Wow, the Food and Drug Administration (FDA) just can’t stop delivering AI milestones!

The agency just launched Elsa 4.0, a major upgrade to its internal AI assistant, now capable of building personalized AI agents for staff workflows, analyzing datasets, creating visualizations, and converting scanned documents into searchable text.

They're also consolidating 40 internal data sources, systems, and portals into a single agency-wide platform called HALO.

This is the same FDA that last week announced real-time clinical trial reporting with AZ and Amgen piloting the approach this summer (read my previous Note on this).

Two moves in quick succession, one facing outward (how trial data reaches the agency), one facing inward (how staff actually work with that data once it arrives).

The inward-facing part tends to get less attention, but it might matter more. Real-time data reporting only works if the agency can actually process and act on it at speed. Consolidating 40 fragmented data sources into one searchable platform and giving reviewers AI-powered tools to work through it is a kind of infrastructure that makes the real-time trial vision potentially more plausible rather than aspirational.

Worth watching whether this is a sustained modernization push or a burst of announcements...

2026 is the year of agentic AI, for sure. I am not saying it will suceed, early to say, but the push is real and widespread.