Marinos Berger

In the run-up to the Leaders in Finance AML event on 1 October in Amsterdam Pakhuis de Zwijger, we spoke with Marinos Berger, Sr Solutions Lead Fraud & Compliance at SAS. We discussed how financial institutions can make AML more effective by leveraging data, AI and collaboration to reduce low-value work and focus more on actually preventing financial crime.

Could you briefly introduce yourself and your role at SAS?
I am a Senior Solutions Lead at SAS based out of Amsterdam working with Banks and Financial Institutions across the BeNeLux region on Fraud, AML and KYC. After more then 12 years in this field my focus is simple to describe and hard to do. Helping institutions move from proving they worked hard to proving it worked. I have seen many banks buy very similar programs under different names and I am curious about why the same challenges keep coming back.

Looking at the current AML and financial crime landscape, what are the most important developments you see for financial institutions today?
First, from July 2027 all EU countries will have to follow ONE rulebook. If effectiveness does not improve now, we lose our favourite excuse – namely the fragmentation. The real questions therefore is whether AMLA will measure activity or results.

One of the central themes of the Leaders in Finance AML Event is balancing effective AML compliance with cost efficiency.
Cost and effectiveness only compete when money goes into work that does not stop crime. Last year for example Dutch Institutions sent over 3 Million unusual transaction reports to the FIU and about 3 in a 100 were declared suspicious.

Many institutions are trying to move towards a more risk-based approach.
It is mostly a data problem I would say. You can only treat risk differently if you can see it and most financial institutions still lack one complete picture of their customers across their systems. So it requires resolving the “who-is-who” risk scores that move when behaviour moves and the confidence to apply lighter checks where the risk is ultimately low. 

AI and advanced analytics are becoming increasingly important in AML and financial crime prevention.
I think whether we should use AI or not is no longer the question. What matters is what we point it at. Doing the same work faster or finding what we miss today? I personally see the most potentials in three places:

1. In Networks – connecting customers, accounts and devices to reveal structures no single alert shows
2. In Triage – gathering more context and grouping alerts so that investigators spend more time on real cases
3. In Learning curves – how quickly a new criminal pattern becomes a working detection. If AI shortens that, it automatically makes you better and if not, it only makes you cheaper

At the same time, institutions need to remain cautious when using AI and more complex technology.
If an Investigator cannot explain the score they will most probably not act on it. Watching the feedback loop is clearly important. When AI decisions train the next model it learns its own blind spots. So keeping people reviewing what the model ignores is essential and judging new models by the same bar as the old rules they replace and not a higher one.

Data sharing and cooperation are often mentioned as essential in the fight against financial crime.
Fom next summer on Article 75 of the AMLR allows sharing on higher risk customers with safeguards. The law is becoming less of a blocker and what remains is trust and data.

What is one practical example of where data or technology can make AML efforts more effective without simply adding more complexity?One customer = One case. Today the same individual can trigger a monitoring alert, a screening hit and a fraud flag, each handled separately so nobody sees the full story. Linking all records beloning to the same persona and grouping them into one case is the way we should consider as a standard workflow. This way the investigator handles one case instead of five and sees the full picture. This would have quite an impact – less work, better decisions and no new system. Less noise and much more meaning.

Looking ahead to the AML Event on 1 October, what is one question or topic you hope will be discussed by banks, regulators, technology providers and other stakeholders?
What would we stop doing tomorrow if we were measured on what we prevented rather than what we documented? Everyone agrees the system must become more effective yet few of us say which parts of our own work could stop without anyone noticing. I would therefore love to leave the event with one concrete thing we collectively agree to stop doing ideally.