FINRA Releases Report on Securities Industry

The Financial Industry Regulatory Authority (FINRA), a private organization that contributes to the regulation of the financial industry, has released a report on Artificial Intelligence in the Securities Industry. The report discusses the application of AI in the securities industry with relation to brokers and dealers. It summarizes the primary application of AI by brokers and dealers with reference to three areas – communications with customers, investment process, and operational functions.

The use of technology to enhance the customer experience, through the usage of natural language processing by using virtual assistance to automate tasks and communication along with targeted outreach improve the overall interaction cycle. Further, for the investment process the report notes the significant impact the technology has on account management, particularly broker account management by maintaining holistic consumer provides to leverage customer profiles based on aggregating customer information across the internet. It can allow brokers to leverage a highly customized profile of each user to meet their specific requirements. This could even affect the manner of trading to maximize the efficiency and price-performance of trades.

The report also notes the significant impact the technology could have on compliance and risk management, by offering real-time surveillance and monitoring to automate otherwise mundane tasks. The report also acknowledges the necessity for model explainability, noting the “blackbox” nature of artificial intelligence and machine learning models. However, at the same time most players in the securities market require a certain degree of accountability to achieve compliance with security laws. Accordingly, the FINRA recommends ensuring some degree of explainability to ensure compliance with legal requirements. Further, they also acknowledge the dangers of potential bias within this data, and recommend verifying data and vetting the data to ensure minimal bias.