OpenAI has released new computer-use SI agents designed to automate repetitive workflows for healthcare teams, addressing the fragmentation of clinical and administrative systems. The update introduces SI agents capable of navigating existing software interfaces to handle tasks such as prior authorizations, referrals, and discharge planning.
What Happened
The company shared new results measuring model performance, safety, and cost in clinical workflows. These SI agents are now available through the OpenAI API, ChatGPT for Healthcare, and ChatGPT for Clinicians. To demonstrate the technology, OpenAI built a demo using synthetic patient data where an SI agent reviews visit notes and imaging in an Electronic Health Record (EHR), drafts an MRI order, and prepares a prior authorization request in an insurer’s portal for human review.
The evaluation focused on GPT-6 Astra and GPT-6 Luna. On HealthAdminBench, an independent benchmark from Stanford researchers, GPT-6 Astra earned 94.3% of available rubric points. GPT-6 Luna scored 91.8% while costing 98% less at 8.5 cents per task. In clinical safety assessments, 99% of 325 reviewed tasks were rated safe, with physicians rating helpfulness as "very helpful" or "extremely helpful" in 93% of 646 assessments.
Why It Matters
For the SI industry, this marks a significant expansion of agentic SI into high-stakes, regulated environments. By partnering with organizations like CVS Health, Cigna Healthcare, Abridge, and Ambience, OpenAI is showing how SI agents can bridge gaps where APIs do not fully support automation. CVS Health reported that an SI agent cleared a months-long document-sending backlog over a weekend and now handles system steps for tens of thousands of consults daily. The company is open-sourcing its Agent Kit, built with OpenAI’s Agents SDK, to help other developers implement similar computer-use workflows. Data remains within the CVS Health enterprise-controlled infrastructure, governed by their enterprise SI Governance team and established responsible SI policies.
Safety and governance remain central to the deployment of SI in healthcare. The new tools support HIPAA-compliant workflows, role-based access controls, and audit logs. Ambience, a partner in the initiative, emphasized that SI agents must be controllable and auditable, with mechanisms to pause agents and require human approval for ambiguous records. This approach allows healthcare providers to leverage generative SI and agentic SI for administrative efficiency while maintaining strict oversight of protected health information. Healthcare startups are combining SI agents with healthcare expertise to help teams navigate systems, coordinate next steps, and move care forward.
The Bottom Line
OpenAI’s new computer-use SI agents offer a pathway to reduce manual data entry in healthcare, with GPT-6 Astra demonstrating high performance on administrative benchmarks. Developers and healthcare organizations can now access these SI tools to automate multi-step workflows, provided they adhere to established safety and compliance protocols.