AI · 18 November 2026
AI agents begin handling US prior authorisation requests
OpenAI is reportedly proposing computer-use agents, GPT-6 Astra and GPT-6 Luna, to automate US healthcare administration. Luna is reported to handle prior authorisations at about 8.5 cents each.
What happened
OpenAI is seeking to automate administrative work across US healthcare using two models, GPT-6 Astra and GPT-6 Luna, deployed as computer-use agents. Rather than relying on bespoke integrations, the agents operate existing clinic software and insurance portals in the way a human staff member would, according to reporting by The Decoder.
The most concrete example in the reporting is prior authorisation. Luna is reported to handle these requests at about 8.5 cents each. The same reporting says CVS Health and Cigna Healthcare are already deploying the technology.
The source material does not give a specific announcement date for these deployments, so the timing should be read as reported rather than independently confirmed.
Why it matters
This marks a shift from AI that answers questions to AI that carries out multi-step work inside systems built for people. Computer-use agents do not need a vendor to rebuild a clinic's software or a payer's portal, because they navigate the interfaces already in place. That lowers the barrier to automation for organisations running on legacy systems, which describes much of healthcare administration.
For service leaders, the change moves where the experience is designed. Patients and clinicians still feel the outcome of a prior authorisation as a wait, a follow-up or a denial, yet more of the work in between is performed by software. Operators will need clear escalation routes, audit trails and defined accountability for cases where an agent completes a form incorrectly or not at all.
By the numbers
- About 8.5 cents per prior authorisation task handled by GPT-6 Luna, as reported
- Two agent models, GPT-6 Astra and GPT-6 Luna, proposed for US healthcare administration
- Two named organisations already deploying the technology: CVS Health and Cigna Healthcare
The Renascence take
Most coverage will focus on the cost per task, but the cheaper form is the easier part of the story. The harder question is what happens to the patient who never sees the paperwork, and whether the staff who once owned that work still understand why a claim moved the way it did.
Automating a task does not remove its cognitive load; it shifts that load to whoever must check, explain and recover from the output. Customer-obsessed operators should treat agent-completed administration as a service design problem first, mapping every handover point and designing the moment when a patient asks why a decision was made. The organisations that win here will be the ones that keep a human able to explain the outcome, not just the ones that cut the cost per form.
Sources
This briefing was written by our Newsdesk, synthesising reporting from the outlets below. Follow the links for the original coverage.
FAQ
Questions we get on this topic
More in AI
Stay ahead of CX
Get the signal, not the noise.
The stories shaping customer experience — plus the Journal and Experience Loom — in your inbox.


