Artificial Intelligence: Do Risk Managers Get It?

When it comes to artificial intelligence, investment analysts, customer service managers and compliance managers seem to agree that the technology can help them make informed decisions more quickly and reduce investment, credit and regulatory risk.

What about risk managers? They don’t understand how to incorporate AI into their risk metrics or participate in the decision-making process to determine whether it should be implemented at all, say industry analysts and operations specialists. “The financial industry is moving at a faster pace with artificial intelligence than the skill set of risk managers,” explains Steve Culp, senior managing director for finance and risk services at global consultancy Accenture in Chicago. “Risk managers can’t afford to be left behind.”

The reason: they will either miss the business opportunities AI offers or implement it incorrectly. Mistakes at one financial firm could ultimately have a domino effect on multiple financial institutions. A 45-page report just released by the Financial Stability Board, calls for more specialist staff to oversee risk models for AI which could lead to “unintended consequences” if they are too opaque. “If multiple firms develop trading strategies using AI and machine learning models, but do not understand the models because of their complexity it would be very difficult for both firms and supervisors to predict how actions directed by models will affect markets,” says the FSB, which represents central banks and regulators for the G20 economies.

As a branch of computer science, artificial intelligence gives machines similar traits to human reasoning. The term now incorporates machine learning, or the ability for computers to learn by ingesting large quantities of data, as well as natural language process or the ability to read or produce text. Culp sees the largest applications of AI in evaluating credit quality, improving know-your-customer procedures and supporting robo-advisory work in wealth management.

Yet about 69 percent of the 475 senior risk managers at banks, insurance companies and other financial service firms surveyed by Accenture worried that a shortage of skills in new and emerging technologies would impede the effectiveness of the risk functions. Only ten percent of the respondents to the survey conducted in January 2017 say that their risk management teams have the internal resources to perform their tasks leveraging AI in specialized areas.

Another study conducted by the risk management trade group Global Association of Risk Professionals (GARP) showed that 67 percent of 200 risk management professionals questioned think of AI as making a profound change on business practice. However, 60 percent have no role when it comes to how AI is implemented at their organization. About 70 percent have no plan to incorporate AI into the risk management process.

Risk managers surveyed by GARP would rather leave the work of implenting it to someone else. About 30 percent of the respondents to GARP’s survey say that the IT department is better equipped to handle AI, while 27 percent say that business line managers know what’s best. That leaves 17 percent thinking that a research and development department has more knowledge. The remaining 30 percent say no one understands AI.

“Part of the challenge of AI is the lack of a commonly-agreed adoption model,” says Hayden Shaughnessy, a research fellow at the University of California in Irvine who helped craft GARP’s survey. “That model would be the process for moving from initial thinking through proving out a particular concept and then prototyping in order to prepare for implementation.”

AI, Shaughnessy believes, inherently lacks the necessary key performance indicators and benchmarks that could make it easier for risk managers to present a business case to other C-level executives. The potential for AI to change workflow — either make tasks easier for employees or reduce their need to do the work altogether — is also not well understood by risk managers.

What to Do

Samuel Won, managing director of Global Risk Management Advisors, a New York-based risk technology and management advisory firm, is confident that it is only a matter of time before AI is incorporated into risk modeling.  “Artificial intelligence is nothing more than an advancement of computer technology, such as what was developed by black box trading and fuzzy logic,” he explains. “Ultimately, it comes down to whether the AI coding was written correctly and tested properly, as well as whether the institution using AI understands the limitations for risk management.”

Risk managers don’t need to know as much as technologists when it comes to understanding AI. However, they do need to have a role at the decision-making table. “Risk managers need to be able to successfully work with technologists to translate and convert their risk management subject matter expertise into machine learning which is the underpinning of AI,” says Won. Of course, banks should include risk managers as well as business line managers and compliance managers in the decision-making process to devise the rules for how AI should work.

The question then becomes do risk managers really want to make the effort to understand thir role in implementing AI? Not always, is the answer. “Risk managers have expressed hesitance,” acknowledges William Fayerweather, vice president of strategy for Broadridge Advisor Solutions, the wealth management technology unit for business outsourcing giant Broadridge in New York. As a result, risk managers become a hindrance rather than a help in the adoption of AI.

Listening to IT experts explain the nitty-gritty of how AI works won’t be enough to persuade risk managers of its merits. In the case of Broadridge’s AI application, which harnesses machine learning and natural language to analyze customer data and make investment recommendations, internal business lines at prospective user firms were able to convince risk managers that AI held the potential for increasing revenues.  When risk managers were able to review the results of three-to-six month trails in parallal with existing procedures, it was enough to persuade them to ride the AI train.

Five unnamed financial advisory firms are now using Broadridge’s AI application to sift through client data, recommend new products and suggest new communications language about suitable services in either emails, mailed letters or social media postings. Compliance departments no longer have to read each message for every client separately. About 90 percent of the messages can be automatically approved without the need to be rewriten.

“It’s all about monitoring the rise of the personalized message targeting the right customers at the right time,” says Fayerweather.