FedCiv Agencies Are Deploying Agentic AI. The Workforce Isn’t Fully Prepared.

# What's happening now

# Which agencies are using agentic AI

• Food and Drug Administration: The FDA uses agentic AI to support meeting management, pre-market reviews and post-market surveillance. Jeremy Walsh, the FDA's first chief AI officer, said the technology is expected to help reviewers streamline tasks and "ensure the safety and efficacy of regulated products."

• Counsel, Taxpayer Advocate Services and the Office of Appeals to process customer requests. Salesforce emphasized agents will not disperse funds or make final decisions.

# Scale and oversight metrics Budget reported a surge in federal AI adoption in 2025. The Federal Agency AI Use Case Inventory, published in April, identified 3,611 AI use cases across agencies as deployments expand.

# The workforce problem Multiple signals in recent reporting and oversight work show workforce capacity is the limiting factor for safe, effective AI adoption:

• Government Accountability Office (GAO) findings: GAO warned that prior workforce cuts at the IRS reduced the number of employees with skills needed to develop and support AI tools. The IRS has not created a workforce plan to address the gap.

• VA implementation issues: During a House Veterans' Affairs subcommittee hearing, Rep. Nikki Budzinski reported that some automation and AI pushed to VA employees produced incorrect information and even slowed production.

• Calls for training infrastructure: GAO's chief scientist, Sterling Thomas, recommended creating a federal digital services academy to help agencies build AI-ready teams. At the VA, Jessica Salyers, chief learning officer for Veterans Health Administration, acknowledged growing training needs and said the VA is open to industry partnerships to develop skills for AI integration.

# Practical implications for managers and policymakers

• Short-term: Expect more pilot projects and targeted deployments in contact centers, review pipelines and case processing. These systems will require human oversight design, robust testing and clear limits on automated decision authority.

• Mid-term: Agencies need workforce plans that map existing roles to AI-augmented responsibilities, define required competencies (data engineering, model governance, prompt engineering, human-in-the-loop design), and fund upskilling or hires.

• Policy options: Centralized training resources—such as a digital services academy—contracting partnerships for reskilling, and agency-level plans tied to specific AI use cases can reduce deployment risk.

# Near-term events Oct. 29 will include a panel, "Digital Labor in the Workforce of Tomorrow," where agency leaders and industry providers will discuss agentic AI impacts on federal jobs and training approaches.

Agency deployments show agentic AI can automate multi-step work across mission areas, but deployment speed outpaces workforce readiness. Concrete workforce planning, training programs and centralized support mechanisms will determine whether agencies can scale agentic AI without introducing operational or safety risks.