OpenAI pushes AI agents to automate US healthcare paperwork
OpenAI wants its GPT-6 Astra and GPT-6 Luna models to handle administrative healthcare tasks, while CVS Health and Cigna Healthcare are already deploying the technology.
OpenAI wants its upcoming GPT-6 Astra and GPT-6 Luna models to take over administrative work in the US healthcare system, from filling out medical forms to filing insurance requests. The company is pursuing this through computer-use agents, software that operates existing clinic systems and insurance portals in the same way a human staff member would, clicking through screens and entering information.
According to the report, the smaller Luna model is designed for routine work such as prior authorizations, the approvals insurers require before covering certain treatments or medications. The cost of such a task is reported at about 8.5 cents each, a figure that highlights the economic appeal of automating repetitive paperwork that currently consumes large amounts of staff time in clinics and hospitals.
The push is not purely theoretical. Companies including CVS Health and Cigna Healthcare are already deploying the technology, suggesting that large healthcare and insurance organizations are willing to test AI agents in day-to-day operations rather than only in pilot projects. The administrative burden of American healthcare is widely cited as one of its most persistent problems, and automation of this kind is positioned as a way to reduce it.
The distinction between the two models matters for how the technology would be used. Astra and Luna are presented as part of the same planned generation, with Luna assigned the high-volume, low-cost tasks that make up much of the everyday workload in medical offices. The report does not specify how the larger Astra model would be applied, leaving that part of the strategy unexplained in the available material.
The development also raises questions that the report leaves open. Automated systems that interact directly with insurance portals and medical forms touch on accuracy, accountability and patient data, areas where errors can affect whether care is approved or delayed. Whether the agents will operate with human oversight, and how their mistakes would be corrected, is not described in the available material.
Because the tools work through the same interfaces that staff already use, they do not require clinics to replace their existing software. That design choice is part of the appeal, since it allows automation to be layered onto established systems rather than requiring new infrastructure. It also means the agents inherit whatever complexity those portals contain, including the repetitive verification steps and form requirements that the report identifies as the target of automation.
For readers following the intersection of technology and faith-informed debates about human dignity and work, the shift invites a broader reflection. The question is not only whether machines can complete forms efficiently, but what roles human staff, patients and communities retain when the administrative fabric of care is handed to software. Administrative workers in healthcare are the people most directly affected by such a change, and the report does not say how their roles would be affected or whether staff reductions are part of the plans.
Source: the-decoder.com
5
