Page 35 - GS230702

Education

and current transaction are reasonable. This data must be In the event-driven world, a bank just has to make sure a
fed into the model and assigned a score. payments channel sends the right event to communicate
with the fraud detection or AML system, and receives the
The score also depends on authentication requests. same events to get the “yes or no” back.
Typically, if you can identify a user together with their
mobile phone, banks pass the transaction because they The alternative is not an option
are comfortable they know who the user is. But if a similar It's a much easier integration than attempting this via
scenario occurs where the user has reached the same score, standard REST APIs, which becomes far more challenging
but there is no biometric data or mobile authentication, and will need to be built differently for every channel a
this would likely trigger a blocking or flagging of the bank has now, plus any new channels. This means banks
transaction for escalation. may have to change models based on changes in user
Now add AI and ML behavior, as well as changes driven by new products
and services, or to counter new types of fraud or money
When a bank has built a database of models, and new laundering.
transactions can be checked against the models and given
an accumulated score, AI and ML step up to the plate. With standard REST APIs, every time a bank adds a new
Aided by EDA, they can make rapid decisions and enable channel, it has to change the way AML and fraud systems
companies to flag abnormal transactions in real-time work, because they have to know about this other channel.
across all channels. In the event-driven world they don't know and don’t need
to know.
Layering these data models with AI/ML enables banks
to gain ground on fraudsters and money launderers. Mc With EDA, banks can accurately support a high volume
Kinsey researchers (https://tinyurl.com/4a73xtmt) wrote, of transactions in the quickest response time, balance
“Recent enhancements in machine learning are helping transaction authentication and authorization with fraud
banks to improve their anti-money-laundering (AML) detection without decreasing customer satisfaction,
programs significantly, including, and most immediately, and route events securely across the whole payments
the transaction monitoring element of these programs.” ecosystem with efficiency.

To be fully effective, AI and ML need a big data set. They can A platform for the future
only make decisions based on access to historic datasets. EDA also provides a platform for the future, allowing
The first thing to do is to "train" the model by buying data banks to innovate outside of just countering fraud and
or scraping from a bank's own historical datasets. Then money laundering. According to PwC (https://tinyurl.
the model runs through several fraudulent transactions, com/2d8mhcry) EDA will help traditional banks compete
so it is now trained on what a fraudulent transaction looks in the new world: “[B]anks need to deliver products and
like. The objective is to build an understanding so the AI/ services faster in order to compete," PwC wrote. "A large
ML can pick out the right (fraudulent) activities. bank, with its legacy systems, can now compete against an

Speedy policing of fraud and money laundering online mortgage lender—and deliver a broader portfolio of
products to customers with more speed.”
Ideally, banks should build one model set for fraud and
one model set for money laundering and implement both Newer fintech market entrants have significantly less
models across all transactions and payment channels. technical debt than traditional FIs. Imagine a new FX rate
And this is where EDA enables them to leverage their provider that can provide payments to every country and
fraud and money laundering data models and use AI/ML give customers the best FX rates. Everything is built on
technology in true real time across an ever-expanding a modern infrastructure anyway; there is no legacy core
number of payment channels. banking app; everything is microservice; everything is in
the cloud.
EDA allows banks to build an enterprise IT architecture that
lets information flow between applications, microservices But EDA as an approach to enterprise IT architecture can
and connected devices in a real-time manner throughout help traditional banks introduce new services and link
the business. applications quickly and at scale, ensuring they can match
these agile competitors and provide customers with the
EDA works with a middleman known as an event broker, instant kind of feedback they seek, while not being held
which enables what’s called loose coupling of applications. back by large volumes of existing technical debt. The
This is essential because it means applications and devices challenge for larger banks is to move more toward real
don’t need to know where they are sending information, time—even with a large amount of technical debt.
or where information they’re consuming comes from. But Mat Hobbis is chief Architect FSI at Solace, https://solace.com, which
the event broker does.
helps enterprises adopt, manage and leverage event-driven architec-
ture. Contact him at linkedin.com/in/mat-hobbis-609758.
35