In 2015, the UK government announced the workplace pension scheme, sometimes also called the occupational or work-based pension. The scheme mandates that employers provide pension plans for all employees. The scheme is good for employees and, admittedly, creates new business for insurers. However, it also created an unprecedented problem: a vast number of employees suddenly started registering for pension plans and had to be quickly cleared using Anti Money Laundering (AML) checks and a Jurisdiction Index Check (JIC). The index ranks jurisdictions according to their secrecy. It helps understand tax havens and the flow of capital. These processes are critical to regulators. The reason is simple: the insurance industry offers the flexibility to park large sums of money and recover them with relative ease - a fact that is attractive for money launderers, terrorist groups and underworld operators processing dirty funds. Traditional due diligence to measure risk and combat financial terrorism, naturally, falls short of these sophisticated requirements.
The AML process checks the identity of individuals and organizations carrying out financial transactions. It screens them against various global watch lists. The AML process also has to be used across national and regional jurisdictions. This makes AML complex, time consuming and expensive. Failures in AML processes leads to hefty regulatory penalties. For a leading insurance provider that was providing the workplace pension scheme to employers, the status quo was not maintainable. More importantly, it did not snag every suspicious business entity. A solution had to be found – and it was Artificial Intelligence (AI) with their existing BPM platform that came to the rescue.
Gaps in existing AML
The traditionally used manual AML process looks like this: an employer registers on an insurance portal and provides details of his/her organization. The AML and JIC checks are done manually with the operations team going offline to research data on the company and see if it is trustworthy, if it has branches outside the UK, if it has subsidiaries that may not be AML compliant, if director or owner or board member is politically exposed, etc., and then come back with “yes/no” and “accept/reject” responses to the workflow.
A typical, operational AML system has several components ranging from business process management to content management and analytics. However, these systems are often not linked and are quite dependent on human intuition and skepticism for determining if a customer is compliant with AML process. Most existing AML systems cannot provide automated insights and action.
AI for better control
Today, AI systems are becoming increasingly popular within organizations for functions like BPM. These AI systems leverage technologies like Robotic Process Automation (RPA). Aside from producing data, insights and decision support from unstructured and scattered data, these AI systems also learn and adapt, evolving continuously towards higher levels of performance. These systems run an infinite ‘learn-analyze-predict-decide-act’ loop that makes them more intelligent with every iteration.
In this scenario, a potential approach is to leverage BPM and boost it with AI capabilities. BPM streamlines AML workflow orchestration with product-specific business rules to onboard and maintain multi-product and multi-jurisdictional businesses. BPM capabilities like intelligent routing, case management, and escalations allow users to maintain and audit more recent regulatory rules. BPM provides a connected platform for conducting AML and embedding AI into it, with support for integration with external systems for automated checks.
Wipro’s recommendation to the insurance customer was to use AI to guide the AML team. Bots were scripted to mimic human actions for processes such as opening a website and extracting ontology-based information (registered name of a business entity, the registered address of the entity, the board of directors, type of ownership, etc). The bots looked for information based on the profile of the business entity. If it were a recruitment firm being scrutinized, the bots would examine Facebook or LinkedIn for information. This is very different from a law firm being scrutinized - for which the bots would examine news sites. Checks and balances were placed by providing adequate control to human operators who would be presented the information gathered by the AI. The insurer’s AML team could then, whenever necessary, override decisions made by the AI.