Looking forward in 2021, banks are under increased pressure to review large volumes of documents within short timeframes. Reasons for this are numerous. For example, due to a pending change in the sunset of the London Interbank Offered Rate (LIBOR), banks must review hundreds of thousands of LIBOR Adjustable Rate Mortgage (ARM) contracts to classify the fallback path and other terms.

Since LIBOR impacts a large universe of products in addition to mortgages, banks must plan for the recast of every in-scope contract to their LIBOR alternative. In other situations, such as qualified financial contract (QFC) remediation under resolution planning, banks must evaluate large groups of counterparty and liquidity agreements with complex terms, often with substantial cross-linking. A third driver of document review is loan forbearance in retail and commercial banking, as lenders grapple with a myriad of loss mitigation challenges, sometimes resulting from bank consolidation.

Purely manual file review solutions are not ideal in these situations because they add to project costs and duration, as file reviewers can only review so much contractual language during a given timeframe. In addition, the increasing complexity of document review often implies more significant staffing. You can scale expertise; however, improving efficiencies becomes challenging beyond a certain level. Consequently, manual reviews are more closely associated with more extensive fixed staffing than other, more dynamic approaches.

In comparison, fully automated document reviews with no manual intervention, which may be desirable in theory, are often difficult to attain and prone to error. Limiting factors include the consistency of file review language to be cataloged, low file quality, and the number/complexity of search terms to automate.

An automated solution that incorporates optical character recognition (OCR), for example, often only yields 80-plus percent accuracy without complementary manual review to provide ‘human correction.’ In addition, while machine learning (ML) and artificial intelligence (AI) elements can provide a lift, full reliance on these technologies to auto-decision complex outcomes may be too daunting to consider, given their potential legal and/or other consequences.