Catching fraud off guard: How artificial intelligence will power next-generation fraud mitigation
Controlling fraud and money laundering amounts to a perpetual game that requires a dynamic and evolving process as fraudsters continuously adjust and adapt their techniques. But relying on business rules or models that use historical data with technologies such as neural networks, deep learning and data mining have only short-term value. This is because they lack the capacity to do the following:
- Adapt. While fraudsters adapt and evolve their techniques, the legacy technologies cannot respond to behaviors that change continuously.
- Personalize. To successfully protect and serve customers, employees and audiences, we must know them by their unique and individual behavior over time—and not by static, generic categorizations.
- Self-learn. An intelligent system should learn from every activity associated to each specific entity.
A next- generation solution is needed that does not rely on pre-programmed rules and/or models based on historical data. The goal: to detect new fraud or money laundering schemes as they arise. Here is how that solution works.
From real time to one-on-one: What effective risk monitoring looks like
Effective real-time fraud prevention and anti-money laundering (AML) monitoring requires a system characterized by several key features. As implied above, it must adapt: that is, continuously update individual profiles to learn behavior over time. Also, it must self-learn to reflect and adjust its parameters to thrive in new environments. In this case, imagine a plumbing system that autonomously notifies the plumber when it finds water dripping out of a pipe and detects incipient leaks.
As for other characteristics, it should:
- Be data agnostic. The proliferation of payment types and methods requires flexible technologies that can manage data in any format.
- Work in real-time. Solutions without real real-time capabilities (before authorization) lack efficiency: They do not proactively prevent fraud losses.
- Profile behavior on a one-to-one basis. It must learn from each individual activity and adapt to the specific situation/behavior of every entity of interest over time. This continuous one-to-one behavioral analysis provides real-time actionable insights and effectively reduces fraud while it increases profit.
- Offer multiple layers of protection. To efficiently prevent fraud, a solution must provide protection across all layers.
Artificial intelligence, naturally: Smart agents technology
“Smart agents” technology is the only solution that overcomes the limits of the legacy machine learning technologies to allow personalization, adaptability and self-learning. Simply put, a smart agent can hold entire conversations using natural language technology that understands the intent and meaning of customer questions. But there is more: It creates a virtual representation of every entity that learns and builds a profile from the entity’s actions and activities.
For payments, a smart agent is associated with each individual cardholder, merchant or terminal. The smart agents associated to an entity (such as a card or a merchant) learn in real time from every transaction and collect specific and unique behaviors over time. There are as many smart agents as active entities in the system. Decision making becomes specific to each entity and no longer relies on universally applied logic, regardless of their individual characteristics. And because smart agents self-learn and adapt, they continuously update individual profiles from each action performed.
What’s more, they do not rely on pre-programmed rules nor try to anticipate every possible scenario. Instead, the profiles specific to each entity—cardholder, merchant, device, etc.—grow with various activities, including card present, card not present, ACH or wire. This one-to-one behavioral profiling provides unprecedented visibility across all channels into the behavior of cards and merchants.
Since smart agents focus on updating the profiles based on the actions and activities of the entity, they store only the relevant information and intelligence rather than incoming raw data, which achieves enormous compression in storage.
By contrast, legacy technologies in machine learning generally rely on databases. A database uses tables to store structured data—but tables cannot store knowledge or behaviors, which artificial intelligence and machine learning systems require. Enter the smart agents, which bring a powerful, distributed file system specifically designed to store this information.
This distributed architecture also allows lightning-speed response times (less than 1 millisecond) on entry level servers, as well as end-to-end encryption and traceability. This allows for unlimited scalability and resilience to disruption: It has no single point of failure.
A comprehensive risk monitoring solution must combine the benefits of existing artificial intelligence and machine learning techniques with the unique capabilities of smart-agents technology. The resulting comprehensive solution is intelligent, self-learning and adaptive. And through the application and implementation of smart agents, banking leaders will display intelligence of another kind.