The U.S. Food and Drug Administration (FDA) is introducing agentic artificial intelligence (AI) models to improve its regulatory workflows, as part of Commissioner Marty Makary’s ongoing efforts to modernize the agency.
Agentic AI involves machine learning “agents” that solve complex problems by simulating human decision-making processes. According to IBM, these agents are capable of planning, reasoning, and carrying out multi-step actions while following specific guidelines and maintaining human oversight.
The FDA stated that for its staff, agentic AI will help manage meetings, conduct pre-market reviews and post-market surveillance, validate reviews, assist with inspections, and support other administrative tasks. The use of agentic AI by employees is voluntary. To encourage adoption, the FDA will organize an Agentic AI Challenge over two months in which employees can develop solutions using the technology. The results will be presented in January.
“We are diligently expanding our use of AI to put the best possible tools in the hands of our reviewers, scientists and investigators,” Makary said in a prepared statement on Monday. He added that the agency is positioned to “modernize with tools that can radically improve our ability to accelerate more cures and meaningful treatments.”
This new initiative follows the launch of Elsa, a generative AI tool introduced by the FDA in June 2025. Elsa was designed to enhance efficiency by assisting staff with reading, writing, summarizing information, assessing product safety profiles, comparing drug labels quickly, creating databases for non-clinical applications, speeding up protocol reviews and scientific evaluations, and helping identify high-priority tasks.
Both Elsa and the new agentic AI program operate within a secure cloud environment without access to data from regulated industry sources.
