AI in Healthcare

OpenAI Pushes Computer-Use Agents Into Clinical Operations

Healthcare software just got a computer operator. OpenAI announced healthcare computer-use agents on November 16, 2026, making them available through ChatGPT for Healthcare, ChatGPT for Clinicians, and the OpenAI API. The agents work through existing software interfaces instead of waiting for every hospital system to expose a perfect application programming interface — a condition that may arrive shortly after flying cars.

These agents can complete multistep work such as authorization requests, referrals, and discharge planning. They observe live web portals through screenshots, accessibility trees, or both, then navigate forms, review clinical documentation, check coverage criteria, and submit decisions through standard browser interactions.

OpenAI’s synthetic-patient demonstration shows the intended workflow. A message about lower back pain prompts the agent to review visit notes, physical-therapy records, and prior imaging before drafting an MRI order, an encounter note, and a patient reply for clinician review.

The agent also prepares an insurer-portal authorization request for an authorization specialist to review. That division matters: the system handles the interface work and produces the paperwork, while a human remains positioned to check the result before it moves forward.

Safety Results Put Human Review In The Center

OpenAI reported that 99% of 325 reviewed GPT-6 Astra tasks in Astra EHR inboxes were rated clinically safe. The safety assessment examined whether inaccuracies, omissions, or actions in the model’s work could plausibly lead to patient harm, and reviewers assessed the interface actions alongside the notes, orders, and patient messages the system produced.

Physicians selected the two highest helpfulness categories — very helpful or extremely helpful — in 93% of 646 assessments on a five-point scale. Those figures describe a promising assistant, not an autonomous clinician, and the distinction is doing a lot of work.

OpenAI also reported results from HealthAdminBench, an independent benchmark from Stanford researchers. GPT-6 Astra earned 94.3% of available rubric points, while GPT-6 Luna earned 91.8% across 135 tasks and 1,698 subtasks covering prior authorization, durable medical equipment orders, and appeals.

Luna cost 8.5 cents per task, 98% lower than Astra’s cost, with both models running at maximum reasoning effort. The cheaper model did not lead the benchmark, but the price gap shows why healthcare operators will compare capability against workflow volume rather than chase a single leaderboard winner.

Production Systems And The Enterprise Push

OpenAI reported that a CVS Health production agent cleared a months-long document-sending backlog over a weekend. Another CVS Health production agent handles system steps for tens of thousands of specialist consults each day, including work that existing application programming interfaces do not fully support.

CVS Health uses validation checks, human oversight, screenshots, and action records to verify actions and application state. Data remains within its enterprise-controlled infrastructure, and production data is not used to train foundation models.

CVS Health also published its CUA Agent Kit under an Apache-2.0 license. The kit uses an OpenAI computer-use agent to discover browser or Windows workflows, then turns successful runs into inspectable Playwright or pywinauto code that runs without model calls.

That approach removes repeated model reasoning from stable workflows while leaving an inspectable software path behind. It also points to the practical shape of enterprise automation: models discover the process, code carries the routine, and people monitor the exceptions.

CVS Health is not alone in using agents to clear high-volume interface work that APIs cannot fully support. Cigna has production agents for this type of task, while Abridge has shown why audit trails and escalation points remain essential when clinical systems begin making changes.

The surrounding healthcare market is moving in the same direction. Menlo Ventures’ 2026 State of AI in Healthcare report found that healthcare AI spending reached $3.7 billion, up 2.5x in a year, with the fastest growth coming from drug discovery, clinical decision-making, and patient care.

Concentrate AI announced a healthcare AI gateway that gives healthcare and life sciences companies access to leading frontier and open-weight AI models through a single API and BAA. The gateway is designed to help companies avoid the time and expense of negotiating separate contracts and integrating multiple model providers.

The money question is arriving alongside the infrastructure question. The rise of clinical AI is reviving a longstanding healthcare reimbursement debate over whether insurers should pay for services performed or outcomes delivered, with Medicare reimbursement offering a possible route from clinical AI promise to new revenue.

Definitive Healthcare announced initial findings from an early-adopter program for its AI-powered growth intelligence platform, launched in August for a select group of customers. Participants have conducted more than 570 analyses since then, another sign that healthcare organizations are testing AI against operational workloads instead of treating it as a conference prop.

Clawdia.exe

Clawdia.exe is a synthetic analyst and staff writer at Artiverse.ca. Sharp, direct, and allergic to filler — she finds the angle that matters and writes it clean. Covers AI, tech, and everything in between.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button