Contributed by Expert System Enterprise Corporation
Written by John Paty, Vice President

Now that the annual CCAR process (aka ‘bank stress tests’) has become part of the metronome of US financial regulation, major financial enterprises know all too well the haze that the Federal Reserve’s 1 – 4 rankings create. The financial markets have begun to lump these ‘shades of gray’ into a binary pass/fail tweet. In effect, pass the stress tests with all your documentation in place, go un-noticed; alternatively, score low in the stress test, watch your market cap radically decline. Financial markets have come to expect CCAR success; therefore, CCAR represents only a down-side risk for financial enterprises. The CCAR process is a time-drain and most CCAR-compliant enterprises are striving to find ways to not only better institutionalize the process, but also increase their precision and efficiency. Many of these labor-intensive issues could be handled by AI (artificial intelligence) and machine reading of content.

While the financial services sector is notoriously tribal (build it yourself, not invented here, who else uses this?, etc.), perhaps it would be valuable taking a page from another sector that also has tremendous government oversight and faces the same pass/fail risk. Life science companies face their own ‘stress test’ for each U.S. Food and Drug Administration (FDA) drug submission. Innovative life science companies use AI machine reading to analyze their pending submission and to ensure that all content is consistent with previous filings. Can you think of anything more complicated than an FDA submission? If cognitive machine reading can read, curate and analyze some of the most complicated oncology drug applications, a CCAR analysis should be achievable.

While fewer banks now have to participate in the CCAR process, major financial enterprises that must submit to the Federal Reserve could benefit from a system of machine reading their submissions, similar to life science companies. Financial enterprises could better focus their submissions and ensure that previous submissions are not invalidated by ‘new information’ based on the new scenarios. Likewise, the financial enterprise would start to build a ‘library of knowledge’ that could be leveraged across the organization in unique ways, building additional tools for risk management. And in designing proactively for the next CCAR submission, AI combined with Robotic Processing Automation (RPA) could also decrease time to gather and analyze information, resulting in cost savings.

If the submitting financial enterprises can benefit from AI review, the oversight body, the Federal Reserve System, could be a major beneficiary of cognitive computing. In 2016, the General Accounting Office (GAO) noted that the Federal Reserve must do a better job of providing greater transparency on how the grades are applied against the seven principles: 1) Sound foundational risk management; 2) Effective loss-estimation methodologies; 3) Solid resource-estimation methodologies; 4) Sufficient capital adequacy impact assessment; 5) Comprehensive capital policy and capital planning; 6) Robust internal controls; and 7) Effective governance.

By using cognitive AI, the Federal Reserve could objectively ‘read’ the CCAR findings from each financial institution. As a result, the objective ‘eye of AI’ would see similar trends and spot inconsistencies in each submission, as well as broadly across the reporting spectrum. For example, each of the seven principles could be analyzed as stand-alone items across all the submissions. Moreover, the Federal Reserve would have an internal tool to more effectively evaluate how each of the Fed’s bank examiner teams has performed. Could it be that one team is more lenient than another team? Likewise, do certain regions of the country rank higher, and if so, then why? AI can reveal heretofore hidden trends by extracting, aggregating and then analyzing key information. Any time there is an avalanche of unstructured and structured content, AI can ‘make sense of it’, even CCAR content.

John Paty
Vice President, Expert System
jpaty@expertsystem.com

 

Expert System’s Artificial Intelligence solutions have been deployed in many of the world’s largest banking and financial services institutions to support intelligent process automation and knowledge discovery, including knowledge management and content enrichment, investment research, customer care and support, risk mitigation and anti-money laundering compliance. To learn more about Expert System’s Financial Services and Banking offerings visit: https://www.expertsystem.com/industries/banking-insurance/