Situation
Banks are required to monitor clients and their transactions for signs of financial crime, but the high volume and variety of signals make manual planning challenging. Where private individuals might conceal criminal activity through large cash or cryptocurrency transactions, companies could try to circumvent sanctions with creative company structures. Each form of potentially criminal activity requires specialised investigation by the financial institution.
To improve operational efficiency in these investigations, banks often arrange their analysts in specialist teams and assign specific types of investigations to each team. With millions of clients and a multitude of transactions, this results in a complex planning problem.
Our client, a Dutch bank, had its own manual planning and distribution mechanism. But as the financial crime department grew rapidly – due to increasing regulatory requirements – it became difficult to do this manually. With a shortage of skilled analysts to do these investigations, the backlog grew, as did the pressure to stay in control of the backlog.
So how could the bank prioritise actions? While it is important to take a risk-based approach, there are also KPIs that stipulate the time within which signals need to be processed. For example, a sanction circumvention signal has a high risk-based priority, but a lower-risk cash transaction that has already been on the backlog for a couple of weeks needs to be picked up, too, at some point. What if the backlog for one team piles up, whereas another team’s backlog shrinks? And there might be a propensity for analysts to cherry-pick easy cases and leave complex cases on the backlog. There are lots of considerations.
Approach
All these questions were addressed in the smart orchestration tool that Valcon developed and implemented together with the client. The tool manages the workflow for two types of incoming work: mandatory periodic client reviews and event-driven signals. Periodic reviews for low-risk clients – which constitute the majority of the client base – consist of repetitive tasks. Together with RPA (robotic process automation) teams, Valcon automated these reviews using robots that check the consistency and recency of client data between sources.
On the other hand, event-driven signals require manual investigation and must be assigned to a specialised team. The tool aggregates event-driven signals before distributing them to teams by combining operations research techniques with simple but effective business rules.
For example, a company might start a new business in another country, causing both a change in organisational structure and a change in transaction patterns. These two changes trigger two separate signals in the bank’s monitoring platform. With the previous system, this could result in two investigations assigned to different analysts. Using graph theory, our tool recognises the signals are related and allocates them to the same analyst. This is more efficient, because just one analyst is apprised of the context and deals with the whole situation.
The success of the tool is also attributed to its integration with various operational systems that are used by analysts. Although the detection of financial crime efforts is centralised, client units still have their own IT systems and processes. Rather than building tailor-made solutions for each client unit, Valcon has built standardised solutions to which various client units, business lines or subsidiaries can be onboarded. Customisation remains in the connection to end-user systems, but not in the planning solution itself.
Results
The Valcon tool makes financial crime detection more efficient in three different ways:
- Automation of periodic reviews: robots can now quickly perform reviews that would otherwise have taken thousands of hours of repetitive manual work.
- Efficient allocation of manual work: our tool enables the analysts to pick up more files per week than they could otherwise do. Without the tool, files would have changed hands multiple times before they land at the right team, or multiple analysts could be working on related cases.
- Automated file distribution: the planning and distribution of files themselves is automated, and subject matter experts can spend their time on process optimisation rather than manual file distribution.
Further automation now is the name of the game, and together with the client, Valcon is exploring what can be automated next.
