Financial firms are toeing the waters of artificial intelligence (AI) and machine learning (ML) with focused initiatives to enhance operations and make more informed decisions. Some of the largest financial institutions in the U.S. are using AI to execute some trades and apply the technology toward detecting fraud. The ability to use this emerging technology to solve for specific issues is part of what makes it compelling and why startups with targeted-use applications are proliferating.
While the technology promises to provide data-driven insights into the processes that run your business and lead to lower costs, better risk controls and more efficient operations, it is not expected to give rise to an enterprise-wide revolution so much as an evolution. Quietly and gradually, application by application, AI and ML are expected to turn today’s higher cost and labor-intensive tasks into machine-based activities, freeing up staff for more client-facing activity.
Defining Terms: The Difference Between AI and ML
AI and ML are not one and the same. AI applications attempt to process language, understand complex data sets, and then synthesize the analysis into conclusions—much like humans do. As a result, AI is being applied to more cognitive tasks. For instance, it is being used to make chatbots more helpful and the information they communicate more personalized.
ML is a close cousin, but more analytical. Its best use is in sifting through immense amounts of data, seeking complex patterns and any gaps or deviations from standards. ML programs help with predictability and classification tasks, making it more appropriate for back- and middle-office activities.
Both AI and ML are iterative in that these applications learn from experience. Self-coding learners that they are, they improve on their own, as experience in a task accumulates. As such, they require minimal additional new coding and continually update themselves, minimizing the disruption and costs associated with upgrading. AI and ML applications are essentially their own latest version.
What the Technology Can Do
While companies and external investors are making investments in AI, more dollars have been flowing toward the development of ML applications. According to a 2016 McKinsey Global Institute Study, 60 percent of the $8 to $12 billion in AI funding went specifically into ML due to the quicker coding capability.
Where financial service applications are concerned, many of the applications that have emerged so far are focused on trading. For example, they seek to enhance the information available to traders, improve the efficiency of execution and post-trade processing, and strengthen the protections against misconduct. These ML applications are essentially tools that enhance the work humans do by empowering them with better information.
Accordingly, the types of back-office tasks that the technology is expected to enhance, speed up, and eventually take over include:
- Document scanning and the parsing of the content each piece contains for use by other programs.
- Spotting errors, fraud and deviations from policy faster so preventive actions may be taken or noncompliant violations can be investigated, reported, and addressed sooner.
- Improving the confidence level of counterparties to a transaction through quicker verification and confirmation.
- Optimizing processing, authorizing, and identifying the need for margin calls.
- Providing ongoing, real-time information regarding a firm’s capital on-hand and the accurate segregation of client funds from their firm funds.
- Risk assessment across a firm’s entire exposure.
What the technology can’t do is overcome bad data. The same problem that undermines less powerful technology still holds true—bad data in leads to bad data out. Ensuring the quality of the inputs is even more vital to the effectiveness of AI and ML-based programs and apps, since the technology is constantly building on its “knowledge base.”
The technology is also susceptible to an added complication: programming biases. If it isn’t monitored, it can inadvertently pursue a biased learning pattern that could lead to poor decision-making. Guarding against that is where humans come in.
Upgrading the Trade Settlement Process
While ML might be applied to speed up the way trades are executed, confirmed and settled, its best use is likely to be in exception processing. Instead of having middle-office employees go through the labor-intensive process of identifying the reasons behind a trade being rejected, then investigating, confirming and fixing the trade, ML applications may be able to take over error resolution.
ML can also be used to analyze historical data on failed trades to understand the causes and learn to identify the underlying factors for quicker intervention and possibly even prevention. This would not only reduce exceptions and their related costs, it could eventually do so before a client’s account, a firm’s trading limit, or the clearing firm’s capital position are impacted.
Across the industry, AI and ML technology that helps enforce regulations and keeps firms compliant is also receiving attention and investment. These applications have the potential to sort through vast amounts of data looking for patterns that can flag instances of noncompliance—intentional and unintentional. This would result in quicker action on both the part of firms and their regulators.
A Disruption for the Bottom Line
For FCMs and securities firms, making better use of data is the promise of this emerging technology. More than decreasing the uncertainty systemic to trade execution, it can empower you to make better informed decisions about your capital needs and operations. In that sense, AI and ML applications could be well-timed solutions for the problems your firm faces today.
Disclosure: The information in this article has been obtained from sources deemed reliable; however, we do not guarantee its accuracy. This information is not intended to be legal, investment or tax advice and should not be relied upon. MB Financial Bank, N.A. and its affiliates do not provide legal or tax advice. You should review your particular circumstances with your legal and tax advisors. Member FDIC
SVP, Division Manager, Commercial Banking
MB Financial Bank