The impact of modern technologies such as artificial intelligence (AI) and machine learning on job security has been a widely discussed topic today. Of course, AI has advanced very quickly in the last few years and evolved to outperform humans in several areas. Technology has also proven its ability to do astonishing tasks that humans cannot do. There are many reasons to think that employers will replace humans with AI-enabled machines, which are more economical, leading to the creation of an ever-growing pool of unemployed humans. At the same time, there are opinions that this worry of AI-induced job losses is unfounded. Some say that technological advances are not a new phenomenon. From the steam engine to electricity to the internet, the chain of technological advances has created more jobs than they eliminated, according to them. They argue that AI and machine learning will pave the way for new careers we cannot imagine.
AI in regulatory compliance
If we take the case of the BFSI sector, and specifically the regulatory compliance function in it, the AI wave has started working but not in full measure. Having learned a lot from the 2008 financial crisis and being burdened by more stringent regulatory requirements, banks are really in need of efficiency and effectiveness improvements. As one of the heavily regulated sectors across the globe, banking cannot easily entrust machine learning with a seriously important task such as compliance. However, pilots and small projects focusing on machine learning-powered solutions, commonly known as RegTech solutions, have been running on bank premises with a view to testing their efficacy. These solutions are primarily applied in the areas of suspicious transaction monitoring and alerting.
Machine learning and compliance job
Before discussing if machine learning will impact jobs in the compliance area, we will need to understand the actual tasks of compliance professionals. Most banks have a rule-based compliance management system that produces alerts based on set transaction thresholds. The compliance staff at a bank reviews these alerts for possible anomalies and trends. They also look for emerging suspicious financial crime patterns and the possibility of reducing false alerts. The compliance personnel also make incremental adjustments, in tune with their company’s risk tolerance, in order to improve the efficiency of system alerts. This task requires a lot of experience and expertise. Rule-based systems typically follow a unidimensional approach where they produce alerts based on transactional value and more than 90% of alerts generated are false. The sole reliance on internal expertise is also risky as there are chances to miss less frequent and emerging typologies. In addition, updating of rules to reduce false positivesis often a slow process as it has to be done manually, and banks are currently not in a position to afford that.
The promise of machine learning to the compliance function is significant efficiency and effectiveness improvements in terms of holistic risk coverage, reduced false alerts and better detection of suspicious activities. Given the glaring inefficiencies, even a marginal improvement with machine learning will help banks save huge amounts of money. Financial institutions can train supervised machine learning algorithms on prior alerts to update or fine-tune rules and reduce false alerts. Supervised algorithms can compare current rules with previous investigation results to make changes in the rules, and this is a continuous automated process running the background. The use of unsupervised machine learning algorithms can effectively help identify new typologies without the help of rules. Some solutions as the Tookitaki Anti-Money Laundering Suite offers semi-supervised machine learning, which combines the best of both supervised and unsupervised algorithms.
The use of machine learning has proven to increase efficiency and effectiveness at many banks. For example, Tookitaki’s 6-month long pilot project involving its machine learning-powered solution at United Overseas Bank saw a 60% and 50% decline in false positives for individual and corporate names, respectively, for name screening alerts. In addition, transaction monitoring saw a 5% increase in true positives and a 40% drop in false positives.
Compliance Jobs at Risk? Not at the Moment
As of now, full automation of compliance tasks at banks with the help of machine learning is not done. Machine learning-enabled RegTech solutions have the capability to eliminate low-level, repeatable, manual processes but we are not sure if banks have dedicated posts for these processes. Machine learning solutions are not fully replacing all human involvement in compliance processes at present rather they assist compliance professionals to be more productive and quicken decision making. Full automation of compliance tasks is still a distant dream as machine learning solutions are still not mature. They need to undergo further testing, research and refinement to attain their full capabilities. Therefore, AI cannot be the compliance decision-makers at financial institutions at least as of now.
However, compliance professionals need to equip themselves and adapt to the change in order to stay relevant. Future compliance officers should be good at both the technology and business sides of the compliance job. They should have a proper understanding of how a machine learning solution works. Also, they should be able to explain to authorities the reason for the system alerts and how the solution reached into a conclusion with regard to a particular suspicious case.
