Latest from the Quantamize Blog

AI Replacing Analysts Writing Reports

Jun 27, 2018

Start Your Free Trial Today

MiFID II, imposed in January 2018 as a replacement of the original MiFID guidelines which were enacted in November 2007, is a legislative framework designed to protect European investors. MiFID II includes a string of new regulations, including requiring investors to pay for equity research as a standalone product, rather than bundling the costs into trading commissions. This development has had a significantly negative impact on European research houses, with numerous firms reporting that their research revenue has declined by upwards of 30% due to the new MiFID II restrictions. 

As a result of decreasing research revenue streams, many European banks are seeking cost cutting initiatives to maintain profitability. One of the cost reduction efforts by these banks is the automation of research and analyst reports. Commerzbank, a German finance group, is working with content automation firm Retresco to adapt Retresco’s technology to determine if it can also generate basic analyst reports. Retresco had originally intended to produce automated sports reports.

The generation of company earnings reports has garnered significant attention from Commerzbank. Michael Spitz, the head of Commerzbank’s research and development unit Mainincubator, posits that the firm’s technology can currently produce quarterly earnings reports that cover 75% of the material that a human analyst would be able to. Quarterly earnings reports tend to have a uniform structure, and the source documents used for earnings reports are often generated using a relatively homogenous reporting standard. The relative consistency of earnings reports improves the probability that a machine learning program will be able to successfully navigate these reports and extract useful information. Natural language processing tools can then be utilized to contextualize relevant data.

Despite the significant potential benefits of using AI to conduct research, there are roadblocks to consider. Developing automation on such a large scale often takes time to successfully implement. Commerzbank’s machine learning project, for instance, is still in its infancy and it may take several years to produce quality reports. There are further concerns about the ability of AI to handle unique cases, which may require critical thinking by a human. If the inputs or sources for a report are not standard and require a higher level of cognition, it may be beyond the ability of an AI system to analyze these reports. There is uncertainty regarding machine learning’s ability to ever produce work that is truly comparable to that of a quality researcher. 

Regulatory demands may also be an issue when implementing an automated reporting system. Every research publication that is generated must meet certain criteria in order to be deemed compliant with the financial guidelines in place. A machine learning program can be taught regulatory rules, but there are nuances and different use cases that may be incomprehensible to a machine. 

Concerns regarding the time, resource costs, and technical ability necessary to develop a functional AI research mechanism do not seem to be deterring many banks. Pressure to maintain high levels of profitability, in combination with slowing revenue, is driving banks to find alternative methods for meeting performance goals. Cost reduction through automation appears to be a logical step towards this objective.


To view the article source for this piece: