Nevada official ‘cautiously optimistic’ about AI solution to reducing SNAP error rates
jetcityimage via Getty Images
After launching a new data analytics program earlier this year, the Nevada Division of Social Services is seeing early gains in identifying erroneous or fraudulent SNAP payments.
Like dozens of states across the U.S., Nevada is facing a steep bill if the state cannot bring down payment error rates for the Supplemental Nutrition Assistance Program before October 2027. But the state’s Division of Social Services is seeing early success with a new data platform that could be a crucial tool to achieve compliance, one official says.
After President Donald Trump signed the “Big, Beautiful Bill” last July, states became subject to new rules for administering the Supplemental Nutrition Assistance Program. One of those requirements is that states must have a SNAP payment error rate below 6%, or they will have to pay a larger share of costs for the program, which is partially funded by the federal government.
Nevada, for instance, could pay nearly $50 million in additional SNAP costs in 2028 if they do not comply. Indeed, in fiscal 2025, the state had a payment error rate of 6.22%, according to a new report from the Agriculture Department. The national payment error rate was 10.62%.
The Nevada Division of Social Services recently turned to a data analytics solution developed by the SAS Institute that has helped officials save approximately $330,000 in misspent SNAP payments across 3,000 case reviews, Kelly Cantrelle, deputy administrator of program and field operations at the Nevada Division of Social Services, told Route Fifty.
There is no silver bullet to meeting the congressional threshold, but “we are hoping that [with] technology initiatives like this … we will start seeing that error rate come down,” said Cantrelle.
“We’ve got well over 200,000 SNAP cases, and there’s no way that our workers could look at all those cases every single month,” she said, adding that the SAS platform “is going to help us … put an eye on cases that we can’t look at.”
The data analytics platform, which went live in April, uses AI and machine learning to analyze beneficiary data from DSS and external sources, such as income databases, to alert staff of risk factors that could lead to incorrect SNAP payments, Cantrelle explained. Based on the data, the platform generates a monthly risk assessment of SNAP determinations that are most likely to be erroneous.
DSS staff can then triage and address at-risk cases more efficiently, reducing time and money spent to review and correct cases on an ad hoc basis. For example, errors in shelter expenses are a common mistake made by staff, Cantrelle said. Employees may mistype a client’s reported rent payment into their case management system, and that could result in an over- or underpayment outcome.
A report from the new system can more quickly flag that incorrect data point by cross-verifying average rent incomes in the client’s ZIP code, alerting staff that the original recorded payment does not align with existing data sources, Cantrelle explained.
The system will also flag potential fraud by notifying staff of instances where, for example, multiple addresses or phone numbers appear in SNAP cases, which is indicative of fraudulent activity. In fact, the system identified 45 fraudulent cases in its first month of implementation, Cantrelle said.
DSS’ quality assurance team can then more quickly amend the incorrect data point or, when needed, contact a client for further information to ensure the next or any future SNAP payments are calculated correctly, she said.
Preventing misspent dollars leaving the agency door is “much easier” than correcting an incorrect payment after the fact, Cantrelle said. Even if the payment error was the agency’s fault, the client still has to pay the amount back months later.
“It’s way too soon to say we’re going to make [the federal deadline],” she said, “but I am cautiously optimistic.”