As the $12.2 billion First Commonwealth Financial Corp. (FCFC) of Indiana, Pennsylvania, was busy acquiring five banks and another 27 bank branches between 2015 and 2025, the operations team at its subsidiary bank kept equally busy converting all the data from those institutions to its core banking system. That process also involved testing and validating the new data.
While the testing was automated, Teresa Thompson, vice president of operations for First Commonwealth Bank, says the data validation process was not. “People are still looking at data manually, they’re comparing it manually — and that’s time consuming,” she says. “When we were acquiring banks. . .you’re doing that as an extra add-on to your normal day. So, you’re working 12-,14-,16-hour days to get that in.”
Thompson hopes advances in artificial intelligence might soon lead to automated data validation. But her bank is just starting to experiment with AI. “It’s coming,” she says. “We have a little bit of AI, but we have an AI team that is new, and they’re starting to develop.”
Thompson is hardly alone in waiting for an AI-based solution. Operations executives at most financial institutions are still validating data manually after core conversions or software updates.
When it comes to converting an acquired institution’s customer data to its existing core, a bank or credit union must first run a representative sample of that data through a replicated test environment. The sample can include customer account information such as deposits, loans and interest rates. The testing is done to ensure the core software can read and process the incoming data correctly. Otherwise, data fields can contain the wrong information, interest calculations could produce errors or payment schedules could be inconsistent.
While those tests are designed to catch processing issues, they cannot ensure the data itself is completely accurate. That means someone at the financial institution must also manually look at the new data that’s been entered into the system, and compare it to the data from the original records to make sure it matches.
Thompson says if things such as balance amounts or rates don’t match the original data, it becomes both a customer service and compliance issue. “The interest rate on a loan has to be correct or you’re going to be accruing interest incorrectly and the customer’s going to be very upset. It’s very high customer impact,” she explains. “Our auditors are [also] looking to make sure that the data that we have is accurate.”
Helen Rivera Astorga, an automation analyst at the $8 billion Hudson Valley Federal Credit Union, based in Poughkeepsie, New York, has never been through a core conversion, but she has dealt with the aftereffects of a conversion without good data validation. “There were a lot of times when there was no information or documentation of what was done,” she says. “So, that’s when you start having little issues that you have to figure out from scratch and find out what happened at the beginning to get to where we are now.”
Jonathan Baltzell, vice president and president of bank solutions for the core banking provider Jack Henry & Associates, says he knows how important data validation is, and just how time consuming it can be. “We stepped back and looked at our implementation process and said, ‘Where are the biggest points of friction? What’s the biggest lift for the institution?’ And it’s the data validation,” he says. “Because they have to sit there [and say], ‘Here’s a screen print from my existing core. Here’s what it looks like in conversion to my new Jack Henry core. Do they match?’”
Baltzell says larger commercial core banking customers can take a month or more to complete the testing and validation process.
Core Software Updates
Many banks and credit unions also perform data testing and validation after core software updates. Milind Pathak, a community financial services consultant who spent more than 25 years with core banking provider Finastra Group Holdings, says those updates are designed to fix problems or enhance the core’s capabilities. But they can also cause unforeseen issues.
“With core banking [systems], especially the ones where deposits and lending and everything is built in, what happens is they will fix something, but they might break something else,” Pathak says. “So, that’s what they’re looking at. ‘Is there anything else now broken because they have fixed this?’”
While most such issues are discovered during testing, some are only noticed during the data validation process. For example, a comparison of account records from before and after an update might show certain interest amounts have changed. The bank or credit union may then find it was caused by a bug in the update that their testing didn’t uncover. Thompson says one software update caused First Commonwealth’s core to stop producing billing statements.
Rivera Astorga says that type of issue puts an additional time burden on the operations team. “It is a major problem because you have to come up with something in an efficient time,” she says. “You have to stop what you’re doing just to fix that one problem.”
AI to the Rescue?
The search for an automated solution could soon be on the way with the emergence of better artificial intelligence. Pathak believes banks and credit unions will soon be able to use generative AI tools that can automate data comparisons and identify any discrepancies in a matter of minutes. “I’m just going to tell AI, ‘Write a program to compare the data in the core to the data from the original records,’” he says. “I see AI automating a lot of that.”
Baltzell says that is exactly what Jack Henry is doing. They are using generative AI to build a data validation software solution for their client institutions. “In this model, AI helps build the tool, but the validation is ultimately performed by the tool rather than by AI making ongoing decisions,” he says. “We’ve seen promising results but we’re still early in testing and evaluation.” Jack Henry’s objective is to help banks and credit unions improve the efficiency and accuracy of their data validation, but not eliminate governance or human accountability.
Pathak says most core vendors are working on AI solutions, and expects them to be out in the near future. “I think the quality of the overall [data validation] system is going to go up,” he says. “It will take some hiccups before we get there, but it will happen.”
Thompson says it has to happen if smaller financial institutions are going to remain viable. “We have to be better, faster, quicker in order to stay competitive,” she says. “So, the more efficient you are, the better tools that you have, the more you can do.”
Previous