By Patrick Craig, Partner, Ernst & Young LLP, EY EMEIA Financial Services Advisory and Financial Crime Leader and;
Mark Gregory, Senior Manager, Ernst & Young LLP, EY EMEIA Financial Services Advisory
Introduction
Combating money laundering poses a major challenge for the financial-services industry and beyond. The United Nations Office on Drugs and Crime estimates that the annual cost of money laundering and associated crimes runs anywhere from US$1.4 trillion to $3.5 trillion a year.Every day anti-money-laundering (AML) professionals face increasingly complex threats and are tasked with analyzing a growing volume of data. Yet even as they strive to counter real-time threats, they are overwhelmed with the high levels of manual, repetitive, data-intensive tasks that are inefficient and often fail to disrupt criminal activity.
Can artificial intelligence (AI) provide an effective solution to the imbalance between effort vs impact, transforming AML compliance, as it promises to do for so much of the financial-services industry? (AI spans a breadth of fields, techniques and technologies and as such can be hard to define. For the purposes of this article, we adopt the Financial Stability Board definition of AI as “the theory and development of computer systems able to perform tasks that traditionally have required human intelligence”. Machine learning is a key sub-field of AI, and it is developments in machine learning that have powered many of the recent successes of AI.)
Certainly AI has the potential to enable a step change in AML capability and provide a means to scale and adapt to the modern threat of money laundering. Yet, at the same time, many in the industry are sceptical about the effectiveness of AI solutions and the extent to which AI could and should be trusted like human capabilities. To fully explore and realize the potential of AI, the financial-services industry needs to better understand its capabilities, risks and limitations, and establish an ethical framework through which the development and use of AI can be governed so these emerging models can be proven and ultimately trusted.
The current anti-money-laundering landscape
The financial-services industry has evolved significantly since the Financial Action Task Force (on Money Laundering) (FATF) published its first recommendations in 1990. Since that time, an increasingly varied and complex set of products and services, globally connected markets and digitally enabled channels have inadvertently enabled highly organized and sophisticated money-laundering activity to proliferate and diversify. For example, analysis has highlighted the growing threat of high-end money laundering and its links with cybercrime, organized fraud and professional service providers.
As a result, both the industry approach and regulatory framework are failing in the fight against money laundering. Despite the vast resources deployed by financial institutions to combat money laundering, the current approach is not delivering sustainable results. Money-laundering activity is estimated to be between 2 and 5 percent of global gross domestic product (GDP). In the United Kingdom alone, the social and economic cost is estimated to be least £24 billion a year. According to a report by Europol, around 10 percent of the Suspicious Activity Reports (SARs) filed by financial-services institutions lead to further investigation by competent authorities. Worse still, Europol estimates just 1 percent of criminal proceeds in the European Union end up being confiscated by the authorities. Yet this inefficient system of AML accounts for an estimated 4 percent of annual financial-services revenue, and tallied up a global industry spend of around $8 billion in 2017.
As financial-services companies struggle to contain compliance costs and the threat of money laundering continues to evolve, it’s no surprise that business leaders are seeking change and innovation in AML.
How can AI help the financial sector to fight money laundering?
AI has been a field of research since the 1950s; however, its capability has grown rapidly in recent years—underpinned by advances in computing, greater availability and quantity of data, and increased AI research and development. Interest in AI has risen significantly, driven by greater awareness of the capabilities and applications of AI, such as virtual assistants and robotics across industries as diverse as health care, government and manufacturing.
Current compliance processes are dominated by high levels of manual, repetitive, data-intensive tasks that are both inefficient and error prone. The AML technology that supports these processes relies heavily on expert systems that, in general, have not advanced since their introduction more than a decade ago.
Common AML pain points for organizations are typically the high caseloads and human effort involved in customer due diligence, screening and transaction-monitoring controls. AML transaction monitoring has been a particular problem area for many banks and has come under significant criticism in recent years. Existing transaction-monitoring controls typically generate high levels of false positive alerts and significant operational workloads. The cost issue is often further amplified by inefficiencies in the investigation process, which create a low return on the effort employed versus the impact of transaction-monitoring controls.
