Banks detect roughly 2 percent of global financial crime flows, despite dedicating 10 to 15 percent of their workforce to anti-money-laundering activity, and as much as 85 percent of that effort is administrative rather than analytical (McKinsey). That is not a story about banks failing to try. It is a story about where all that effort actually goes: into producing records, not into building anything the institution keeps once the person who did the work has moved on.
Compliance already costs enough
Financial institutions across EMEA, including Saudi Arabia among the markets surveyed, spent an estimated $85 billion on financial crime compliance in a single year, with costs rising at 98 percent of institutions and labor costs cited as the leading driver by 72 percent of them (LexisNexis Risk Solutions, Forrester Consulting 2024). Technology spending tells the same story: 70 percent of EMEA institutions saw compliance and know-your-customer technology costs rise over the same period, and the figure climbs to 74 percent in the Middle East specifically (LexisNexis Risk Solutions, Forrester Consulting 2024). Adding headcount, or adding software, to a problem that is already consuming 10 to 15 percent of the workforce is not a growth strategy. It is a sign that the effort itself is not compounding into anything durable.
What actually breaks is not effort. It is memory
TD Bank’s 2024 anti-money-laundering penalty, one of the largest in US banking history, traced back to a monitoring gap that quietly compounded across 92 percent of the bank’s transaction volume for years before it was caught. Danske Bank’s own $2.06 billion penalty followed a similar pattern years earlier. Neither failure was a single missed alert. Each was a blind spot nobody could see accumulating, because the reasoning behind individual decisions was never captured anywhere a person, or a system, could later review it. Replacing a single AML analyst costs between $75,000 and $150,000 once recruitment, training, lost productivity, and knowledge transfer are counted (WorkFusion). What that figure does not capture is the part that cannot be rehired: the judgment a departing analyst applied to the last hundred borderline cases, none of which was written down anywhere durable enough to outlast their employment.
Saudi Arabia already requires the record. It does not require the reasoning
SAMA’s Anti-Money Laundering Law already requires banks to maintain customer due diligence files and transaction records, and to keep that documentation current as customer risk profiles change. That satisfies the letter of the requirement. It does not solve the underlying problem, because a retained record shows what happened, not why a specific reviewer decided it was acceptable, what pattern they recognized, or what exception they made and under what condition. When that reviewer leaves, the record stays. The reasoning does not.
Decision memory means the reasoning survives the person who made it
A system built to retain decision memory does not just log that a transaction was cleared. It captures the exception, the condition that triggered it, and the judgment applied, so the next case that looks similar can be handled the same way without a human reconstructing the logic from scratch. A sanctions-screening override, a credit exception granted under a specific circumstance, a reconciliation break resolved a particular way: each becomes part of what the bank knows, not just what one analyst happened to remember. This is engineering work built inside the bank, on the bank’s own data and under its own policies, not a tool bolted on afterward: it takes the exceptions and corrections a compliance team already makes every day and turns them into a durable record the institution keeps, independent of who made the original call. Meeting a record-keeping requirement was never the hard part, and no bank needs help doing that. Retaining the judgment behind even one analyst’s decisions for the length of their tenure, so it does not leave when they do, has been the actual gap all along.
None of this reduces headcount to zero, and it should not try to. It reduces how much of that 85 percent of administrative effort has to be repeated every time someone new joins the team, and how much of a bank’s compliance capability quietly walks out the door every time someone leaves it.
That is a different kind of compliance capability than the one $85 billion a year is currently buying across the region. It is also the kind that ambition, not just cost control, requires. Once a decision is retained rather than just logged, the next question follows immediately: can the bank explain that decision to a regulator, in full, on demand.
Schedule a Meeting
Sources
McKinsey & Company. “How Agentic AI Can Change the Way Banks Fight Financial Crime.”
LexisNexis Risk Solutions and Forrester Consulting. “True Cost of Financial Crime Compliance Study, Europe, the Middle East and Africa.” March 2024.
WorkFusion. “Hidden Costs of AML and KYC Operations.”
SAMA (Saudi Central Bank). Anti-Money Laundering Law.