AI Pushed Doctors Onto the Picket Line

Monday morning, September 14, 2026. Instead of walking rounds, about 150 doctors at Allina Health's Mercy Hospital in Coon Rapids and its Unity campus in Fridley, Minnesota, walked a picket line. Hospitalists, OB-GYNs, psychiatrists, critical care physicians. Four days off the floor. Believed to be the first strike by private-sector hospital physicians in U.S. history.

Pay and sick leave were on the list. So was something newer: a demand for a contractual say in how AI touches patient care.

What the doctors are actually objecting to

This isn't a strike against diagnostic algorithms or ambient scribes. It's a strike about who holds the pen when software starts making suggestions inside the workflow.

Dr. John Wust, one of the striking physicians, put it directly: "We feel like we're being excluded from decisions that affect the care of our patients. And without physician input, we believe that medical resources are going to not be allocated in a manner that's best for everyone in our community." He also described a narrower, sharper concern: AI tools embedded in the EHR nudging how visits get coded, tilting toward what maximizes billing rather than what reflects the actual visit.

That's the part worth sitting with. Not "the AI got a diagnosis wrong." The AI is shaping incentives quietly, inside documentation, in a place doctors don't get to audit or override before it affects the record.

Allina's response leaned on process: physician-led forums and governance structures, the company says, ensure "broad clinical input and thoughtful review before changes are implemented." The hospital also noted the union represents roughly 10% of physicians across the two campuses, and that contingency staffing kept both sites open through the strike.

Why this is a governance story, not an AI-quality story

Most clinical AI coverage this year has been about accuracy: benchmark scores, FDA clearances, time saved on notes. Allina is a different kind of data point. It's what happens when AI is good enough to be embedded everywhere in the chart, but the people using it every day have no formal channel to say "not like this" before it ships.

A billing-nudge feature doesn't need to be wrong to be a problem. It needs to be opaque, and it needs to sit between the doctor and the patient without the doctor having asked for it. That's a structural complaint, not a model-accuracy one, and it's not solved by a better algorithm. It's solved by giving the physician visibility into what the tool is suggesting and why, and a real mechanism to push back.

This is the same fault line we keep seeing in ward-level AI adoption generally: point tools get installed into the EHR by administration, optimized for throughput or revenue capture, and the clinician discovers the behavior after the fact, mid-shift, with a patient in front of them. Verification and override need to be part of the interface, not a committee meeting six months later.

The part that should worry every hospital running AI in the chart

Strikes are lagging indicators. By the time 150 doctors walk out over a first contract, the friction has been building for a long time. The AI-in-the-EHR complaint likely didn't start this month. It just became one more item on a long list of things doctors felt they had no say in.

Any health system rolling out AI-assisted coding, AI-assisted triage, or AI-assisted anything inside the clinical workflow should read this as an early warning rather than a one-off labor dispute. The doctors aren't asking AI to leave. They're asking to be in the room when it's configured, and to be able to see what it's doing to their notes and their patients' bills in real time.

That's a low bar. It's also apparently not the default anywhere yet.

Sources

• Striking Allina doctors seek protections as AI plays growing role in patient care, Star Tribune via Yahoo News
• 150 Allina Health doctors start four-day strike after contract negotiations fail, CBS News Minnesota