Human AI combo strengthens fight against financial crime
Money laundering is known as ‘the crime that fuels other crimes’ and encompasses various schemes that bad actors use to wash dirty money and make it clean. In theory, stemming the flow of capital that appears to originate from legitimate sources, but doesn’t, is an effective way of damping down criminal activity. But this is easier said than done. The United Nations Office on Drugs and Crime (UNODC) estimates that the amount of money laundered worldwide annually is in excess of 2% of global GDP ($1.7 trillion, in today’s figures). And governments have been enacting anti money laundering legislation since the 1970s. But there is something that could give criminals pause for thought – artificial intelligence (AI), with its capability to digest big data, combined with human-in-the-loop curiosity.
Trillions of data points
Without algorithmic help, anti money laundering teams will struggle to cope with the sheer number of financial transactions that happen every day. The spending patterns of a global population of 7.9 billion people could easily run into trillions of data points over a year. And to separate the good money from the bad, analysts need to consider additional factors too. This includes regional trends, whether the payment is being made by an individual or on behalf of a company; together with other targeted queries.
In 1990, The Financial Action Task Force (FATF) – an intergovernmental body – provided financial institutions with a list of 40 essential measures to combat the misuse of financial systems for laundering drug money. Subsequently, these have been expanded on to address the funding of terrorist acts and terrorist organizations; and to stem the proliferation of weapons of mass destruction. FATF’s mandate highlights that laundered money flows to bad places. And while in a Hollywood movie it might be actors on a dynamic chase that stop the villains, in reality, it’s financial institutions that have the power to shut down the funding of illegal activity.
Financial crime toolkit
Helping analysts to comply with anti money laundering recommendations – which includes an obligation to Know Your Customer (KYC) – are a variety of financial crime tools that bring AI into play. As numerous examples have shown, AI and machine learning algorithms are very effective at generalizing large data sets and identifying patterns of normal versus anomalous behavior. Today, AI-powered solutions let analysts monitor financial details more rapidly than has been possible previously. But relying on algorithms alone can leave gaps in a financial institution’s defenses. “Criminals are creative, computers are not,” Francisco Mainez, Product Manager at Lucinity, told TechHQ.
Financial institutions have to constantly sharpen their ability to assess risk and anticipate threats, which requires human skill sets as well as computational expertise. After beginning his career in the Spanish Marines and working in military intelligence, Mainez brought his experience in human intelligence systems and data management across to the financial sector. His previous roles include positions at Standard Chartered Bank and HSBC. And what Mainez found particularly appealing about Lucinity’s approach to fighting financial crime is a solution that’s being called ‘augmented intelligence’.
How can HR policies tackle AI bias?
Recognizing the benefits that both humans and machines bring, the company provides software that aims to bridge that gap and deliver the best of both worlds. It’s a concept that has appeal. Asking staff to pore over hundreds of spreadsheets every day doesn’t play to human strengths. But getting AI to do the heavy lifting and prioritize cases for review, puts analysts in a much stronger position to begin their investigations. “Humans can think out of the box,” said Mainez.
As data scientists know all too well, one of the challenges in applying AI and machine learning is the amount of preparation that’s required. Data can be messy, and analysts can spend as much as 80% of their time preparing inputs, leaving only 20% for more important analytical tasks. Lucinity’s API, which supports anti money laundering software as a service (SaaS), strives to flip things around so that users spend 80% of their time on the analysis instead. And the platform has been configured to make ingesting data, integration, and publishing steps as simple and seamless as possible.
To picture the software in operation, let’s consider a hypothetical scenario – for example, the system raises an alert that customers are paying off their mortgages too quickly. Humans in the loop could then step in and put their creative thinking to work – a task made easier as various data has been integrated and is at analysts’ fingertips. Queries could examine the job titles and financial history of account holders – are they students; are the mortgage repayments commensurate with their income? Regional factors could be considered too – for example, do banks in that country encourage early mortgage repayments?
To guide analysts and streamline the process, the workflow is structured across multiple layers. But no design choices are made at the expense of removing human creativity – the element that could prove critical in upsetting the plans of bad actors. Humans and machines working together enable wide coverage and adaptability to changing conditions. Anti money laundering systems were put to an extreme test during the global pandemic as national lockdowns and other restrictions meant that spending patterns had to be relearned. But the software proved to be sufficiently flexible – for example, in being able to distinguish legitimate food delivery payments from transactions made to drug dealers masquerading as take-away vendors.
Anti money laundering software combining AI with human creativity – featuring intuitive user interfaces and accessible information – could help financial services providers work more effectively with law enforcement. Suspicious activity reports (SARs) raised by banks, solicitors, accountants, estate agents, and other stakeholders are critical in providing law enforcement with intelligence from the private sector that would otherwise remain off the radar. And ‘augmented intelligence’ can contribute by removing false positives and increasing the yield of actionable intelligence.