Picture two images side by side.
In the first, a gleaming AI system holds the strings of a white-coated physician puppet. The doctor moves when commanded. Schedules are set by algorithm. Productivity targets are defined by administrators. Documentation requirements are handed down without negotiation. The physician performs.
In the second image, the physician holds the controls. The AI handles the prior authorizations. The AI drafts the notes. The AI routes the after-hours messages. The physician thinks. The physician leads. The physician finally practices medicine.
Both futures are available right now. The difference is not the technology. It is who holds the strings.
And if you are employed by a large health system today, there is a better-than-even chance someone else is already reaching for them.
The Promise That Felt Like Rescue
The hospital employment model arrived as a gift.
Physicians who had spent years building practices, managing staff, negotiating leases, fighting payers, absorbing overhead, were offered a way out. Sell the practice. Join the system. See patients. Let someone else worry about the rest.
It was a compelling offer, and it wasn’t made in bad faith. The administrative burden of running an independent medical practice had genuinely become crushing. Electronic health records designed for billing, not for medicine. Prior authorization requirements that consumed hours physicians should have spent with patients. Payor contracts that squeezed reimbursements while costs climbed. For many physicians, hospital employment looked like the answer to a prayer.
They signed.
The Trap
What they discovered, often slowly and then all at once, was that the burden had not actually been removed. It had been repackaged.
The administrators who took over the business side of medicine had their own agenda. They had paid significant sums to acquire physician practices, not out of generosity, but because physician employment is extraordinarily valuable to health systems. Employed physicians generate referrals that flow through the system. They feed the imaging centers, the labs, the surgical suites, the specialty departments. In health system economics, an employed physician is not merely a clinician. He or she is a revenue engine.
Once that equation is understood, everything that followed becomes predictable.
Work Relative Value Units, wRVUs, were implemented in the early 1990’s to standardize physician payment. But, over time and in the hands of hospitals, wRVUs became the metric of control. See more patients. Bill more encounters. Generate more wRVUs. The pace of practice accelerated. Support structures were quietly eroded: fewer front desk staff, broken equipment that took months to replace, supply shortages that would have been unthinkable in a physician-owned practice. The message, never stated directly but impossible to miss, was clear: do more with less, and do it faster.
Physicians who had joined for relief found themselves burning out at rates that should alarm every patient in America. Many left medicine entirely. Some left the country for slower, patient-focused practice. Others stayed, diminished, running through appointments on a timer.
I know this trajectory, and not in the abstract. I have sat with physicians living it. And I have sat with one who reached the end of it in a way that still stays with me when I think about what we are asking of the people we trust with our lives.
The Moment That Changed Everything
A physician I work with, an internist with subspecialty training, a partner in a thriving practice before a health system acquisition, described the years after the sale in terms that most employed physicians will recognize immediately. The pressure to produce wRVUs intensified annually. Colleagues burned out and left, concentrating more patients onto fewer physicians. Support staff was cut. Equipment broke and stayed broken. The system that had promised relief had become the primary source of distress.
Then one day at work, under a load that had long since crossed every reasonable limit, the physician lost consciousness. Today, there’s no memory of passing out, just of coming back.
And in that moment, sprawled out on the floor of a clinic that a large healthcare system owned, one the physician once felt could not be left without enormous professional and financial consequences, the decision arrived with perfect clarity: I’m done.
What that physician built from that moment forward looks almost nothing like what was left. A small group practice, designed from the ground up to sustain physicians and serve patients. The physician’s income today is a multiple of what the health system paid. Vacation time has increased. The pace of practice is not just manageable — it’s been described to me as enjoyable. And the patients, if you spend any time in office, seem to notice the difference. The warmth is not performed. It cannot be performed. It is what happens when a physician is not running on empty.
The physician grabbed the strings. Everything changed.
The Next Puppet Master
Here’s what concerns me most about this particular moment in medicine.
The factory-ification of healthcare was always driven by a simple institutional logic: maximize the value extracted from every physician. As long as physicians were necessary, as the irreplaceable center of every clinical encounter, there was a floor beneath their leverage. They could be squeezed, but they could not be eliminated.
That floor is now in question.
Voices within healthcare administration, including Mitchell Katz, CEO of NYC Health + Hospitals, one of the largest public health systems in the country, have begun making the case that AI can replace physician functions. Not just assist. Replace.
The argument takes different forms depending on who is making it, but the logic is consistent: if an AI system can handle diagnosis support, documentation, patient communication, and clinical decision-making, why maintain the expensive, independent-minded, occasionally inconvenient human being at the center of care?
This is not a fringe position. It is being articulated by people who run institutions that employ thousands of physicians.
For a physician employed by one of those systems, the trajectory should be alarming. Hospital employment was the first transfer of control, i.e., from the physician to the institution. AI-as-replacement is the second. The puppet master changes. The physician remains the puppet.
The Other Possibility
But AI is not inherently the threat. That’s the critical thing to understand.
In the hands of an independent physician, one who controls her own practice, the same technology that administrators want to deploy as a replacement becomes something else entirely. It becomes the escape hatch that hospital employment was supposed to be.
Consider what AI does well in a clinical practice. It listens during patient visits and drafts notes in real time, eliminating the hours physicians spend on documentation after the workday ends, what the profession has grimly taken to calling “pajama time.” It navigates prior authorizations, the soul-crushing insurance bureaucracy that consumes physician and staff hours alike, handling paperwork and appeals without the frustration and futility that burns people out. It routes and drafts responses to routine patient communications, appointment requests, test result follow-ups, so that when a patient speaks with a practice employee on the phone, that person is unhurried and fully present.
What this means for a small independent practice is not incremental improvement. It’s a structural shift in what is possible. A practice that once required substantial administrative overhead to manage its workload can now run lean, with a smaller staff who have the time and mental space to actually care for patients. Not simply in the clinical sense. In the human sense.
The physician has cognitive bandwidth for medicine. The staff has room for care. The patient experiences something increasingly scarce: a practice where the people who answer the phone seem genuinely glad they called.
This is what physician-controlled AI looks like. The efficiency gains stay with the physician and the patient; they do not flow upstream to an institution that will convert them into higher wRVU targets and smaller support budgets.
The puppet master dynamic reverses. But only if the physician holds the controls.
The Window
Physicians possess something no health system administrator and no AI system can replicate: the trust of their patients.
That trust is earned through years of presence, through showing up, through the particular kind of attention that medicine at its best demands. It is, right now, the most durable asset in healthcare. And it is entirely portable. It goes where the physician goes.
The case for independence has always been sound. What has changed is that AI has made it more feasible than it has ever been. The administrative burdens that once required institutional infrastructure can now be handled by tools a small practice can deploy, control, and adapt to its own needs. The barriers to independence are lower. The promise of hospital employment, relief from the burden of running a practice, has been exposed by an entire generation of physicians who bought that promise and paid dearly for it.
The window is open.
It will not stay open indefinitely. As AI embeds itself deeper into health system operations, as the tools that could liberate independent physicians get captured and reconfigured to serve institutional productivity, the leverage physicians carry today will erode. The time to act is not after AI has been fully institutionalized. The time to act is while the physician still holds the strings.
The two images remain available. One is a future in which AI becomes the latest mechanism of control, another layer in the factory, another justification for reducing the physician to a function that can be optimized, measured, and ultimately replaced. The other is a future in which the physician uses AI the way any skilled practitioner uses a powerful instrument: deliberately, in full control, and entirely in service of the work that matters.
Which future you inhabit is still a choice. Choose carefully. Choose now.
