Banking on GenAI: 3 techniques for fighting fraud and financial crime

Fraud, risk, and compliance departments worldwide face a rapidly evolving and expanding workload, fueled by increasingly complex modern-day fraud and financial crimes. According to data from the FBI and Deloitte, generative AI (GenAI) is expected to rapidly increase fraud losses over the next few years. But even as AI technology galvanizes bad actors’ tried-and-true deceptions – deepfakes, phishing schemes, payments fraud, and scams of all kinds – it can also fortify financial firms’ defenses.

Already, investigators commonly rely on machine learning (ML) models to uncover activities that fit known fraud schemes as well as identify new and emerging trends in fraud. In addition to detecting established fraud modalities, ML algorithms can readily flag suspicious transactions as fraudsters alter their methods. Investigators also use natural language processing (NLP) and text analytics to extract data such as transaction value, where the transaction took place, IP addresses, and other documents.

More recently, anti-fraud professionals have begun exploring the fraud fighting potential of generative AI. A recent fraud technology survey by the Association of Certified Fraud Examiners (ACFE) and SAS found that 8 in 10 (83%) of expect to add GenAI to their anti-fraud tool kits by 2025. 

How might it help them make better, faster decisions to curtail fraud and financial crimes? Below are a few ways GenAI can augment more traditional AI and ML techniques to help investigators out-maneuver their cunning adversaries.

GenAI as an investigator’s digital assistant

Fraud and compliance investigations typically involve a massive amount of data to review, including myriad financial records, untold volumes of financial transactions, and droves of records of outside companies. This type of information is not only extremely time-consuming to examine, but the process of extracting pertinent evidence (e.g., key people, addresses, phone numbers, relationships, etc.) hidden in the data deluge is typically challenging and cumbersome. New information learned from one facet of the investigation often demands scouring previously read reports, making established processes repetitive and laborious.

Enter large language models (LLMs), a form of GenAI that can help investigative teams find relevant data points and connect the dots between them faster and more easily. Given how LLMs work, it isn’t hard to imagine how an LLM-powered “digital assistant” could deliver great value to investigators, quickly cataloguing and interpreting data to answer questions and extracting the most relevant information. Digital assistants of this type can generate summary narratives, highlight key details, identify potential gaps and conflicts within the investigative process, and even suggest follow-up tasks.

What sets this approach apart is GenAI’s adaptive nature. Like other forms of AI technology, it learns and evolves with user feedback, constantly refining its models and providing deeper contextual understanding within the investigative domain. Done right, this dynamic interaction can help ensure accuracy, explainability, and transparency at all points in the process.

Additionally, the collaboration between LLMs and traditional AI further enhances many other aspects of fraud and compliance investigations.

GenAI for conversation analysis

Conversation analysis has redefined how investigations approach digital exchanges. This capability can help revolutionize risk assessment in investigations by ingesting and organizing transcripts from digital exchanges on mobile and other devices. The feature presents the data in an easy-to-use viewer which shows the exchange between two or more people.

The tool’s key benefit lies in its ability to navigate and select highlighted key terms, which accelerates the process of identifying opportunities in massive conversation logs. This is useful in fraud schemes such as account takeover and accessing new credit lines via mobile banking or telebanking. Applied to a conversation between a fraudster and an online agent, for example, GenAI-driven conversation analysis can quickly reveal “red flag” behavior such as multiple requests for non-monetary changes to an account – e.g., requesting a new card, adding an authorized user, modifying personally identifiable information (PII)manipulation, or changing of an email address.

The injection of GenAI capabilities into financial firms’ conversational analytics tools will allow them to reach deeper into large chat logs to identify behaviors of potential concern and assess ongoing fraud and financial crimes risks.

GenAI for testing and optimizing fraud and risk systems

Data is the lifeblood of any AI algorithm – but what if the bank’s data is sensitive or lacking sufficient volume? Synthetic data is algorithmically generated data that mimics real-world data. Organizations use synthetic data generated by AI when real data is unavailable, inadequate or inappropriate due to:

• Sensitive or private information.
• Prohibitive cost.
• Hand-labeling inefficiency.
• Bias or imbalance.
• Rare-scenario data shortages.

According to Gartner, by 2026, 75% of businesses will use GenAI to create synthetic customer data – up from less than 5% in 2023. Many banks are already exploring use cases.

Another recent Gartner study highlights how synthetic data will promote financial inclusion by enriching credit risk decisions and help banks enhance fraud and financial crime prevention. For example, financial firms can use synthetic data to:

• Train their machine learning models to detect fraud or recognize illicit payment patterns indicative of potential money laundering.
• Conduct penetration testing and reduce false positives by simulating novel fraud attacks.
• Safely share data for software development and testing without compromising data privacy or security.

Looking ahead

The use of AI technologies marks a significant milestone in advancing banks, insurers and other financial services organizations’ investigative capabilities for fraud and financial crimes. Notably, the technological synergy of the aforementioned GenAI techniques presents a transformative opportunity for investigators to revolutionize their risk, fraud and compliance operations, leading to more effective detection and prevention strategies.