Original article was published by Raffaella Aghemo on Artificial Intelligence on Medium
Artificial Intelligence and FINRA Report on securities market (PART TWO)
The FINRA Report, relative to the use of the instrument of Artificial Intelligence in the securities industry, of which we had already spoken in previous editorials, relative to the first section of the document, in the second section of the same, deals with the use of the instrument by the broker-dealers. Such use is proliferating in the securities sector, so much so that some large companies have set up centres of excellence, aimed at examining, sharing, and building expertise and creating synergies related to the use of AI in their organizations, incorporating technology startups for “appropriate contamination” in terms of business.
There are three broad areas where IA can make a difference: customer communications, investment processes and operational functions.
In communications with customers, we explore new ways in which companies improve the customer experience and outreach targeting (i.e. how to reach customers or potential customers), through the use of virtual assistants, or through analysis of customer request emails, to speed up response times, as well as through an appropriate customization of the needs and profiles of the individual customer. In addition, companies have also indicated that tools are being explored to determine whether individuals would be interested in certain services, based on the customer profile and browsing history within the companies’ websites.
Regarding investment processes, the document explores how companies are using AI to assist in brokerage account management, portfolio management and trading.
Finally, as far as operational functions are concerned, the document explores how companies are using IA to assist them in compliance, the Know-your-customer (KYC) identification process, risk management and administrative tasks.
Although the use cases outlined in the Report may offer several potential benefits, but also potential challenges, costs and regulatory implications, each firm or company should conduct its own due diligence and legal analysis of any IA application to determine its usefulness, regulatory impact and potential risks, and establish appropriate measures to mitigate those risks.
In addition, the use of IA applications does not release companies from their obligation to comply with all applicable securities laws, rules and regulations.
The use of AI in applications to improve customer experience has gained considerable leverage, not only in the securities industry, but generally in the financial services industry. IA-based customer support applications largely involve NLP and ML-based tools that automate and customise customer communications.
The tools listed in the document related to the Customer Experience, are:
Virtual Assistants. A virtual assistant is an AI application that interacts with human beings, using voice recognition and speech synthesis and is programmed to perform certain tasks, providing answers to basic customer requests, such as account balances, portfolio availability, data market, address changes and password resetting (at some companies, they may also have the ability to accept and process commercial orders within certain thresholds). In addition, some companies are integrating artificial intelligence-based interactive voice response (IVR interactive voice response) systems into their call centres to respond to basic caller requests or to collect sufficient information to facilitate call routing to the relevant human customer service agents, using NLP (including speech-to-text/text-to-speech conversion, tone recognition and text generation), ML and sophisticated customer authentication tools, including the use of facial recognition, fingerprint and voice biometrics.
E-mail enquiries. Some companies are using IA-based applications to view and classify incoming customer emails based on key features, such as sender identity, email subject line, and automatic review of the email message itself. These applications can also automatically respond to emails containing common or routine requests, while delivering emails with complex requests to the appropriate staff. Companies have also noted the use of similar applications to process and manage internal requests (e.g., those received from internal help desks), to provide automated responses where possible, and to direct more complicated requests to dedicated expert departments.
Instead, the tools designed to improve investment processes and reported in this Report, go through the so-called Brokerage Account Management and the subsequent Portfolio Management and Trading, i.e. the management functions of brokerage accounts, which facilitate an almost symbiotic customization of the client’s needs, and the removal, even potential, of any possible friction.
The processes of support and reinforcement for the Brokerage Account Management provide:
Holistic customer profiles. Companies are starting to develop IA-based applications that create holistic, real-time client profiles that incorporate information from a wide range of sources, such as client assets (held both inside and outside the company), spending patterns and debt balances obtained through data aggregation tools; updates on social media and other public websites; browsing history on the company’s website and mobile applications; and past communications (e.g., from emails, chat messages and meeting notes). All of this information is analyzed using IA tools to provide a broader picture of the client’s needs, along with customized suggestions on which investment products might be of interest to the client, while also highlighting a cautious approach to the use of these “invasive” methodologies for various legal, regulatory and reputational concerns.
Personalized research. as indicated in the previous section, tools based on Artificial Intelligence can offer clients social media data and related sentiment analysis on investment products and asset classes.
The tools for an optimization of Portfolio Management and Trading, i.e. portfolio management and trading, provide:
Portfolio management. Artificial Intelligence tools draw on large amounts of data available from both internal and external sources, including non-traditional sources such as social media and satellite images, in order to identify insights that can signal a price movement.
