Evaluating AI adoption in AML programs

Financial services organizations can use AI technology in their AML programs to align with other AI-based enterprise initiatives and create efficiencies in their processes. Additionally, organizations can use AI to combat fraud and reduce false positive rates, streamline know your customer (KYC) and AML frameworks, and enhance transaction monitoring systems.

Combating fraud and reducing false positive rates. Financial services are implementing AI technology, but bad actors also employ AI toward their own ends. In fact, fraud rates have increased significantly since the emergence of AI technology, and since the COVID-19 pandemic began in 2020, check fraud has increased by 385%. However, organizations that use AI in fraud prevention, specifically in alert or event evaluation, have seen 70% fewer false positives, based on a study by the technology and information services firm CSI.

Streamlining KYC and AML processes. Integrating AI into an organization’s AML program can improve many aspects of business processes, including increasing efficiency and accuracy and assisting in real-time detection. Additionally, AI-based applications that use deep learning and large language models (LLMs) can enhance KYC and AML programs by improving identity verification, as AI technologies can help confirm or reject a new customer’s identity in real time. Strengthening KYC and AML programs is critical in protecting customers and avoiding potential fines. In 2022, $5 billion in fines were levied against financial services organizations for various AML, sanctions, and KYC system failures.

Improving transaction monitoring. AI technology can enhance transaction monitoring for AML compliance. By using advanced analytics and machine learning algorithms, AI-powered systems can identify suspicious patterns and anomalies in real time or near real time. These AI-based transaction monitoring solutions can analyze vast amounts of data from multiple sources, including customer profiles, historical transactions, and other relevant information. The algorithms can detect behaviors that deviate from a user’s past activity or exhibit characteristics indicative of money laundering or terrorist financing. The ability to generate timely alerts for these suspicious activities allows compliance teams to focus their efforts on the most high-risk cases rather than on false positives.

Some potential ways AI technology can be deployed in a financial crime program include:

KYC onboarding AI models. An AI-based model can automate identity verification for new customers by analyzing documents, historical data, and other metrics. The model can quickly analyze the authenticity of documents provided by the potential customer and identify any discrepancies. This automation can streamline the onboarding process, enhance the customer experience, and reduce the risk of identity fraud while complying with regulatory requirements.

Predictive analysis for risk assessment. When using AI, predictive analysis can assess the risk profiles of customers and transactions based on behavioral patterns, historical data, and counterparties. An AI model can identify patterns and anomalies in transactional data, enabling organizations to identify potential risks associated with the customer. Financial services organizations can more effectively mitigate risks and prioritize resources by predicting the likelihood of suspicious activities.

Suspicious activity report (SAR) narrative generation. An AI-powered process can automate the generation of detailed narratives for suspicious activities identified by the organization’s AML system, improving efficiency in compliance reporting. Using natural language processing algorithms, an AI model can swiftly analyze data including transactional, customer, historical, and previous SARs to generate initial drafts of narratives that provide insights into potentially suspicious activities. Organizations can significantly reduce the time and resources required for compliance reporting by implementing an automated process.

AI graph analytics for network analysis. AI-generated graph analytics can review networks of financial transactions and relationships, uncovering hidden patterns indicative of money laundering and supporting proactive risk management. Using machine learning algorithms to assess transactional data, such models can identify unique clusters of related entities and transactions and help organizations flag potential instances of suspicious activity for further investigation.

Behavioral biometrics. Using AI, behavioral biometrics can authenticate users and detect suspicious activity in real time, enhancing security measures and fraud detection capabilities. AI-trained behavioral biometrics can help financial services organizations reduce the risk of unauthorized access or fraudulent transactions, improve customer experience, and provide insights into customer behavior.

Alert models. AML compliance teams rely heavily on rule-based transaction monitoring systems, which often yield a high rate of false positives that strain resources and hinder investigators’ efficiency. To mitigate this issue, organizations can develop AI models that automatically learn complex patterns from customer, transactional, and historical data. These models focus on the individual alert level, assigning each alert a score ranging from 0 to 1 based on its likelihood of resulting in a SAR. By integrating these predicted probabilities into organizations’ AML systems, organizations can better allocate resources by prioritizing high-probability alerts, reduce false positives, and improve efficiency.