OpenAI has launched computer-use agents that now automate medical order preparation, insurance authorisations and cross-platform patient data retrieval for healthcare teams.
The agents can work across electronic health records, insurance portals and other clinical software, handling tasks such as drafting orders, submitting prior-authorisation requests and pulling patient information from multiple systems.
In a demonstration, an agent used synthetic patient data to review medical records and imaging notes before drafting an MRI order and a clinical note, then prepared an insurance authorisation request for a specialist.
Major health organisations are already integrating the technology: CVS Health used an agent to clear months of document-sending backlog in a single weekend, while Cigna Healthcare leverages the models to aggregate data from more than ten clinical and administrative systems for care managers.
Safety testing shows strong clinician confidence: 99 percent of 325 tasks were rated clinically safe, doctors found the model's responses very or extremely helpful in 93 percent of 646 assessments, and GPT-6 Astra scored 94.3 percent on the Stanford-developed HealthAdminBench benchmark.
Healthcare firms can control which applications and patient records agents access, restrict actions and require staff approval, with role-based access controls and audit logs; OpenAI does not use organisational data to train its foundation models.
The computer-use features are available through the OpenAI API, ChatGPT for Healthcare and ChatGPT for Clinicians, and existing customers can enable them via ChatGPT Work and Codex.