Research shows financial institutions want proactive fraud prevention

64% of large banks want supervised and unsupervised AI for fraud prevention

As artificial intelligence (AI) and machine learning (ML) become more commonplace in the business world, financial institutions are realizing the role these technologies play in their sector. Use cases include preventing fraud, insider trading, and money laundering while identifying opportunities for better customer service and new markets.

As you may have read previously on this blog, Brighterion collaborated with PYMNTS to analyze how AI and ML are being used by financial institutions (FIs) across the U.S. We surveyed 200 financial executives from banks and credit unions with assets ranging from $1 billion to over $100 billion, resulting in 12,000 collected data points for analysis.

We were surprised by some of the things they told us, while others confirmed what we suspected. For instance, in our joint report AI Innovation Playbook: how FIs are using AI and ML, we learned that larger institutions were further down the path of adoption than smaller ones. That made sense; they have more assets to track, are often more spread out geographically, and have bigger budgets to dedicate to IT and administration.

What surprised us, however, is the large number of FIs that have learning systems but don’t leverage the full potential. There is a clear opportunity to use technology more effectively.

The most used ML technology is data mining, with a 70.5 percent adoption rate. The problem is that data mining is not necessarily the most effective tool for the job. Data mining supports payment services well, but it’s only used for this purpose by 45 percent of FIs.

Large banks also reported they use fuzzy logic, business rules management systems (BRMS) and neural networks for data insight, while smaller banks tended to use one or two tools. As we discussed last time, BRMS is a very common form of machine learning, with 59.5 percent of banks reporting they used it for fraud prevention. Unfortunately, BRMS uses a predetermined set of rules, whereas fraudsters continually change their methods to stay ahead of their targets’ knowledge.

Few financial institutions (5.5%) leverage AI that uses both supervised and unsupervised learning. In essence, the platform analyzes data and builds and evolves profiles as more information is received, preventing fraud in real time as it easily identifies anomalies as they occur.

Smart Agents, developed by Brighterion’s data scientists, enable 10 or more machine learning technologies to interact and develop a true picture of each individual profile, using both supervised and unsupervised learning.

Smart Agents can understand the full behavior of any entity or individual. Once the data is received, Smart Agents enrich the data, saving our customers hours of tedious work and allowing them to focus on their businesses. They pull data from multiple sources, regardless of format, type, complexity or volume, providing unlimited scalability (see Aite Group’s report which names Brighterion #1 for scalability) and no disruption—no single point of failure.

Learning and making real-time observations from interactions with human users, Smart Agents apply this knowledge to create virtual representations of every entity with which they interact, building a digital profile that optimizes customer-facing payments and banking services. If there are 200 million cards in an ecosystem, there will be 200 million Smart Agents analyzing and personalizing their services to a degree that other ML systems cannot accomplish.

Yet very few FIs use Smart Agents, substituting a combination of ML tools for true AI systems. Decision makers need to understand and accept that ML offerings are not viable substitutes for true AI. ML simply cannot carry out certain functions without considerable human intervention and using it in these areas produces little benefit compared to what can be achieved with Smart Agents.

Low usage does not mean FIs — especially larger ones — are not interested in Smart Agents. Approximately 64 percent of those holding more than $100 billion in assets reported being “very” interested, with 9 percent “extremely” interested.

Some smaller banks also expressed interest in Smart Agents, with 13 percent saying they would consider adopting the technology.

Compliance, money laundering, credit delinquency and other forms of fraud happen in real time and are not predictable. Get the full report to learn more about the state of AI in banking.