The news: The FDA is introducing agentic AI to help its drug reviewers, investigators, and scientists carry out more complex tasks.
- The new tech builds on the agency’s rollout of generative AI (genAI) platform “Elsa” in May to accelerate reviews and scientific evaluations and identify high-priority inspection sites.
- 70% of the agency’s staff voluntarily use Elsa, per the FDA.
- The new agentic AI is optional as well, but alongside it, the FDA launched an “Agentic AI Challenge,” inviting staff to build solutions to showcase at its scientific computing event in January.
- The FDA’s drug regulatory centers are down more than 1,000 employees, per FDA data, as part of the administration’s plan earlier this year to cut 3,500 FDA employees in total.
Why it matters: Agentic AI is the next advancement in genAI for pharma and healthcare, adding predictive and contextual capabilities—but industry adoption is still in the early stages.
- 73% of pharma leaders are planning or piloting agentic AI projects, with adoption expected to increase quickly over the next 12 to 18 months, per a global MIT Technology Review Insights survey of 250 senior leaders from May to July 2025.
- A recent McKinsey analysis found 75% to 85% of pharma workflows have tasks that could be automated by AI agents.
- Agentic AI commercial initiatives could translate to 5% to 9% reduced spending and 4% to 8% increased revenues over the next five years, per McKinsey.
Implications for pharma companies: AI-driven review is becoming a regulatory norm. Drugmakers should expect their data submissions, post-market studies, and inspection materials will be evaluated by both human reviewers and AI agents.
Pharma companies need comparable AI tools internally to validate data and identify issues before they reach regulators. However, human oversight will remain essential. While agentic AI can speed analysis, it still carries risks of errors and fabrications. AI outputs should be treated as decision support rather than final determinations.