AML/CFT Artificial Intelligence

Artificial Intelligence Scenarios Strengthen Your BSA Program

April 29, 2020

Read Time: 0 min

Key Takeaways

Abrigo introduced five new machine learning scenarios

Abrigo recently introduced five new scenarios to BAM+, their industry-leading AML software, that utilize artificial intelligence (AI) and machine learning (ML) to help BSA departments catch more suspicious activity. The five scenarios focus on:

All credit transactions – This scans against every credit transaction, regardless of the type of transaction.
Other credit transactions – This monitors credit transactions that are not currently mapped to a specific transaction type (like cash)
Other debit transactions - This monitors cash transactions that are not currently mapped to a specific transaction type (like cash)
Remote deposit capture
Daily ACH Debit Amount Anomaly Detected

Machine learning scenarios learn from normal transactions

These scenarios are based on machine learning, meaning they will get “smarter” the longer they work in your system. As transaction anomaly detection scenarios, they detect any abnormalities in an account’s actual behavior against its stated behavior, similar to how spike scenarios currently work on a rules- or behavior-based system. They will also expand your current monitoring capabilities because they monitor for transaction types that were not previously looked at by other scenarios. This will also cause a normal uptick in alerts.

Being able to explain the AI/ML in your system is essential

AI/ML scenarios can cause a headache for BSA professionals with the mere mention of them. Most BSA officers feel that they cannot even explain how AI works so therefore they cannot use it in their systems. AI shouldn’t be secret or proprietary information, especially with AML software. In a joint statement issued in December 2018, FinCEN highlighted the need for financial institutions to use AI to strengthen their BSA programs, but they cautioned that institutions still must be able to explain the systems to regulators.

While these five artificial intelligence scenarios might be more technical than our current rules and behavior-based scenarios, you shouldn’t be afraid to use them. Our team designed these scenarios in-house using the Microsoft ml.net framework, based on the anomaly detection algorithm. The models are powered by transparent machine learning models that easily allow the end user to explain how they work to regulators, a key component to any AML technology.

Use transparent machine learning scenarios to strengthen your BSA program.

learn more

What is ml.net?

It is an industry standard machine learning framework that uses proven AI methods that are explainable. This means BAM+’s AI/ML scenarios are included in the only transparent AI system on the market. The code is open source and standard, so there’s no mystery or “proprietary information” behind how it works which should put BSA professionals’ minds at ease come exam time. Our product team created our AI/ML scenarios with an ml.net anomaly detection algorithm that uses the Gaussian Kernels of adaptive bandwidth, which is the industry-standard mathematical method to detect anomalies in patterns of behavior. We will ensure you know exactly how and why these scenarios work the way they do. This is the best way to detect anomalies according to industry experts and others even outside Abrigo.

AI will enhance your BSA program and not replace the human element

Similar to all of our current scenarios, each AI-powered scenario is completely customizable to fit your institution’s unique risk profile. You can set an institution-specific amount threshold with each scenario and apply them to user-defined groups. This allows the system to learn from each transaction and more precisely detect potentially suspicious behavior. It doesn’t utilize industry-standard data to assume that each individual account (convenience store, retail outlet, etc.) has a similar transaction history. It takes into account each individual account’s transaction patterns and detects anomalies from that. The strength of machine learning lies in its precision and limited false positives.

Other BSA software may claim to have AI built into their programs, but can you actually explain to your regulator how they work come exam time? With our AI scenarios, we will provide you the documentation so you can prove to your regulator that you fully know and understand how your system works.

By developing and adding these five AI-powered scenarios, we are enabling BAM+ users to spend more time on things at which humans are more effective. Computers are better at finding patterns so we are automating some of the more tedious parts of your job. Because you are better at looking at an alert and knowing if it should be escalated to a case because your gut tells you something is off, that is where you should be spending your valuable and limited time.

Artificial intelligence should be used to strengthen your BSA program and help you better fight financial crime. While AI relies on human knowledge, it will never replace your human instincts or the need for a fully-staffed BSA department.