The Huge Opportunity for Artificial Intelligence in Investment Operations and Accounting

In the middle and back offices, the depth and breadth of industry expertise will be a bigger driver of AI innovation than data science 

By Tom McMackin, SVP, Marc Zimmerman, SVP, and Scott Kurland, MD, of SS&C

Although AI technology is increasingly being embraced by most industries to drive enhanced operational efficiencies, customer experience and financial performance, the financial services industry was actually one of its earliest pioneers. The ability to predict stock market movements has been the holy grail for institutional investors since the inception of securities exchanges. It is little wonder that this industry segment first began research with predictive analytics as early as the 1950’s and 60’s. A decade later, capital markets began to see the development of the first algorithmic models that were used to accelerate trading decisions. Today, these tools have evolved into the High Frequency Trading (HTF) systems that execute millions of transactions daily.

In very real terms, AI offers the ability to make faster and smarter decisions, translating into billions of dollars a year for financial services institutions. It should come as no surprise then that investment banks and hedge funds have been pouring significant funds into related software technology and research for years.

Investment Operations and Accounting – A New Financial Services Frontier for AI

To date, the major share of the AI spend within capital markets has been squarely focused on enabling front-office trading and customer-facing business functions, while middle- and back-office operations have remained largely unchartered territory. Investment operations and accounting systems have become increasingly sophisticated in their efforts to address the industry’s ever-changing accounting standards and regulatory compliance requirements, but thus far these applications are not ‘smart.’Instead, they are typically legacy solutions driven by hard-coded rules and processes that enable the inflow of structured data from external sources, such as counter-parties, custodians, securities exchanges,and clearing and settlement systems.  Unlike front-office trading operations that utilize AI to make more enlightened decisions based on Deep Learning and Big Data analysis, the goal of the middle and back office is to perform accounting, regulatory compliance and other operational tasks with the utmost efficiency and precision.

No “Do-Overs” in the Middle and Back Office

Clearly there are big opportunities to leverage AI in the middle and back-offices. Examples include automating reconciliation processes, reducing the burden of exception management, and enabling faster, more effective remediation of identified errors. However, the middle and back-office calls for a substantially different approach to the use of AI — one that is driven as much by the depth and breadth of expertise in investment operations and accounting as it is by the AI technology itself. In the front office, bad decisions one day can be compensated by better decisions the next. Anomalies are expected as part of the asset management and trading process. Not so in the middle and back office, where numbers either add up correctly or they don’t. If they don’t, the exceptions must be quickly identified, reconciled and repaired to avoid undesired downstream consequences with regulators, auditors and stakeholders.

More Sophisticated AI Tools Don’t Always Produce More Sophisticated Results

Today’s most sophisticated, front-office trading models leverage advanced AI tools like Deep Learning with Advanced Neural Networks (ANN) to uncover new patterns, and reach insights and conclusions through interpretation of data – similar to how the human brain functions. Unfortunately, Deep Learning models are not yet an exact science. Like human brains, these tools can draw inaccurate conclusions. In the more precise world of investment accounting, there is little room for opinion — human or machine. Machine learning models must be thoroughly trained and tested to produce very specific and accurate outcomes. They must be designed to identify only appropriate patterns, then suggest or trigger appropriate actions relevant to operations and accounting processes. This takes highly specialized investment operations and accounting expertise, not only with regard to middle-to-back-office functions, but also across a continually expanding landscape of asset types, transactions, markets, regulations and industry operating models.

In the investment accounting space, machine learning can be used to reduce time and cost associated with data inquiries by providing relevant context. Once the machine learning model identifies an issue and where it resides, it can either suggest the proper course of action to resolve it, or autonomously initiate the appropriate workflow through “Intelligent Workflow Automation” (IWA). IWA technology learns from user behavior to identify and automate the appropriate workflow processes needed to locate and resolve the problem without manual intervention.For example, in the investment operations area, reconciling position holdings from custodians can be time consuming if there are differences in the quantities due to out-of-date factors. Pattern recognition algorithms can discover these breaks and resolve them quickly.

It is essential that the behaviors of IWA models are thoroughly scripted and tested by highly experienced and knowledgeable investment operations and accounting experts to ensure that target automations correctly perform the tasks at hand. No matter how sophisticated the models are, they will not produce valid results unless they are guided by specialized investment accounting and operations domain expertise to know what anomalies, parameters and drivers to look for.

Increasing the Value of AI with a Single, Unified Platform

What the financial services industry generally calls “integrated investment operations and accounting systems” are essentially a series of disparate functional applications or modules that are loosely wired together in an effort to more efficiently perform a range of middle-to back-office functions and services. To the extent they are effectively integrated, they endeavor to exchange data and automate hard-coded transactions from beginning to end in a serial sequence commonly referred to as “Straight Through Processing (STP)”. However, here no pattern recognition is involved –it is simply one specific event triggering another specific event with no enhanced intelligence.

Today’s modular investment operations and accounting systems lack the unified architecture needed to exploit the full value of AI. The full potential of machine learning and intelligent workflow can only be realized when an application can holistically recognize patterns and pinpoint exceptions across relevant functions, activities and data anywhere in the system. Only then can the system autonomously initiate the best actions or recommend the most appropriate next steps.

This process, however, is problematic where separate applications or modules have been cobbled together to look, but not really act like one. To truly enjoy the full benefit and value of AI in a middle and back-office investment setting, institutions will need a unified technology platform that provides a single database and user interface together with a rich collection of pre-integrated functions and common services. The platform also needs to be able to support all the diverse assets, transaction types and industry operating models that define an institution’s businesses. Successful solution pioneers in this new space will likely have long and successful track record in the industry, deep expertise in wide range of asset types and industry operating models, and an aggressive mergers and acquisitions strategy to continually deepen and widen that expertise.

Bottom line – AI tools have the potential to bring huge efficiency gains and cost savings to middle- and back-office investment operations, especially when embedded in software applications that singularly support ready access to all required data. However, successful deployment of these technologies also requires deep domain knowledge and expertise on the part of the application provider to truly optimize the capabilities and benefits of this innovative technology.