How Do You Think Ai Can Improve FDA's Internal Process?

Harpreet Singh, MD, chief medical officer at Precision for Medicine and former FDA Oncology Division Director, discusses the FDA’s implementation of an artificial intelligence system known as Elsa and its potential impact on internal processes and drug review.

Singh clarifies that Elsa was introduced after her departure from the agency, but she maintains “a very active pipeline into FDA” and is familiar with key features of the system. She emphasizes that Elsa is a closed system, meaning it can only access internal FDA data. For example, if a reviewer is evaluating a novel breast cancer therapy and wants to compare it with approved agents, Elsa, at least as she understands it currently, would not necessarily provide all pivotal external data needed for a full comparative assessment. That limitation, she suggests, constrains how far AI can replace traditional, holistic regulatory judgment.

Despite these boundaries, Singh is optimistic about AI’s potential to improve efficiency. She sees value in Elsa’s ability to aggregate large volumes of information, surface key data, summarize critical points, and detect less obvious signals. Drawing on her own training, she describes a “very classic” FDA review approach in which she directly interrogated data sets, performed some programming tasks, and conducted detailed line-item data review. She believes an AI agent can be trained to assume parts of that technical workload, especially in sorting and synthesizing complex data.