Many of AI’s capabilities can alleviate these pain points. They include the ability to:
-
Drive insight and value from large volumes of complex data that are often involved in due diligence, risk assessment and monitoring activities, leading to better risk outcomes.
-
Learn from and adapt to changing environments and inputs, helping firms to keep up with the rapidly changing financial landscape and risk profile.
-
Automate repetitive tasks that are currently handled by humans, operate at scale and take decisions at speed to reduce costs and focus human engagement where there is the highest value added.
-
Reduce error and improve consistency in processes and decision-making.
Much has already been written about the threat AI poses to the current workforce. Despite AI’s obvious strengths, however, industry experience (particularly in the manufacturing and engineering sectors) has shown that machine learning performs best in collaboration with human insight. The combined capabilities of human insight and machine learning can drive more effective solutions than deploying humans or AI in isolation.
The increasing use and understanding of how AI could be applied and integrated with human activity in AML is driving new thinking in the important process of know your customer (KYC). AI could bring increased breadth, scale and frequency to existing holistic KYC reviews in a way that better integrates ongoing screening and monitoring analysis. Risk and detection models can assess and learn from a richer set of inputs and produce outcomes in the context of both the customer’s profile and behaviour. By leveraging AI’s dynamic learning capability coupled with skilled investigators, this model could be used to augment operations, provide quality control and even be used to train new resources.
So what’s holding AI back?
Despite the obvious attributes that AI offers AML, its adoption in this crucial part of the financial- services industry is still very much in its infancy. So what are the barriers that AI must overcome if it is to fully deliver on its promise?
Certainly the current low levels of maturity, industry adoption and regulatory guidance on AI aren’t helping an industry that is deliberately cautious by nature and with rules of compliance that create hurdles when it comes to innovation. At the same time, many in AML are uneasy about the increased technical complexity and reduced transparency of AI solutions, and there are also concerns about the data quality and the specificity and consistency of intelligence used to train AI. The need for new infrastructure, technology and talent poses significant problems as does the level of change and disruption to operating models and business processes that integrating AI brings. Underlying all these practical concerns is an existential threat—the potential unknown and unquantified risks of embracing the machine.
Building trust to meet compliance
For AI to make the leap into the AML mainstream and win trust from financial institutions, it will need to prove it can meet the many challenges posed by AML compliance. To do so, financial institutions will need to shape their AI AML operations with the following considerations in mind:
-
Institute strong governance in order tomanage risk and enable the necessary levels of understanding and documentation that will inform effective decision-making across the life cycle of an AI solution.
-
Define scope, objectives and success criteria at the outset in order to establish clear performance indicators and parameters that link to a well-defined risk-appetite statement. This will be critical to tracking whether the outputs from the AI are meeting objectives at an acceptable level of risk and in helping stakeholders mitigate the risk of unintended use and outcomes, as well as appropriate and fair use of data.
-
Make the AI design transparent in order to effectively surface key design decisions, assumptions and known limitations.
-
Collaborate with multiple stakeholders including firms, vendors, regulators and governments in order to define best practice. Collaborative efforts can underpin wider adoption and identification of further benefits but also set standards for appropriate governance and controls to manage the safe development and deployment of AI-enabled solutions.
-
Focus on data inputs and ethical implications to minimise bias and mitigate concerns that AI will increase AML sensitivity to poor data quality—already a major challenge for many financial institutions.
-
Apply robust testing and validation, engaging early, deploying incrementally and reviewing regularly to ensure effective integration into business processes, as AI can bring significant disruption to compliance processes and the institution’s operating model.
Ultimately, it’s clear that the current AML approach is struggling to keep pace with modern money-laundering activity and that AI offers a great opportunity not only to drive efficiencies but to also identify new and creative ways to tackle money laundering. And while AI continues to pose challenges and test the industry’s appetite for risk, the question all financial institutions should be asking is: Can we afford not to embrace AI in our AML, and get left behind? Ultimately, when integrated with the right strategy and with the right focus on building trust, innovating with AI must be seen as a risk worth taking.
The views reflected in this article are the views of the authors and do not necessarily reflect the views of the global EY organization or its member firms.