An Integrated Approach to Fraud

As criminals gain in sophistication, banks are at risk of complacency. A new approach employs a smarter combination of AI and allied technologies.

With digital transactions becoming the daily norm, so is financial crime. Criminals are becoming even savvier in how they target banks, payment providers. and financial institutions. To combat the evolving threats from financial crime, FRAML (FR for fraud combined with AML for Anti-Money Laundering) has emerged to align internal fraud detection and anti-money laundering compliance. But how prepared are we?

Criminals have been quite successful in their money-laundering schemes, costing in the neighborhood of 2% to 5% of global GDP. Between $800 billion and $2 trillion is laundered worldwide annually, of which $300 billion is laundered in the U.S. market alone.

This means that the United States is responsible for roughly 15% to 38% of all the money laundering happening annually. According to the Federal Trade Commission, bank fraud rose by 25% from 2021 to 2022, and it’s projected this growth will continue if the proper anti-fraud safeguards aren’t put in place.

FRAML is expected to deliver a synergized front against modern-day financial crimes, as it provides seamless integration between fraud detection and anti-money laundering mechanisms and overall case management. But is it all that it’s cracked up to be?

Fraud And Compliance

Historically, fraud detection and AML compliance operated separately from one another. Each had its own protocols, data systems, regulatory-compliance procedures and individual teams. There wasn’t a lot of communication between the teams and each operated in its own silo of automation and detection.

Criminals could benefit from this lack of communication because it could cause vulnerabilities. This lack of communication could also give way to various inefficiencies. Even though the goal of both of these departments was to combat financial crimes, no one ever sought to integrate operations and their findings because they were fighting crimes from different commercial behaviors and, often, even different central cultures.

For example, AML does not focus on benefiting an organization financially. And banks are too often at the mercy of fast-evolving regulations, which they are required to implement and report on. Implementing siloed approaches for both departments not only leads to vulnerabilities, such as overlooked threats, but often can also lead to redundant investigations and case management in completely different solutions.

For instance, one Asian bank cited success by analyzing cross-channel data, where it could ultimately detect organized-crime rings. It was able to identify the variety in the crime ring’s fraud attempts and structuring behaviors. In their own little silo of operations, a crime ring’s behavior would not necessarily have been detected.

It was only through a more holistic view of the crime ring—which included integrated fraud prevention and AML—that the bank could see its exposure in potential chargeoffs, which turned out to be a half-million-dollar risk. This discovery warranted the head of the bank’s group security to quickly deploy a new investigation model dedicated to organized crime rings, irrespective of fraud or AML bias.

Operating in silos has long been scrutinized. As more and more financial institutions are seeing strong results from use cases that prove a more integrated approach can greatly reduce the quantity of financial crimes, FRAML is gaining traction.

Now, regulatory bodies are starting to emphasize innovative technologies that strengthen the effectiveness of anti-fraud and financial-crime risk management. In the U.S., for instance, the Anti-Money Laundering Act of 2020 stresses the necessity for financial institutions to implement risk-based programs and solutions to help prevent money laundering and terrorist financing.

AI’s Role

Artificial intelligence within a FRAML approach aids in bringing about more accurate results. AI technology can help in more seamlessly automating data sharing and in providing more insights between various departments like AML compliance, fraud, and security.

The increasingly hyper-digitization of fraud over the last decade has been identified as a substantial prerequisite for a strong offense against money laundering. And it ultimately supported the development of the FRAML compliance framework. This framework further enhances real-time detection and analysis of criminal activities and brings predictive modeling to the forefront, thus significantly reducing false positives.

FRAML is well-suited to mitigate the risk of money-mule and social-engineering scenarios, and it adheres to Customer Due Diligence (CDD) requirements.

But there is one fraudulent risk that we have recognized that mostly impacts mobile-money applications and their telecommunication service providers.

In many scenarios, consumers will use their phones as a “bank account” for conducting financial operations, including registrations, logins, financial transactions, or loan requests. The risk of fraud with mobile devices can be mitigated with fraud protection that uses AI and supports FRAML by integrating fraud prevention, AML compliance, and credit management throughout the consumer’s journey.

For instance, when a new customer signs up to register for an account, AI technology would start performing customer segmentation and risk scoring based on a myriad of input data, while automatically scanning watchlists and sanction lists regarding current customers.

The AI technology continues working behind the scenes in making comparisons of fraud-related and compliance-related evaluations throughout the entire customer lifecycle. The processing occurs in real time and is dynamically triggered by customers whenever they begin a new action or operation.

