An anti-money laundering system.

Context

Anti-money laundering is not a reporting exercise. It is a continuous operational process, staffed by people who have to review, decide and document, inside timeframes and standards set outside the bank. The software that supports it is used every working day by teams whose decisions carry regulatory and reputational weight.

This engagement involved designing and building an AML system for a top bank - an environment where the system sits alongside core banking infrastructure and established enterprise processes.

Why this class of system is demanding

The difficulty in AML software is rarely a single algorithm. It comes from the combination of constraints that have to hold at once:

• Breadth of inputs. The relevant information is spread across customer records, transaction data, and reference information that each live in different systems with different update cycles and different definitions of the same entity.
• Workflow, not just detection. Producing a signal is the start. The system has to carry that signal through review, decision, escalation and documentation, with every step attributable and reconstructable later.
• Operational reliability. A compliance workflow that stops working is an operational incident, not a backlog item. Availability, recovery and data integrity requirements follow from that.
• Usability under volume. The people using the system work through queues. Interface decisions that look cosmetic have a direct effect on how much work a team can complete and how consistently.
• Change over time. Rules, thresholds and expectations change. The architecture has to allow that without a rebuild each time.

Why it is relevant to other work

Most of what makes AML software hard appears in other mission-critical enterprise systems: multiple sources of truth, a human decision in the middle of an automated process, an audit trail that has to survive scrutiny, and no acceptable failure mode where the system quietly does nothing.

That is the same problem shape we work on in other sectors - a process that matters, spread across systems that were not designed to cooperate.