The ACH Network

Have you ever received a Direct Deposit from your employer? Have you ever sent a loved one money via Venmo or Paypal? There are many other kinds of transactions that fall under this category, but chances are, you have come into contact with the ACH Network. The ACH Network, also known as the Automated Clearing House Network, is a network that is used to electronically transfer money between bank accounts in the US. The ACH Network is governed by a Financial Institution that creates and enforces the rules required for the ACH Network to operate safely and efficiently. If a transaction is flagged as a risk, it is the Financial Institution’s responsibility to investigate the transaction and determine if the Originator broke the rules.

The Original Way

About 4 million transactions are sent through the ACH Network weekly and the amount is constantly growing. In the past, the Financial Institution leveraged a third-party system that analyzed all 4 million transactions using the rules that the Financial Institution had created. The system flagged about 2000 that were potential risks. Then, someone would go through all 2000 transactions flagged and determine if they were actual risks. If they find a risky transaction, a case gets created and the person could potentially be fined for fraudulent activity or affecting the health of the ACH Network. 

The Challenge

The challenge that the Financial Institution faced was that not only were there thousands of transactions to sift through, they were unable to close the application without losing all of the alerts. Once the application was closed, any record of the alerts was gone. Because there was no coming back to the alerts once the application was opened, all 2000+ alerts had to be analyzed within one sitting. The other challenge was that querying the data would take a long time, so the system would time out and not give any results back. All of this prompted the Financial Institution to depart from the third-party system and build their own.

The Journey Begins

In 2019, Wiweeki started their journey with the Financial Institution to improve and optimize their Risk Analysis System. Since the alerts were constantly deleted, Wiweeki was working with virtually no data. “What we did have was the rules that the Financial Institution applies to each ACH transaction,” says David Winslow, Sr. Data Scientist at Wiweeki, “so we could implement those rules and that would narrow it down to about 1300 alerts.” So, Wiweeki went on to build a rule-based artificial intelligence model defined by “if-then” statements using the predetermined rules.

Not only did Wiweeki build an AI model for the Financial Institution, they also built a reporting engine on top of it instead of deleting the alerts after they’ve been investigated. The Financial Institution was then able to create reports with filter criteria so they could drill down to the details of every transaction. “Now, if the Financial Institution wants the details of a transaction that happened at a certain time in the past, they can easily get to it, unlike before,” says Venkatesh Kalluru, CTO of Wiweeki. 

The Journey Continues

Of the 1300 transactions that get flagged every week, only an average of two go through as risks that need to be investigated. Right now, there is no automated process that goes through the 1300 alerts to get to the truly risky ones. This is where Wiweeki sees room for more innovation.

“Now that we have over two years worth of data, we can build a smarter model,” says Mr. Winslow. 

“We are going to build a machine learning classification engine that will go through the 1300 alerts. Based on the historical evidence of which kinds of alerts were flagged as true risks versus resolved, the AI software will flag the ones that are high risk and alert the Risk Manager to watch out for them immediately,” says Mr. Kalluru. With smarter AI, 1300 alerts will be reduced to only a handful. 

The more data and the more decisions made, the smarter and more accurate the machine learning model will become. When the Risk Manager makes decisions, they are looking at things outside of the original rule, running reports, and so on. “We want to be able to replicate that thought process and do the decision-making work for them so they can focus on other work that requires a human brain,” says Mr. Winslow, “and we can do that with artificial intelligence by looking at what they’ve done in the past, getting feedback, and applying it to the model.”

It is Wiweeki’s mission to keep the ACH Network healthy and to make continuous improvements that will minimize human effort and increase efficiency. “There is no end in data analytics,” says Mr. Kalluru, “just constant evolution.”