A Potential Industry Standard

While FRAML offers lots of advantages for financial institutions and their customers, it does come with some regulatory hurdles. Indeed, it can be difficult to navigate the maze of compliance requirements, which at times can inadvertently impede an integrated approach. Hence, end-to-end FRAML processes have been long in the making and are only now being evaluated and implemented as a universal standard in banking.

So far, it seems there are various degrees of FRAML integration, ranging from semi-siloed operations to fully integrated applications. Bigger, more established financial institutions tend to have larger but fully separated teams. These institutions are typically slower to become more fully integrated. On the flip side, smaller or younger financial institutions tend to be more agile. They tend to see the value of integrating fraud and AML teams and operations. Overall, the degree of integration depends largely on a financial institution’s situation (resources, capabilities, requirements).

As with all system silos, the major challenge in achieving full integration usually boils down to data connectivity seen as the orchestration of data between a labyrinth of disparate systems. Legacy systems have been running for years, and it can be quite difficult to replace or reconfigure these outdated solutions to accommodate FRAML operations. However, as financial loss and risk continue to grow, larger institutions will integrate their operations more and more with each passing year.

Emerging Technology And Adoption

Even AI is evolving, as today there is hybrid AI. This combines the decision-making process of classical knowledge-driven AI techniques (for example, cognitive intelligence, or dynamic profiling) with the adaptive learning capabilities of data-driven techniques (for example, machine learning).

In the framework of FRAML, this translates to systems being not just rule-followers but also rule makers. Hybrid AI can analyze vast datasets, learn from them, and identify hidden patterns of fraudulent activity and money laundering that would escape human analysts or conventional intelligent rules-based systems.

In detecting complex financial crimes, hybrid AI systems can flag unusual transaction behaviors while considering contextual information to reduce false positives, without compromising detection rates. These systems adapt to emerging fraud tactics and the latest money-laundering schemes.

Financial institutions appreciate how hybrid AI systems foster adaptability and evolve with emerging fraud patterns, but keep humans in the loop. Financial institutions can further modify the dynamically adapting rules, simulate new rules and, above all, immediately take them into live operation, even if the system has not yet learned the rule in question on its own.

Classic machine learning is dependent on the maturity of the ML models they use. These must be intensively trained before they can be used for the first time with a reasonable detection rate. But the time required is not available to financial institutions in the age of instant payments. If a bank’s fraud-prevention team, for example, relied only on ML algorithms to learn fraud patterns, the criminals would have plenty of time to cause a lot of damage.

Current systems that maintain dynamic profiles for different entities (for example, a bank account) can detect “potentially fraudulent” transactions in real time through advanced rule sets. Through dynamic rule systems based on cognitive intelligence, situationally adaptive rules can also be created to help avoid the risk of a particular event. Thus, through the ability to handle independent rules for different data within the same transaction, it is then possible to create “fingerprints” for customers and identify unknown patterns.

In some isolated cases, these patterns could remain under the radar of a model based entirely on machine learning. As an example, imagine a transaction where the amount is 30% higher than the average amount drawn by that customer in recent months. There are also several new accounts for the customer, with amounts that are close to the maximum daily amount allowed. There are also unknown IP addresses, and the countries of the transaction are different from the customer’s.

Each of these transaction anomalies alone may be strange and suspicious, but not enough to raise the proper alarms. With the hybrid AI model, advanced techniques and machine learning work hand-in-hand, making software incorporated  with this technology a more effective alternative for anticipating these types of problems.

If anything seems fishy, you can simulate a rule and immediately test it on a live data set. If there is a problem, you will block it at once.

A critical factor here is the immediacy with which these models can react. Few tools have the ability to work in real time with in-memory capabilities to provide a “millisecond” response to identify and prevent fraud before it occurs. Models combining all available technologies and techniques have a substantial comparative advantage over models focusing on only a few technologies. This allows organizations to achieve significant savings and have a healthier and more satisfied and protected customer base.

The Future of FRAML

FRAML brings many benefits. But, most important, it supports the operational alignment of fraud management teams and anti-money  laundering teams, in addition to complying with data-sharing regulations.

When AI technology is added to financial fraud systems to support the FRAML framework—coupled with the fraud and financial-crime expertise humans can provide—fraud and AML teams alike can realize huge benefits.

They are effectively combining their human resources with efficient technology resources. No matter how you look at it, adopting a more integrated approach will not only combat sophisticated financial threats, but also deliver operational efficiencies.

—Justin Newell is chief executive of Inform North America.