The Department of Housing and Urban Development (HUD) recently added three more AI use cases for the agency to work on – bringing its total number of AI use cases to four – with HUD’s chief AI officer (CAIO) saying the newer cases focus more on machine learning operations.
HUD CAIO Vinay Singh explained on Oct. 15 that the three new use cases are for HUD’s Government National Mortgage Association, or Ginnie Mae. Singh said the three AI use cases for Ginnie Mae actually “predate” President Biden’s October 2023 AI executive order (EO), but that the agency spent the past 10 months assessing them to ensure they aligned with the EO.
HUD published the three additional use cases on Sept. 24 on its AI Inventory website.
“Currently, we only have four use cases that are public,” Singh said on Tuesday during ServiceNow’s AI In Action event. “Three of them were just released … but those use cases – that can be also found on our website – deal a lot more with machine learning operations.”
The first Ginnie Mae use case automates draft counterparty credit narrative reports, which according to the AI Inventory, “enables processing efficiency and reduces errors that can occur in a manual process.”
The second use case uses machine learning algorithms to analyze counterparty risk profiles of mortgage issuers and detect potential risk areas.
The third and final use case for Ginnie Mae automates the analysis of Master Sub-Servicer (MSS) transaction data. Through automation, HUD said, “the machine learning algorithms enhance the efficiency and accuracy of key reporting processes.”
“The reason that programs such as Ginnie Mae or even FHA [Federal Housing Administration] would be pulled into AI years ago – before this executive order – is because the industry was using it,” Singh explained.
“Ginnie Mae and FHA are pulled and involved very heavily with the banking industry,” he added. “AI is 50-60 years old, and banking has been playing in this space for at least a decade. So, Ginny Mae and FHA had to look at tools that would accelerate their involvement in the market and staying level in the market.”
As for the other use case, which Singh said has been out for about a year, it aims to assist HUD’s Office of Policy Development and Research (PD&R).
HUD requires grantees of its formula block grant programs to submit what are known as “Consolidated Plans” to identify and assess affordable housing, community development needs, and market conditions. However, Singh said the PD&R team “has just not been staffed to be able to look at those,” and to conduct in-depth analysis of the plans.
“The PD&R use case is essentially more of a research initiative, and what it intends to do is look at consolidated plans that come in at the local, state level … [and] essentially assess those,” the CAIO said. “A lot of what AI can do for government that I’ve come to learn is around document summarization [and] insight extraction.”
Singh said HUD envisions “a research document coming out of that” use case.
