Can an AI Copilot Give Kaiser Doctors More Time with Patients?

Kaiser Permanente’s Permanente Medical Group is introducing Nabla’s optional AI Copilot to help clinicians turn patient conversations into notes and other administrative documents. A pilot reportedly saved doctors 1.5 hours of administrative work, but the rollout starts in Northern California and is not intended to replace clinical work.

WTF Index NEUTRAL
◄ Terminator 1 Idiocracy 1 ►

The optional Copilot aims to reduce routine paperwork while leaving clinical decisions to doctors, with little clear lean toward either risk.

Can an AI Copilot Give Kaiser Doctors More Time with Patients?

Kaiser Permanente’s Permanente Medical Group is rolling out an AI assistant from healthcare startup Nabla to help clinicians handle paperwork after patient visits. The tool listens to conversations, transcribes them and uses the information to draft notes and other documents. Its goal is to reduce administrative work, not take over doctors’ clinical responsibilities.

Starting with an optional rollout

The rollout begins with physicians in Northern California. The service will be available to physicians and clinicians in the Permanente Medical Group, but using it is optional. That group includes more than 9,500 physician staff and more than 44,000 non-physician staff.

Nabla and the Permanente Medical Group tested the service in a two-week pilot in August. Doctors in that pilot cut out 1.5 hours of administrative time using Copilot, according to the company. If the tool proves useful, the group may consider expanding it across the rest of Kaiser Permanente’s U.S. footprint.

The agreement was announced as 75,000 Kaiser Permanente workers embarked on a three-day strike over understaffing, burnout and low wages. The labor action was not focused on AI. The article raises the possibility that it affected the speed of the commercial deal, but does not establish that it did.

Turning patient conversations into paperwork

Clinicians often have forms and other records to complete after seeing patients. This administrative work can take hours each day, and some doctors put it off until the end of their shifts. The resulting after-hours work is sometimes called “pajama time,” because clinicians may finish it at home at night.

Nabla’s Copilot is designed to take information from patient conversations and supplementary documents, then prepare documents associated with a visit. Those can include prescriptions, follow-up appointment letters and consultation summaries. The intention is to help with the paperwork that follows care, while leaving the clinical work to the doctors and other clinicians.

The scale of routine computer work is one reason the company is targeting this task. Nabla CEO and founder Alexandre Lebrun said a doctor might make as many as 4,000 mouse clicks in a 10-hour shift. Automating parts of documentation could reduce time spent on those repetitive steps, though the pilot result reported by Nabla does not establish how the tool will perform across the wider rollout.

AI support, with limits

Copilot first launched in March of this year. It initially relied on GPT-3, later upgraded to GPT-4, and has since moved most of its services to Nabla’s own large language model. Nabla still uses GPT-4 in some areas, including to check the work of its own model.

That mix points to a tool built around drafting and verification. It does not change the stated boundary of the Kaiser agreement: Nabla and the Permanente Medical Group are not working on a system to take over clinicians’ patient assessments or other clinical work. The service is aimed at administrative tasks arising from care.

Other companies are pursuing different uses for healthcare AI. Corti, another healthcare startup, raised $60 million to continue developing an assistant intended to help clinicians with patient assessments. Its approach illustrates how AI tools in healthcare can range from paperwork support to systems aimed more directly at clinical decisions.

What the rollout may show

The Kaiser deployment will offer a chance to see whether an AI assistant can ease a specific source of clinician workload: completing documentation after visits. The pilot’s reported time savings provide an early signal, while optional use gives clinicians a choice about whether to work with the tool.

Whether those early results carry over across more clinicians and locations remains to be seen. For now, the plan is limited to an initial Northern California rollout, with broader use only a possibility if the service works well. The central test is practical: can Copilot reduce the paperwork burden without displacing the clinical judgment and patient care that remain with clinicians?