FDA Rolls Out Agentic AI Tools for Staff Reviews and Admin Tasks
The FDA announced on December 1, 2025, that it’s making agentic AI available to all employees. These tools handle multistep tasks like premarket reviews and administrative work, but staff must review every output before using it in any official action. An FDA spokesperson told STAT News the AI stays exploratory and never replaces human judgment.
What Agentic AI Does Here
Agentic AI uses machine learning agents that plan, reason, and carry out multi-step actions, according to BioSpace and Becker’s Hospital Review. FDA built in guidelines plus human oversight for reliable results. It runs in a secure GovCloud with no access to regulated industry data and doesn’t train on staff inputs, as Pharmaceutical Executive reports.
Key Applications
- Premarket reviews
- Review validation
- Post-market surveillance
- Inspections and compliance
- Meeting management
- Administrative functions
Bloomberg Law and Fierce Biotech list these same tasks. Use is entirely optional—FDA Commissioner Marty Makary called it a way to give reviewers better tools, per Pharmaceutical Executive.
Part of Bigger AI Push
This follows the June launch of Elsa, a generative AI for summarizing adverse events, comparing drug labels, and more. Pharmaceutical Executive says over 70% of staff use Elsa voluntarily. FDA also started an Agentic AI Challenge: employees build solutions over two months and present in January 2026. Chief AI officer Jeremy Walsh said it gives staff a powerful way to check product safety and efficacy.
