The federal financial management community, faced with a deluge of oversight work during the COVID-19 pandemic, is turning to automation technology to stay ahead of the workload.
The Department of Housing and Urban Development, for example, is standing up a new risk management platform called Argus, named after a hundred-eyed monster from Greek mythology.
Wilmer Graham, HUD’s chief risk officer, said the risk management platform’s artificial intelligence capabilities will give the agency the equivalent of that many pairs of eyes to monitor the performance of its grants portfolio.
When fully implemented, Argus will reduce the time spent manually monitoring HUD’s program grantees by over 75%.
“Right now, the analytics effort is quite labor-intensive,” Graham said Tuesday at the Association of Government Accountants’ Fraud 2021 conference. The process involves more than 6,000 worksheets that are input manually for review.
With Argus online, HUD will cut the time to complete an upfront review of a grant program from four months to one month.
“The programs can spend more time on actual risk assessment and opportunities to support the grantees. And probably most importantly, remove some of the human subjectivity from the process,” Graham said.
Graham said the successful rollout of the Argus platform will lead to other HUD programs adopting AI solutions to support their fraud risk management work.
“The complexities of our work environment have driven us in the direction that if we’re trying to do things manually, we will burn a lot of man-hours, and it’s no longer practical. We have to move in the direction of automating as much as we can,” she said.
HUD’s Office of the Chief Financial Officer, as part of its fiscal 2022 budget justification, said it’s working with the Office of Community Planning and Development (CPD) to stand up Argus, which will streamline CPD’s risk analysis process.
CPD serves as HUD’s primary office for affordable housing, disaster recovery and the disbursement of COVID-19 spending as part of the American Rescue Plan.
“The proposed process will include more automation, reduce subjectivity, and develop quicker results, thereby allowing CPD to load-level the monitoring workload across the year,” HUD wrote in its budget justification.
Automation is playing a more central role in grants management across most agency programs at a time when government risk experts say their agencies face a higher risk of fraud now than before the COVID-19 pandemic.
More than 300 survey respondents in a survey from AGA and Guidehouse said the most common sources of fraud and waste are benefit payments, payroll and vendor purchases of personal protective equipment during the pandemic.
However, more than a third of federal, state and local entities said they haven’t implemented technology in their anti-fraud programs.
GrantSolutions.gov, a shared service provided by the Department of Health and Human Services, is applying machine learning tools to its oversight work as part of a 10-year modernization effort that also includes moving away from paper-based processes.
Julius Chang, the director of strategic initiatives for GrantSolutions.gov, said the site uses machine learning to review grant recipients’ budgets, work plans and past performance, then flags underperforming cases.
“These are the ones that are sort of outliers, and need your help to really figure out is there any underlying problem with the grant project, or is there something else going on and they need your expertise in oversight,” Chang said.
GrantSolutions.gov processes more than $100 billion in awards across more than 2,000 grant programs and more than 10 agencies.
“We know there are a lot of smart, engaged, dedicated hardworking federal staff members that are interested in social services or infrastructure, and they want to make their projects do well. But without a competent grants management system backing them up, they get bogged down in the compliance aspect of financial systems,” Chang said.
GrantSolutions.gov is also using machine learning to run a future finding probability indicator, a tool for grant specialists that determines the likelihood of a grantee having an audit finding in the next audit cycle.
Mike Wetklow, the National Science Foundation’s deputy chief financial officer and division director for financial management, said his office is also rolling out tools to help support the rest of NSF’s program activities.
NSF’s Burn Rate Explorer, for example, uses machine learning to flag grant transactions that may need further attention. Wetklow said the machine learning tool can flag if a grant recipient is spending money too quickly or too slowly.
“What we’re working towards is having the CFO be a supporting partner to the program office to help them,” Wetklow said.
Wetklow said his office works also closely with the NSF inspector general to flag potential fraud, waste and abuse. But he said the agency is also taking steps to ensure that government-funded research at universities doesn’t improperly fall into the hands of foreign governments.