Artificial Intelligence or “AI” is the new buzzword in Anti-Money Laundering. Nervous?
Recent AML news would have you believe we’re on the cusp of dramatic change that will wipe out thousands of compliance jobs. AI will then transform those few remaining workers into Arnold-like Model T-101 Terminators annihilating alerts, cases, and SARs with emotionless robot-like precision.
My take is different. There are two likely outcomes of AI.
- AI will cure disease, advance all science and then lead mankind to destroy itself.
- AI will create more AML work, not less.
Let’s gloss over the first of these inevitabilities. Once super AI is created and machines actually learn at speeds incomprehensible to human understanding, two things will happen: First, machines won’t need us anymore, and second the political, economic, and social upheaval caused by AI will lead to cataclysmic war and famine.
But don’t worry, experts say this “super AI” is still probably 50 years away.
AI Impact on AML
For AML, the expectation is that AI will cure the overwhelming inefficiency the plagues us. Billions of dollars are wasted as AML teams remain buried under false positives, hunt down information from clunky old databases, and make inconsistent, and sometimes really bad decisions.
An AI driven system that can gather data instantly and be programmed to make decisions when given a set of simple facts will no doubt change AML. But this “weak AI” as it is known, won’t reduce AML workloads, it will actually increase them.
Let’s start with something everyone in AML agrees about: There is too much time spent on finding stuff. Take alerts and cases, for example, where 80% of the time on each matter is spent not on actual analysis and investigation, but on gathering transaction records, KYC information, and risk information such as negative news. Or how about the AML risk assessment, which is just mostly a multi-month (or yearlong) data gathering exercise. If AI can eliminate this wasted time and gather needed information in seconds, then indeed certain AML work will be reduced.
It wasn’t that long ago that the smart minds of science predicted the world was soon to run out of oil. The last 10 years has made it obvious that this theory, known as the “peak oil,” was utterly wrong.
A similar view, call it “peak work,” is what creates the hope that AI will transform AML: All the work that needs to be done in AML is already being done, but this “real” work is overtaken by the “busy” work. AI will eliminate the “busy” work, and we will finally be able to focus just on what remains (the “real” work) and banks will then start shrinking AML departments.
There is a problem with this thinking. The “real” work of AML hasn’t even yet begun.
We Fail to Detect Most Money Laundering
The U.N. estimates that up to $2 trillion dollars a year is laundered.
Add together total money seized by credible law enforcement agencies from around the world each year and the amount of funds involved in activity reported on SARs and you get much, much less than the actual amount of money laundered. The “real” work is to find this money.
So yes, weak AI will clear away a lot of the clutter, and it may even be able to report – without the need for human input – simple suspicious activity like cash structuring.
But imagine the activity that will be uncovered once the problem of disparate and inefficient data gathering is solved. What will happen in banks when unknown customer connections are revealed or when data from innocuous looking wire transfers exposes critical information that up to this point is hidden?
Instead of shrinking AML departments, financial institutions will instead be hiring more investigators to deal with the flood of new, heretofore unknown (and more complex) suspicious activity. And despite what AI developers may say, computers will not spontaneously learn how to analyze and investigate all this new activity. It will take years for weak AI to work in banks and then it will take years for this weak AI to become super AI (general intelligence where machines can replicate or surpass human thinking).
Another thought: Won’t the criminals, who are as well funded (or better funded) than financial institutions also be developing AI to suit their needs? Perhaps their AI will be better than our AI.
Maybe we better hope John Connor really does take down Skynet.