Arlington's Excella is using AI to find fraud risk in government pandemic spending - Technical.ly DC

Civic

Arlington’s Excella is using AI to find fraud risk in government pandemic spending

The Pandemic Response Accountability Committee recruited the Arlington-based Agile agency to look for instances of fraud, waste and system abuse in the $5 trillion pandemic relief packages.

Jimmy Benani, Excella’s director of federal health and civilian markets.

(Courtesy photo)

Since the start of the pandemic in March 2020, we’ve reported a lot on the progression of federal aid packages and Paycheck Protection Program (PPP) loans.

But the necessity of the aid programs doesn’t mean they are without flaws, and with so many agencies tackling the different facets, the support still comes with certain risks, for both recipients and distributors. To counter this, the federal government’s Council of Inspectors General on Integrity and Efficiency created the Pandemic Response Accountability Committee (PRAC) to provide insight into the huge amounts of spending in pandemic relief aid. And PRAC has recruited Arlington-based Excella, an Agile technology firm, for a two-year contract to bring more transparency into the $5 trillion distributed throughout the pandemic via government spending.

The risks, Excella Director of Federal Health and Civilian Markets Jimmy Benani told Technical.ly, primarily come from the fact that various pandemic responses — think the CARES Act, PPP loans, healthcare help and unemployment — all come from different agencies, and they’re not checking in with one another.

“All of that data is housed individually within their respective agencies,” Benani said. “So, what PRAC is trying to do is really collect all of that data into one place, first of all, and then once they have this data, then be able to start to make some decisions and evaluate and then identify fraud, waste and abuse.”

Some of this might come in the form of people committing fraud, like applying for PPP loans when they don’t actually need it or impersonating someone else. It’s worth noting that of the 11 million PPP loans distributed, totaling around $792 billion in loans, the New York Times reported that around 15% have at least once indicator of fraud.

Advertisement

But some of it, Excella VP of Engineering Claire Walsh said, also might come in the form of waste, like government agencies accidentally sending multiple checks to the same person.

“You’re actually doing a much bigger kind of discovery effort, where you’re looking at the variety of different spending that’s happening, you’re looking for patterns, you’re looking for anomalies and you’re basically looking for potential where there might be fraud schemes,” Walsh said. “And they could be different kinds of new use cases and examples than we’ve seen before, because this is a bit of an exceptional program.”

Another large part of the problem, Benani said, is people “double dipping” into multiple government agencies. This could take the form of people intentionally trying to game the system by applying for multiple types of aid or applying to multiple agencies, not realizing that qualifying for and receiving one type of aid might take you out of the running for another.

The problem there, Benani said, is that government agencies don’t interact enough with each other to know this is happening.

“Quite frankly, no one is really looking at that because they’re very focused on their specific data and their fraud,” Benani said.

Claire Walsh. (Courtesy photo)

Excella is taking on these discrepancies by using machine learning and artificial intelligence to process the large amounts of data and make recommendations, although Walsh noted that it will be keeping a human element in the process when making financial decisions. It will primarily be developing a framework to help identify and prioritize pandemic-related spending as well as potential fraud schemes in agencies like the Small Business Administration, Health and Human Services and the Department of Treasury.

“Every time we start an AI project, we’re looking at, you know, what are the potential impacts here?” Walsh said. “Where does it make sense to leverage something and have a machine take an action or make a recommendation? And where does it need to have a human in the loop? So that mindset of a human in the loop is how we approach all of these problems.”

Benani said that presently, there are still many individuals, groups of people and businesses that are still affected by the pandemic. Ultimately, he added, the risks noted above clog the system and prevent the people that still need pandemic aid — like PPP loans, healthcare for loved ones or financial stability in the case of job loss — from getting their necessary funds.

“The government is doing the best they can to provide funds to help individuals and it’s very disheartening when those funds are getting abused by people that are taking advantage of it for their own benefit,” Benani said. “So if we can help make an impact from that perspective, to ensure that those funds are getting to the right people that [they] were intended to, that impact and the mission focus is so strong, that it really motivates the team.”

-30-
Subscribe to our Newsletters
Connect with companies from the Technical.ly community
New call-to-action

Advertisement

Technically Media

Sign-up for daily news updates from Technical.ly Dc