When disputes or regulatory issues arise around fraud, bribery or other forms of malfeasance, outcomes hinge on data. That data must be readily available—a task that requires not just monitoring, but also the ability to predict and detect potentially anomalous patterns. As a compliance or risk executive, your job is not only to protect the organization from risk, but to stay one step ahead of it. That’s why it’s essential to make optimal use of the increasingly powerful (and cost-effective) emerging technologies available today.
Machine learning is a subset of artificial intelligence (AI) that utilizes algorithms and computer power to sharpen the judgments organizations make about voluminous and disparate data. Simply put, it allows machines to learn how to perform certain tasks without being explicitly programmed to do so.
Machine learning already permeates virtually every facet of our lives—it is used every day for image and speech recognition, digital assistants, cyber protection, consumer marketing, medical diagnoses, ferreting out proscribed content on social media platforms, ride-sharing apps, law enforcement and in countless other applications.
Machines could learn how to detect fraud simply by being provided with examples of previously seen fraud cases, and without the need for manually coded business rules. The advantages of such a data-driven modeling approach over traditional business rules are significant:
Machine learning helps optimize the mix between humans and machines in an intelligent, accretive process that “learns” as it goes along. If implemented appropriately, machine learning can help reduce risk, cost and customer friction (by minimizing false positives and blocked legitimate transactions).
The financial services industry—which, given its business model, is heavily exposed to fraud risk—has led the way among industries in leveraging machine learning technology. Its algorithms and data labels pointing to (or predicting the likelihood of) credit card fraud, money laundering and other crimes are relatively mature because of the volume of data processed over time.
Machine learning holds tremendous promise in addressing bribery and corruption risk as well—offering compliance and risk teams a significant boost by processing, identifying and tiering potential anomalies (or “exceptions”) that may be hiding in their data. It can scour transactional data and communications for traditional corruption markers—duplicate payments, improper relationships, offshore bank accounts and the like—and prioritize accordingly.
This technology can also be extremely effective in an M&A context—where an acquiring company needs to quickly process a large number of potential exceptions from a target company or legacy technology. Machine learning can help your team quickly tier these exceptions for faster human action.
Partner, PwC US
Principal, PwC US
Principal, Cybersecurity, Privacy and Forensics, PwC US