London's business center in the 16th century often used the expression of a "black swan" to refer to impossibilities, simply because scientists had not documented the sighting of many such animals. After Dutch explorers first sighted black-feathered swans in Australia in 1696, they were surprised and took the birds home. Since that day, it remained a way to describe unexpected events. The black swan theory has been further developed by Nassim Nicholas Taleb and is often cited in a macroeconomic context and to approach random and high-impact risk in financial markets. The key piece of advice regarding a black swan event is that you cannot predict it, nor can you prevent it. You can only prepare for it.
Many economists and financial professionals working in an industry of risk and correlated numbers find comfort in rationalizing previous events and integrate them into future prediction models. Learning from the past and planning for the future are the adages that drive modern economics. We can find these mechanisms in trading, supervision, risk management, fraud management and even in anti-money laundering controls.
But what if we are ready for a new approach? A paradigm that steps away from the rule-based mindset the industry has embraced for decades. If we look at the example of financial crime, we can already say that the past behavior of criminals that are making illicit payments are not a guarantee of future behavior or the start of a pattern. But then what is? Perhaps the answer to that question is to ask another one: What if the ultimate predictor does not exist? Then the first step would be to stop looking for the unknowns, and rather aim for what we already know. Banks and payments providers, being data-driven entities, are already very capable of gathering information on their customers and beneficiaries. They are overseeing legitimate data points that can help to improve critical and data-intensive functions such as fraud management, compliance and transaction monitoring. Machine learning and artificial intelligence can unlock the value of these core capabilities and do not require the industry to reinvent the wheel. What is more, monetizing and investing in concrete AI opportunities within highly regulated environments such as transaction monitoring sets the bar high while opening doors for the rest.
A closer look at the opportunities shows that the use of machine learning in this space is increasing. In a recent
Industry investment and cooperation is needed but regulatory support is more urgent. The regulatory stance on data privacy and machine learning adoption are currently moving in different directions. On the one hand, U.S. financial supervisors, the Federal Reserve and Federal Deposit Insurance Corp. are pushing banks to improve internal controls and risk infrastructure, while regulatory agencies such as the Consumer Financial Protection Bureau and the Data Protection Authorities in the U.S. and Europe issue careful considerations on the decision-making capabilities of artificial intelligence. Transparency is key in overcoming uncertainties. Knowledge sharing between technologists, AI advocates and policymakers is very welcome, but it would be even better to focus on the daily operational challenges experienced by many back-office workers.
