Beyond KYC: How AI Risk Detection Software Helps Banks Combat Modern Fraud
As fraud activities grow in sophistication, so should the technologies used to combat them. Here’s how AI risk detection software can provide further protection in addition to the traditional KYC process.
Despite the ongoing development of sophisticated tools and technologies to detect and combat fraud, synthetic identity theft and money laundering continue to plague organizations in a myriad of industries. It’s a seemingly endless game of tug-of-war, where technology companies must continually up the ante to fight back against fraud as criminals continue to find ways over, under, or around newly developed fraud detection methods.
Vice versa, as fraudsters become more skilled in committing illicit acts, so too must technology and software companies in detecting fraud and stopping criminals in their tracks.
One way that organizations tackle fraud issues is with the use of KYC (Know Your Customer) processes that attempt to verify the identities of their customers so they can recognize patterns in behaviors. So far, AI risk detection software is poised to play an important role in allowing KYC methods to thrive.
What does KYC mean?
KYC refers to the process of verifying the identities of their customers to determine the potential risk of fraud. The idea is that knowing your customers ― including personal details, spending habits, financial histories, and unique risk factors ― can reduce the chance of money laundering, terrorism financing, and other types of illicit financial activities from occurring.
KYC fraud detection consists of three core steps:
- Customer identification procedures
- Transaction monitoring
- Risk management
When all three parts are addressed, financial institutions should be better positioned to differentiate between authorized and unauthorized or illegal activity.
The Challenges of KYC in Digital Channels
KYC isn’t just good business sense: it’s also part of maintaining compliance. Still, many organizations have unsuccessfully implemented adequate KYC detection methods in their organizations. They’ve racked up an estimated $26 billion in fines related to non-compliance with AML, KYC, and sanctions over the last 10 years, not to mention the immeasurable damage to their reputation for not being in compliance.
Several challenges impede the efforts of traditional fraud modeling, particularly as digital channels continue to blur the lines in the banking industry. To date, KYC has largely relied on document-based verification, including IDs, utility bills, pay stubs, and face verification. But now, many interactions are no longer face to face: customers can open accounts via chatbots, use mobile apps to complete transactions, and place orders using voice assistance, leading to increased usage in unstructured data.
DataVisor VP, Priya Rajan, notes the difference between structured and unstructured data:
“Unstructured data isn’t in the same neat and tidy package that structured data comes in because it’s being collected from a variety of channels. The emergence of structured and unstructured data has complicated KYC processes to the extent that traditional technologies cannot keep up. Reference databases like sanctions lists, while effective during an era when data was centralized, is no longer sufficient. Companies who wish to remain competitive must be able to leverage their unstructured data with confidence.”
With data breaches, valuable information about the customer gets exposed to fraudsters who make use of these supposedly private details to create synthetic IDs that can easily pass the KYC process. On the operational side, increasing the volume of online applications burdens an already costly and inefficient process.
Banks are becoming more competitive with the emergence of challenger banks and are in turn offering their own digital versions of traditional banking services. With that, the need to seamlessly validate customers without increased friction and time lag is becoming more important than ever.
Combating Fraud with Machine How AI Risk Detection Software Can Enhance KYC
Financial institutions are considering new technology tools to address challenges to meet the heightened regulatory scrutiny and the increasing cost pressures that are affecting their KYC processes. Ideally, these tools can help with automating many of the manual processes associated with KYC.
When implemented correctly, the results are twofold:
- Higher accuracy and coverage; and
- Greater operational efficiency
Clustering and graph algorithms look at all user behavior rather than a single user. This allows companies to deliver high detection accuracy with fewer false positives and provide a better customer experience.
In addition, companies need the ability to identify groups of suspicious accounts based on correlated patterns at scale, but doing so can alert teams of potential fraudulent actions. This enables them to confidently make bulk decisions, but also to work through them with efficiency.
While business process automation techniques can help alleviate operational challenges associated with this increasing workload, they cannot deliver the intelligence based on big data at the real-time scale that is needed to proactively address emerging threats from crime rings and coordinated attacks.
Many existing solutions look at events in isolation or heavily rely on historical patterns to predict future risk. Not only are these solutions reactive, but in many cases, they also miss the critical signals that indicate fraud because they don’t know what to look for. This is evident in the case of bot-initiated attacks where many events look normal in isolation, but when looked at holistically are indicative of broader, organized crime activities.
To address these previously unidentified and emerging threats and acting on them before they result in fraud requires transformational thinking. The best approach should not just look at incremental solutions, but rather at solutions that can radically redefine and reshape how fraud is discovered.
In response, AI risk detection software is helping to solve KYC challenges by leveraging big data in real-time at scale. AI’s ability to monitor larger sets of data at a time can better identify larger crime rings by reviewing data holistically.
DataVisor’s unsupervised machine learning (UML) does not require historic loss labels or intense data training to detect anomalies. This allows financial institutions to proactively address potential instances of fraud as they occur rather than reactively review transactional history over time.
See DataVisor’s UML in action by requesting a free demo.