Trading. Securities industry participants are exploring AI tools to make their trading work more efficiently, maximizing speed and price performance. Some examples are the use of ML for intelligent order routing, price optimization, best execution and optimal allocation of block trades, but with the caveat that the use of AI in portfolio management and trading functions can also pose some unique challenges, particularly when trading and execution applications are designed to act independently.
Unusual and untapped circumstances in the formation of the model, such as unusual market volatility, natural disasters, pandemics or geopolitical changes, could create a situation where the model no longer produces reliable forecasts, triggering undesirable commercial behaviour with negative consequences. In addition, the possibility has also been raised that models can learn from each other across the industry, to say the least, that they may be risky, triggering collusive behaviour, opening the door to unpredictable results.
Finally, in the so called operational functions, the last category of the triptych of areas where the use of AI/IV can make a difference, the FINRA Report notes a massive effort by companies in the sector to apply the tool to improve compliance and risk management functions, as broker-dealers have to keep pace with complex and changing national and international regulations, as well as with a rapidly changing risk landscape (e.g. IT security, internal threats and financial risks), in addition to administrative ones.
From a Compliance and Risk Management perspective, we will look at some examples of AI/IV application in this business area:
Surveillance and monitoring. AI technology offers companies the ability to holistically capture and monitor large amounts of structured and unstructured data in various forms (e.g. text, voice, image and video) from both internal and external sources in order to identify patterns and anomalies. This enables companies to oversee and monitor various functions within the company, as well as to monitor the conduct of various individuals (e.g. merchants, registered representatives, employees and customers), in a more efficient, effective and risk-based manner. Market participants have noted that these tools could significantly reduce the number of false positives, which in turn free up the time of compliance and oversight personnel, time to conduct more in-depth reviews of remaining reports, resulting in increased escalation rates. Companies indicate that these tools offer the ability to move from “traditional rule-based systems to a predictive, risk-based surveillance model that identifies and leverages models in data to inform decision making.
Customer identification and monitoring of financial crimes. AI-based tools for customer identification (also called “know-your-customer” (KYC)) and financial crime monitoring programs are also being developed to identify potential money laundering, terrorist financing, corruption, tax evasion, insider trading, market manipulation and other fraudulent or illegal activities. Market participants have noted that many traditional methods of monitoring KYC and financial crimes are cumbersome and not as effective as they would like, often resulting in high false positive rates. As a result, companies have begun to incorporate IA technologies, such as ML, NLP and biometrics, to make their programs more effective.
Regulatory intelligence management. Some industry participants noted that automated regulatory intelligence management programs have the potential to increase overall compliance while reducing both the cost and time required to implement regulatory change.
Liquidity and cash management. Such applications would have the ability to analyze substantial historical data, along with current market data, to identify trends, note anomalies, and make predictions about, for example, intra-day liquidity needs, peak liquidity requirements, working capital requirements, and securities lending demand.
Credit risk management. This could speed up the credit review process by incorporating non-traditional criteria (e.g. information available through the social network). However, some IA-based credit-scoring systems have faced criticism for their opacity and potential bias and discrimination. These models not only analyse traditional credit assessment criteria, such as current financial situation and credit history, but can also identify other demographic factors as deterministic criteria, which could lead to an unfair and discriminatory assessment based on bias in the underlying historical data.
Cybersecurity. Cybersecurity continues to be a major challenge for the financial services industry: “sixty-nine percent of organizations believe that IA will be necessary to respond to cyber attacks.
With regard to automation in administrative tasks, within the above mentioned operational functions, companies are incorporating software enhanced by IA technologies (e.g. ML, NLP and CV) to automate manual, less complex, repetitive, high-volume tasks that traditionally involve significant human time for staff. Companies have indicated that automating such tasks with IA tools has the potential for high performance, cost savings and efficiency gains, especially in the following cases:
Automation of manual processing on paper. That is, automating functions that involve manual document review, such as processing business orders by fax, search, classification and retrieval of documents. These applications include CV and NLP to identify customers, review signatures, read orders, and scan documents. Companies have indicated that these applications not only increase productivity, but also accelerate important processes, such as trade and payment processing.
Document review and information extraction. NLP allows companies to examine significant volumes of documents (e.g. legal contracts, custody documents and loan contracts) at a fraction of the time needed for human analysis, for example, to look for certain clauses or key points within a category of contracts.
Other applications. For example, these tools improve the accuracy and efficiency of existing processes, such as reporting and reconciliation of invoices.
Section Three will be the subject of my next editorial.
All Rights Reserved
Raffaella Aghemo, Lawyer