ACTICO Anti-Money Laundering

Business rules in an AML system are pre-configured criteria used to monitor potential money laundering. Examples include limits on financial transfers, unusual money movements, and suspicious customer behaviour compared to a client’s normal pattern or that of their peer group.

ACTICO Anti-Money Laundering comes with a standard rule set to check transactions and personal data. When an anomaly occurs, the system notifies the money laundering officers and automatically initiates a workflow for further processing of these ‘hits’.

Machine Learning (ML) is a powerful method within artificial intelligence, particularly well-suited to reducing false positives in AML systems. ML models are trained using a combination of compliance officers’ expertise (expert knowledge) and historical money laundering cases (data knowledge). This complements the system’s rule-based verification algorithms, leading to improved detection performance by prioritising true positives over managing false alerts.

An ACTICO project with a retail bank demonstrated the significant potential of machine learning in reducing false positives in the fight against money laundering. A validation dataset showed that that bank eliminated approximately 40% of false positive alerts.

Machine learning supports banks and insurers in customer segmentation. This is advantageous because data is automatically analysed to group customers into clusters based on behaviour, risk profile, and other key factors. This type of analysis helps identify distinct customer segments and uncover meaningful patterns in the data.

With generative AI (GenAI), compliance officers can automate time-consuming research and complex regulatory reporting tasks. This is especially valuable when creating reports—such as suspicious activity reports for Financial Intelligence Units (FIUs) or MROS in Switzerland. Compliance teams are required to provide and annotate legally relevant material.

GenAI supports this by extracting relevant information from complex datasets, evaluating the findings, and generating structured draft texts—reducing manual review effort and accelerating report creation. Laborious document reviews are also eliminated: AML experts in banks and insurance companies can simply use prompts to have the AI search available data sources and deliver digital results, producing a first draft of the regulatory report.

When calculating AML risks, various factors are taken into account—including customer risk (e.g. PEP status), product risk, transaction risk, country risk, and the risk associated with distribution channels.

Even broader sets of information can further enhance risk assessments. These may include partner details, business relationship structures, geographical data (such as countries of residence or nationalities), and external sources like customer screening results.

At ACTICO, business relationships are assigned a risk level based on these factors. While the risk level is generated automatically, it can be adjusted manually where needed. Key elements such as source of wealth, expected transaction volume, deposit levels, and the origin of assets can also be included. Users can easily define additional custom risk factors to tailor the model precisely to their needs. In exceptional cases, analysts can manually assign a specific risk classification.