Piloting an AI-Referral Tool to Reduce Administrative Burden and Improve Resource Matching in Workforce Services

• In Event: Does AI Deliver? Evidence from Training Programs, the Workforce, and Career Pathways

Abstract

Frontline staff in workforce and social service organizations play a critical role in connecting job seekers to education, training, employment, and supportive resources. Yet caseworkers often face significant administrative burden when identifying appropriate referrals, navigating fragmented resource ecosystems, and tailoring recommendations to individual client needs. Generative AI tools offer new opportunities to streamline these tasks, but empirical evidence on their practical value and implementation in workforce settings remains limited. This paper presents findings from a pilot evaluation of a generative AI–powered referral generation tool developed by Nava Labs and implemented with Goodwill organizations in Texas and Pennsylvania.

The referral tool was designed to integrate directly into existing case management workflows and support caseworkers during real‑time client interactions. By synthesizing client information and drawing from trusted, locally relevant resources, the tool recommends resource referrals based on clients’ needs and generates step-by-step action plans for education, workforce training, job placement, and social supports. The pilot applied human‑centered design principles, with caseworkers retaining full discretion over whether and how to use AI‑generated recommendations.

The pilot examined how the tool affected caseworker experiences, workflow efficiency, and perceived service quality. Using a mixed‑methods approach, the study combined survey and tool usage data with qualitative feedback from participating staff. Early findings indicate that integrating the AI referral tool into daily practice was associated with a 9% reduction in reported administrative burden. Additionally, early implementation findings from discussions with caseworkers highlight that successful use depends on thoughtful integration into existing systems, transparency about the tool’s capabilities and limitations, and ongoing opportunities for staff feedback.

This paper contributes to growing evidence on the role of generative AI in workforce and social service systems by documenting early, real‑world impacts on frontline practice. Results suggest that when designed and implemented responsibly, AI‑enabled referral tools can reduce administrative burden, support more efficient service delivery, and amplify caseworker expertise by adopting a “human-in-the-loop” approach. The findings offer practical lessons for policymakers, funders, and service providers considering the adoption and scaling of AI tools to strengthen workforce, employment, and other support systems to improve health and human service outcomes.

Author

• Martelle L Esposito, Nava PBC
  Presenting Author