This article is authored by THE DATA INITIATIVE AB Member David Caruso.
AML news has many believing we’re on the cusp of dramatic change that will wipe out thousands of compliance jobs.
We hear Artificial Intelligence is poised to transform the few remaining AML workers into Arnold-like Model T-101 Terminators annihilating alerts, cases, and SARs with emotionless robot-like precision.
The reality is a bit different.
The promise of AI in AML is overstated.
“Artificial Intelligence” is an imprecise concept.
“Better Software” is a better term then “AI.”
Better Software will create more AML work, not less.
Better Software Can Improve AML Today
Three or so years ago, when the concept of AI first emerged in AML, there was a hazy notion new technology will somehow fix the overwhelming inefficiency plaguing transaction monitoring and due diligence.
Billions of dollars are wasted as AML teams remain buried under false positives and hunt down information from clunky old databases. This is inefficient and costly. But a hoped-for quick fix from AI was never going to happen.
Have you noticed at recent AML conferences and in marketing content how the term “Artificial Intelligence” is morphing into different words like “Machine Learning” and “Robotic Process Automation?” AI is unclear. Machine Learning and Robotic Process Automation are however real things that users can see. Or, to say it another way, “software you can purchase and use now.”
Machine Learning is software capable of being “taught,” through specific examples and instruction. In AML we see Machine Learning improving Adverse Media monitoring.
Experienced analysts know what negative news results are useful to investigations and due diligence. A skilled developer uses a set of these good negative news articles to teach a computer how to better identify content that actually helps in due diligence or a SAR case. This is considered Machine Learning (although it may be more accurate to say “Machine Teaching”).
Robotic Process Automation or RPA automates many of the repetitive manual tasks that burden analysts and investigators. Think about a SAR investigator having to gather check images, monthly statements, KYC files, and negative news. RPA can automate this information gathering, sparing the investigator monotony and saving time.
Better Software Means More AML Work
Both Machine Learning and RPA enable humans to spend less time on tasks better completed by machines. However, it is wrong to then conclude this means less AML work and fewer AML workers.
Picture Machine Learning and RPA clearing out the clutter. As tough a problem as this is for AML, we are now beginning to solve it. This will launch a new period of creative development because (1) progress creates momentum and (2) data scientists and software programmers will have gained a better understanding of AML and be ready to take-on the decade-old transaction monitoring applications on which most AML work depends.
With smart engineers and AML experts turning their attention to this bigger challenge, does anyone think new advanced technology will discover less money laundering and financial crime? Of course not. We detect only a tiny fraction of money laundering today, so would using better systems mean we detect even less? That doesn’t make sense.
We Fail to Detect Most Money Laundering
The U.N. estimates (on the low side) that more than $2 trillion a year is laundered.
The total money seized by law enforcement agencies each year and the total amount of suspicious funds reported on SAR/STRs adds up too much less than the actual amount of money laundered worldwide.
Now, imagine better software, through massive data analysis connecting an innocent-looking wire transfer to the little-known associate of a human trafficker. Imagine detection systems adapting in near real time to new schemes uncovered and shared by law enforcement.
Instead of shrinking AML departments, financial institutions will be hiring more investigators to deal with the rise of new, heretofore undetected suspicious activity. And won’t these investigators need better training and more experience to properly investigate complex financial crime schemes?
And just as we in AML will have better technology and better ways to detect suspicious activity, criminals will have access to the same (or better) technology. Criminals are as well funded as financial institutions and have greater incentives for success. The fight between those misusing our financial systems and those responsible for defending is going to get much tougher.
Change is indeed coming to AML. It is just different change than many people realize.