SSA AI Chief: Agentic AI Is Ready With Governance, Testing – MeriTalk

Source Domain: www.meritalk.com

The Social Security Administration (SSA) believes that agentic artificial intelligence (AI) is poised to support agency missions, but effective implementation requires robust governance, continuous testing, and modernized data infrastructure.
SSA’s chief AI officer, Brian Peltier, emphasized the agency’s transition towards using matured AI capabilities alongside significant technology modernization.
For successful AI deployment, SSA recommends agencies determine if AI is the appropriate solution for a problem, establish clear metrics and desired outcomes, and involve multidisciplinary teams throughout development and deployment phases.
To ensure security and effectiveness, constant monitoring, testing, and safeguarding measures are crucial throughout AI’s production deployment.
Peltier highlighted that AI should enhance service delivery across all customer channels rather than only focusing on digital interactions, thereby freeing up resources for more personalized, in-person assistance.
SSA has successfully integrated AI with a retrieval-augmented generation (RAG) tool, combining policy documents with a large language model, improving internal search capabilities and collaboration, and increasing employee use of the tool.
The RAG tool’s usage has significantly improved the ability of frontline employees to serve customers by providing direct links to authoritative policy documents.
SSA’s AI capabilities also facilitate information security teams and inspectors general, and the tool is accessible through an API for integration into other agency applications.