he added. Bankers say regulatory demands for oversight on research publication could also protect humans in research jobs.
Recall, it was less than a year ago that we wrote about the first AI-controlled ETF. At the time, its creators said it "has the ability to mimic an army of equity research analysts working around the clock, 365 days a year, while removing human error and bias from the process."
Last year, EquBot LLC, in partnership with ETF Managers Group (ETFMG) launched the world’s first ETF powered by artificial intelligence, the AI Powered Equity ETF (NYSE Arca: AIEQ). According to Business Wire, the new ETF uses "cognitive and big data processing abilities of IBM Watson™ to analyze U.S.-listed investment opportunities."
Business Wire explained how EquBot makes investment decisions "EquBot’s approach ranks investment opportunities based on their probability of benefiting from current economic conditions, trends, and world- and company-specific events, and identifies those equities with the greatest potential for appreciation. EquBot and ETFMG expect the fund’s portfolio to typically consist of 30 to 70 of U.S. equities only and volatility comparable to the broader U.S. equity market…the fund’s underlying technology is constantly analyzing information for approximately 6,000 U.S.-listed equities, including company management and market sentiment, and processes more than one million regulatory filings, quarterly results releases, news articles, and social media posts every day.”
The moving of all financial services - including equity analysis - into AI, feels like it could become a major error not only as real human analysts will possibly be needed to reverse work that computers will likely do poorly, at least at first.
A bigger problem is that this "revolution" will come just as the paradigm that has defined markets for the past decade: central bank largesse pushing risk assets higher, fades, and neither AI nor unmanned algos will be able to trade in the "newer normal." Ironically this is precisely the time when humans will be most needed.
But that bridge has yet to be crossed, and until then the main prerogative is to keep costs low.
With that said, it seems unlikely that any bank has the artificial intelligence or automation on the level necessary to effectively dissect the story and the narrative that are behind the numbers yet. Consider every time trading algorithms have misinterpreted a headline, only to be kneejerked back and forth until human traders intervene to "discover" the price.
For banks looking for a quick revenue saver, this option will almost certainly prove to be more trouble than it's worth.