The U.S. Food and Drug Administration has deployed agentic AI capabilities to all employees as of December 1, 2025, building on the June 2025 launch of its generative AI tool Elsa. The agency reports that 70 percent of FDA staff have voluntarily adopted the system for tasks including document summarization, regulatory writing assistance, and research analysis. The FDA’s public report positions the rollout as a model for AI adoption across federal agencies.
Elsa launched on June 2, 2025, ahead of the agency’s original June 30 deadline set in a May 8 announcement. The tool handles foundational generative AI tasks — reading, writing, and summarizing documents — that consume significant staff time in a regulatory agency that processes thousands of drug applications, device submissions, and safety reports annually. The 70 percent voluntary adoption rate is notable because the FDA did not mandate use; employees chose to integrate the tool into their workflows based on perceived utility rather than administrative requirement.
The December 2025 upgrade to agentic AI capabilities represents a significant architectural shift. While generative AI tools like Elsa respond to individual prompts, agentic AI systems can execute multi-step workflows autonomously — retrieving documents, cross-referencing regulatory databases, drafting preliminary reviews, and flagging inconsistencies across submissions. For FDA reviewers evaluating drug safety data, an agentic system can assemble relevant prior approvals, adverse event reports, and clinical trial results before the reviewer begins their assessment.
The FDA’s approach — deploying to all employees simultaneously rather than piloting with select divisions — contrasts with how most federal agencies have adopted AI. The Department of Defense, IRS, and State Department have used phased rollouts with extensive approval processes. The FDA’s aggressive timeline reflects both the agency’s technical readiness and the operational pressure it faces: the drug approval backlog and increasing submission complexity make productivity tools a practical necessity rather than an innovation experiment.
For the pharmaceutical and medical device industries, the FDA’s AI adoption has direct implications. Faster internal processing could reduce review timelines for new drug and device applications. However, it also means the agency can conduct more thorough cross-referencing of submissions against historical data, potentially catching inconsistencies that manual review would miss. Companies submitting applications to the FDA should expect reviewers who are better informed and faster — a combination that raises the bar for submission quality.
