Policy memo (297.38 KB)
| DATE: | January 10, 2024 |
| SUBJECT: | Supplemental Nutrition Assistance Program – Use of Advanced Automation in SNAP |
| TO: | All SNAP State Agencies All Regions |
The Food and Nutrition Service (FNS) has seen an increased interest from state agencies in adopting advanced automation technologies for Supplemental Nutrition Assistance Program (SNAP) operations, including use of robotic process automation (RPA) and artificial intelligence (AI). These advanced automation technologies could be useful tools in assisting state agencies with workload and staffing challenges and improving customer service. FNS encourages and supports state agencies’ use of advanced automation technologies to enhance the administration of SNAP and foster public trust, both in SNAP and in the state agencies’ systems, within the framework of statutory and regulatory requirements.
This memo provides state agencies with guidance on allowable use of advanced automation technologies. Depending on the type and functions of advanced automation technology, state agencies are either required to notify FNS in accordance with 7 CFR 272.15, major changes in program design, or request a waiver from FNS in accordance with 7 CFR 272.3 or section 17(b) of the Food and Nutrition Act (FNA).
Definitions
For purposes of this memo, please refer to the descriptions below:
- Robotic Process Automation (RPA): an approach or system that uses software robots or “bots” that are programmed to perform specific tasks in an existing system using structured inputs and rule-based, non-discretionary logic.
- Artificial Intelligence (AI): a machine-based system that can make predictions, recommendations, or decisions for a given set of human-defined objectives. Machine learning means a set of techniques that can be used to train AI algorithms to improve performance at a task based on data.
Advanced Automation Technologies That Require FNS Notification of a Major Change
State agencies must notify FNS when they make major changes in their operation of SNAP in accordance with 7 CFR 272.15.
Major changes include substantial increased reliance on automated systems for performance of responsibilities previously performed by state merit system personnel, changes in the way that applicants and participants interact with the state’s SNAP agency, or replacing or adding functionality to existing automated systems used in the certification process. Use of bots to perform allowable functions currently performed by a state agency’s automated eligibility system or non-merit system personnel is considered a major change and subject to the major change notification and reporting requirements. This is because, by themselves, bots are not much different from many state agencies’ eligibility systems that have more sophisticated functionality built into their system process. Bots that are considered major changes must conform with all existing statutory and regulatory requirements, including merit system personnel, language access, civil rights protections, and privacy safeguards that prevent inappropriate disclosure of information.
Examples of advanced automated technologies that FNS has acknowledged as major changes include:
| State | Advanced Automated Technology Description |
| Arizona | Uses a voice virtual agent that routes phone calls and accepts periodic reports. The voice virtual agent collects caller responses to periodic report form questions through voice-to-text technology. |
| Georgia | Uses voice and digital virtual agents at their call centers to provide case status information and accept changes in household composition and address. The virtual agents create a flag for the merit worker to review the case and make subsequent updates that require discretion and judgement. These agents can communicate in multiple languages. |
| Georgia | Uses RPA to prepare, but not certify, certain SNAP recertification applications when applicants have electronically provided documents for all reported information. The bot populates the information, identifies mismatches between what is on the application and already in the eligibility system, and creates flags for merit workers to review. The merit worker then reviews the case to recertify the household. |
| Illinois | Uses RPA to review and automatically process a no-change periodic report. If the periodic report is incomplete or indicates changes, the bot creates a task in the eligibility system for the merit worker to process the periodic report. |
| Kentucky | Uses a virtual agent for recipient/applicant to check case status. |
| Kentucky | Uses optical character recognition (OCR) to scan verification documents and upload them to the eligibility system. The merit worker receives a task to assign the scanned documents to a case. |
| Kentucky | Uses a chatbot to respond to general inquiries, such as how to apply for benefits, status of documents, application status, and benefit amount. |
| Kentucky | Uses a virtual agent to automatically process no-change periodic reports. The virtual agent captures questions, responses, and telephonic signatures. Additionally, when an authenticated household calls, the virtual agent will provide a nudge regarding the status of requested information and reminds the household when the information is due. If reports are incomplete or indicate changes, the virtual agent creates a task in the eligibility system for a merit worker to process the periodic report. |
| Massachusetts | Uses a virtual agent that provides households additional access to case information as well as allowing the state agency to communicate important dates. |
| New Mexico | Uses RPA to collect information from other locations, such as from an interface or client portal, and to populate the information into the eligibility system, add case notes, and create a task for a caseworker. These include:
|
| New Mexico | Uses RPA to automatically process no-change periodic reports. If the reports are incomplete or indicate changes, the bot will create a task for the merit worker to process the interim report. |
| Ohio | Uses a virtual agent chatbot to answer general inquiries about case status, benefit amount, and eligibility. |
| Texas | Uses RPA to collect data from the Bureau of Vital Statistics and Social Security Administration to compare against the eligibility system for a perfect match. The bot will automatically update the case to send a Notice of Adverse Action to one person households that are reported deceased. |
Advanced Automation Technologies that Require FNS Approval Prior to Implementation
FNS understands there is a growing field of advanced automation technologies that go beyond those mentioned in the Major Changes section above. The quickly evolving field includes technologies that use logic to flag case record materials or data for additional review beyond standard case processing requirements or that use machine learning as AI. In deciding which advanced automation technologies need FNS approval prior to implementation, FNS considered various factors including the President's Executive Order 14110 (Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence) which requires that executive departments and agencies adhere to eight policies and principles as stated in Section 2.
These policies include requirements for responsible innovation, supporting American workers, equity, civil rights, privacy, and risk management.
Based on such considerations, FNS has determined that state agencies must submit a request to FNS to use AI or any other type of advanced automation technology not mentioned in the Major Changes section, above, for review and approval. Since these types of advanced automation technologies may introduce bias or errors into eligibility determinations, FNS expects state agencies to demonstrate how the design and use of such technology mitigates potential bias. Furthermore, state agency requests must include an explanation of how the advanced automation technology is anticipated to affect program access, the duties of merit systems personnel, system security, equitable program administration, and privacy. States must additionally provide a summary of the procedures that will be used to audit individual actions taken by the technology in support of an appeal and procedures that will be used to monitor the ongoing accuracy and equity of the technology.
In addition, state agencies that would like to implement any type of advanced automation technology not mentioned in this memo that would implement policy outside of the regulations would need to submit a waiver request in accordance with 7 CFR 272.3. Likewise, state agencies that would like to implement advanced automation technology not mentioned in this memo that would implement policies contrary to the FNA would need to submit a request for a demonstration waiver under section 17(b) of the FNA. Such demonstration projects require a robust evaluation and cost neutrality.
FNS encourages state agencies to explore opportunities to innovate and improve the delivery of SNAP benefits. FNS appreciates its partnership with state agencies and looks forward to engaging modernization efforts that help state agencies improve their administration of the program while maintaining a high level of customer service.
Please contact your FNS regional office representatives with any questions.
Moira Johnston
Acting Director
Program Development Division