At the beginning of the pandemic, everyone expected things to go back to normal after a two-week shutdown. Over a year later, nearly everything in our day-to-day operations has changed, and that includes how we interact with financial institutions (FIs).
Because COVID-19 made it difficult for consumers to venture out and run their usual errands, FIs needed to find other ways to provide their services. The only way for them to really keep up with the speedy digitization was through the implementation of AI systems.
To further discuss all things AI, PaymentsJournal sat down with Sudhir Jha, Mastercard SVP and head of Brighterion, and Tim Sloane, VP of Payments Innovation at Mercator Advisory Group.
AI-based banking tools
Jha believes that there were two fundamentally big changes that occurred in banking during the pandemic: the environment began constantly shifting, and person-to-person interactions were abruptly limited. “Every week, every month, there were different ways that we were trying to react to the pandemic,” explained Jha. This impacted virtually every aspect of FIs’ operations.
Companies were forced to take a more digital approach in a very short period of time. While this was something they were working toward pre-pandemic, the pandemic significantly increased the number of ways for people to connect remotely.
“When you are trying to provide the same kind of experience that you were able to do in a physical space—you’re trying to do that in digital space—you need AI to really capture the essence of the interaction and personalize that interaction for the customer that you’re interacting with,” added Jha. “An AI is able to do that. It’s able to sort of ingest all this data in real time and mimic how a human being is going to react to certain situations.”
AI can also adapt quickly to changing conditions, such as increased data being entered into the banking systems. It can then use its capabilities to predict behavior, not just for a particular customer, but for the entire ecosystem. Based on these predictions, AI can provide smarter and faster tools for FIs and their customers.
Sloane noted another change: an increase in coordinated criminal activity. “[Cyber criminals] made a business out of creating new attacks, and exploiting those attacks are effective at scale. They hire gig workers to help execute [these] attacks,” explained Sloane. “They’re really going at this in a big way.” Because of the increase in data, they have more personal information on consumers than ever before, and AI is a critical component in getting and keeping cyber-attacks under control.
AI can better detect credit risk
Pre-pandemic, FIs already had plenty of information on their customers. With this information, banks were able to put in place some standard rules for screening them. Since then however, the amount of data available has grown exponentially.
So what are banks to do with all this data? If FIs want to compete in this newly evolved environment, the answer is AI.
“With AI, [FIs] can, in many cases, create features by combining data in very interesting ways, and [there are] exponential ways that [they] can do that,” said Jha. One of these new features is an updated and more intricate credit risk model. Using AI, banks are able to optimize a number of different outcomes while considering factors that may have been overlooked in the previous model, such as default rate, profitability per customer, and increased credit limits for certain individuals.
“What AI also does much better than the traditional models or rules based systems is the ability to learn from other people’s data, even competitors’ data, without transferring the data itself,” continued Jha. FIs can now transfer the learning from different data sets, making it unnecessary to actually share the data. Additionally, AI can better prevent fraud than the previous methods used.
How to use AI to minimize late payments
A major focus of Brighterion’s solution is the uniqueness of each individual customer’s experience. This includes excellent risk management, from delinquency to collection, which spans across the customer’s lifecycle. “It’s not just [about] identifying [a] bad consumer, the consumer that actually is going to default, but really understanding how we can enhance the customer experience and the entire journey,” said Jha.
Mastercard considers these three questions when finding solutions for minimizing late payments:
- How do we make sure the bank knows how much credit to give a customer?
- When is it justifiable to give a customer more or less credit?
- How do we predict delinquency early?
The focus is not on predicting delinquency right before it’s about to happen, but rather, predicting the majority of delinquency 70 days in advance. This gives the FI an ample amount of time to work with the customer and perhaps avoid delinquency altogether. A plan of action can then be put in place, allowing the customer to set up an installment plan, increasing the odds that they will never reach the point of late payment.
“AI itself may not be able to eliminate default or eliminate late payments, but it can actually provide the tools, to both the consumer and to the banks, to be able to come to a situation where [they] can be much more proactive about these things, and therefore, work out a situation that allows the customer to be happy. And the banks will be happy because they minimize the losses from these situation,” concluded Jha.
Fact vs. Fiction: Myths surrounding AI adoption
The power of AI seems magical, so it’s no wonder some people have trouble trusting it. But at Mastercard, and particularly Brighterion, “explainability” is the goal. This means that “every outcome, every signal that we produce, every recommendation that we give from the model, we want to make sure that we can provide a region code for it,” expanded Jha.
For example, Mastercard does not just tell a customer that a particular score they looked at for a credit decision was high or low; they will provide a variety of reasons for what led to that score. With more research happening all the time, these AI algorithms become increasingly explainable, something that is a critical asset for adoption.
A popular myth that the industry has seen and mostly debunked is that AI systems are too expensive and a number of highly qualified data scientists are required to incorporate AI solutions. Perhaps this myth used to hold some validity, “but we have overcome that,” assured Jha. “With many different platforms that are available today, solutions are almost ready to be implemented with very small changes.”
To put customers at ease, Mastercard can build custom models for them in a short period of time, for example, its 8-12 week program that provides a clear picture of the Return on Investment (ROI) before implementing these solutions. This lets the customer know exactly what they are getting into.
Lastly, some FIs still believe that it takes a long time for AI models to change. “During the pandemic days, we would get asked this all the time, how quickly [Mastercard] can adapt, because things are changing,” remembered Jha. “And all the data elements that were from before, for example, when most institutions gave three to six months of offset of no payment necessary. All the payment history that could be used for character prediction couldn’t be used [anymore].” The models had to react to this situation, and they had to do it quickly.
To combat this, Mastercard used a mixture of techniques, combining many different models to create results. It also used a variety of data sources and velocity signals, most of which are able to adapt in an efficient manner. So while there is a bit of truth to some of these myths, there is always a solution in place to combat and ultimately debunk them.