Skip to content
worth noting AI agents

OpenAI deploys GPT-6 Astra and GPT-6 Luna agents for healthcare administration

only one source so far

OpenAI is deploying the computer-operated agents GPT-6 Astra and GPT-6 Luna, which fill out medical forms and insurance requests by clicking within existing systems. GPT-6 Luna can complete a task for 8.5 cents, and companies like CVS Health and Cigna Healthcare are already testing them.

OpenAI has introduced the GPT-6 Astra and GPT-6 Luna agents, which are taking over administrative work at US healthcare organizations – filling out medical forms, creating test orders, and submitting prior authorization requests to insurers. The agents work directly within existing electronic health record interfaces and insurance portals by clicking and typing like a human, because many of these systems lack an API. According to OpenAI, the agents' proposals are reviewed by physicians and administrative staff before submission.

According to OpenAI, physicians evaluated 325 tasks performed by the GPT-6 Astra agent and rated 99% of them as clinically safe, while 93% of 646 assessments rated them as very or extremely useful. On the independent HealthAdminBench benchmark from Stanford University researchers, GPT-6 Astra scored 94.3%, while GPT-6 Luna scored 91.8% at a cost of approximately 8.5 cents per task, which according to OpenAI is 98% cheaper than Astra. Both models ran at the highest level of reasoning effort.

The agents are already being tested or deployed by CVS Health, Cigna Healthcare, Abridge, Ambience, and Assort. According to OpenAI, CVS Health used an agent to clear a months-old document backlog in a single weekend, and another agent processes tens of thousands of consultations daily; the company published its architecture as Agent Kit on GitHub. Cigna Healthcare uses the agents to merge patient data from more than ten systems.

To work with protected health data, organizations must enter into a BAA (Business Associate Agreement) with OpenAI under HIPAA; the company states that it does not use organizations' data to train foundation models and offers role-based access control and audit logs. In CVS Health's deployment, data remains within infrastructure controlled by CVS, which reviews the agents' steps using screenshots and action logs. The source does not specify where these screenshots are processed or how long they are retained, and does not mention the European GDPR regulation — the offering is currently focused on the US market.

What changed

Why it matters

For healthcare organizations and insurers, this is a tool that can speed up or reduce the costs of highly fragmented administrative work (insurance authorization, test orders, merging records from multiple systems) without having to build API integrations. According to the source, the practical impact will depend on the reliability of the connection to existing systems, the available options for overseeing the agents, and whether clinically or legally significant steps remain in human hands.

Relevant practical impact

What this means

01

For a business

Healthcare organizations and their partners can automate administrative processes (prior authorizations, test orders, merging data from multiple systems) using agents that operate existing interfaces without an API, but they must secure a BAA agreement, access control, and human review of outputs before submission.

Processes
What to decide Before deploying similar agents, verify with OpenAI the terms of the BAA agreement, access control settings, and the human review mechanism for clinically or legally binding outputs.
More business impacts →
Automation GPT-6 Astra GPT-6 Luna OpenAI software agents healthcare administration

Check the original

Event sources

only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.

1
The Decoder (daily AI news) independent context · first detected AI agents are starting to fill out medical forms and file insurance requests in place of human staff