Cross-Agency Matching for Summer EBT Enrollment
← Go to Use Cases
New Jersey

Cross-Agency Matching for Summer EBT Enrollment

SUN Bucks (the federal Summer Electronic Benefit Transfer program) provides $120 in summer grocery assistance to children from low-income families, but eligible children often go unenrolled because the program data lives in different agencies than the benefits they already receive. Matching eligibility manually across the state's Departments of Children and Families, Education, Agriculture, and Human Services would mean searching hundreds of thousands of records and over a million entries — work too time-consuming to do at scale before the summer benefit window.

Year added2025
Status as of Spring 2026Pilot
Primary AgencyCentral IT / CIO Office
Approach and SolutionThe New Jersey Office of Innovation used AI to match public assistance records across four state departments, identifying children enrolled in programs like SNAP who were eligible for SUN Bucks but not yet enrolled. Families flagged through the match were automatically issued cards with the summer grocery benefit loaded on them, with no additional application step required.
Tags
  • Data / Data Sharing
Who it ServesThe Public
Area of ImpactHealth & Human Services
AI Technology TypeGenerative AI
Level of Human OversightAI-assisted decision (human review required)
Impact / OutcomesThe matching effort identified approximately 100,000 additional eligible children — 14% of total enrollees — and delivered more than $24 million in summer grocery benefits that would likely have gone unclaimed under a manual process.

Related

Modernized Benefits Access and Case Management Platform

Modernized Benefits Access and Case Management Platform

DES built A-to-Z Arizona, a self-service client portal developed in partnership with Salesforce, and is layering AI into its child support application. A virtual policy assistant trained on child-care policy and procedure data lets staff ask policy questions and receive answers with source references for verification, and a case analysis assistant is in development to summarize case information and suggest next steps. DES is also preparing to deploy Agentforce to help address backlogs in unemployment insurance, SNAP, and TANF interview processes, with humans kept in the loop where needed.

Virtual Assistant for Benefits Access and Housing Permitting

Virtual Assistant for Benefits Access and Housing Permitting

Some agency teams will use a Claude-powered virtual assistant to help residents apply for benefits, update their information, track applications, and identify additional programs they may qualify for, while other teams will use the technology to streamline housing permitting. The initiative is rolling out first to staff at Maryland Benefits, the Department of Labor, the Governor's Innovation Team, and the Maryland Department of the Environment, building on earlier work including a self-service portal for environmental site assessments developed with the Maryland Digital Service and Office of AI Enablement.

SNAP Policy Navigator for Caseworkers

SNAP Policy Navigator for Caseworkers

The partnership will initially focus on developing the SNAP Policy Navigator, a Claude-powered integration that allows caseworkers to seamlessly access accurate, reliable, and real-time SNAP information across federal, state, and county policies. Beyond the initial pilot, the partnership is intended to produce reusable tools and approaches that can be adapted across states and counties, helping governments modernize service delivery while remaining compliant. Code for America and Anthropic will create a full suite of Claude integrations purpose-built to help administer critical benefits programs in an increasingly complex policy environment and save time for caseworkers. These integrations will focus on assisting with tasks like answering tough policy questions, reviewing eligibility documents, and drafting plain-language communications to benefit recipients.

Medicaid Fraud Detection AI Pilot

Medicaid Fraud Detection AI Pilot

Minnesota's governor proposed a $54 million package to prevent, detect, investigate, and penalize fraud across state programs, including a pilot at the Department of Human Services that would apply AI to flag payment anomalies among Medicaid providers. The system would be trained to surface suspicious billing patterns — such as excessive service volumes, claims for services not rendered, or sudden deviations from a provider's historical norms — much like the fraud-detection systems banks use on credit card transactions.