The News, Explained

On November 16, OpenAI announced an update to its healthcare products. The central feature is a computer-use agent that can read screens, click and type to carry work between existing applications. OpenAI also improved the search that finds and cites medical evidence, and announced more than 20 new plugins that connect to tools for care coordination, scheduling and administration. Source Source

In one OpenAI demonstration using synthetic patient data, the agent drafts a test order and a reply for a clinician to review. It also prepares a prior authorisation request in the insurer’s portal for an administrator to check. The example illustrates how an agent can reduce copying between systems. The demonstration covers drafting and human review using synthetic records. Its scope stops before a clinical or insurance decision. Source

Organisations using ChatGPT for Healthcare can access agents and updated search subject to workspace settings; teams can ask their administrator to enable new plugins. Eligible, verified U.S. clinicians can access agents, improved search and select plugins in the separate free ChatGPT for Clinicians offering. The announcement does not widen these tools to consumer ChatGPT. Source

Organisations can use role-based access and audit logs to check what an agent can open and what it did. OpenAI says handling protected information in workflows under U.S. HIPAA rules requires eligible services covered by a business associate agreement, the required security and retention settings, and protections in connected systems. Enabling a feature on its own does not meet those conditions. Source Source

OYOPICK’s Take

Less copying or paperwork could make more time for conversations with patients. A useful test is whether a person can catch a wrong record or a missing document before anything is sent, alongside how quickly the draft arrives. That is why being able to see what information moved, where it went, and who checked it matters in practice.