The so-called deskilling of physicians is a red-hot subject in discussions of AI in health care. I hate the term “deskilling” just about as much as I hate the term “provider.” It is shallow and smacks of the MBA PowerPoint presentation. What is being lost is not some random skill. It is the physician’s core competence: the ability to synthesize a history of present illness and formulate a diagnostic and treatment plan. The more precise word is authorship. The larger word is sovereignty.
AI is being blamed. My own experience has been the exact opposite: AI strengthened my skills and taught me new ones. A short investigation revealed why. The problem is not inherent to AI. It is how AI is implemented in health care, and who chooses the implementation.
With ambient scribes and AI bolted onto electronic medical records now near-ubiquitous, it is AI, not the physician, that performs the essential intellectual duty: the HPI, the assessment, the plan. The doctor is reduced to reading and approving a plan he did not author. Iterate that a few thousand times and clinical thinking dulls. The greater fear is that newly minted physicians will never develop it at all.
Since “why” is one of my favorite questions, I went looking for the origin. The loss of intellectual sovereignty is neither random nor coincidental. It is the third act of a three-act play, and the first two acts were staged decades ago.
Real physician voices, twice a week
Free, and one click to unsubscribe.
Act I: financial sovereignty. Physicians have been the builders of their own demise from the beginning. Blue Shield was built by state medical societies. The American Medical Association (AMA) fought prepaid group practice hard enough to earn a criminal antitrust conviction, upheld by the Supreme Court in 1943, and it endorsed the service-benefit model for one reason: It preserved fee-for-service and physician-set fees. Then wartime wage controls and a tax ruling made employer-paid premiums the cheapest money in America, and third-party payment became dominant.
The bargain looked like a win. Doctors kept setting prices and someone else paid them. It held only as long as no single buyer was large enough to dictate terms. Medicare and Medicaid ended that. Once government became the dominant purchaser, the purchaser set the price, and commercial insurers followed the benchmark. The trap shut. Adjusted for practice-cost inflation, Medicare physician pay fell 33 percent between 2001 and 2025 while the cost of running a practice rose 59 percent.
The rest was mechanical. Denials and prior authorization forced the doctors to hire staff exactly when they could least afford it. The facility-fee differential, the identical service paying more when billed by a hospital outpatient department, created the arbitrage that funded the acquisition spree. Squeezed hard enough, doctors gave up control of their practices.
Act II: administrative sovereignty. Unwilling or unable to leave the third-party payer system, physicians took employment with hospitals and large health care organizations. The siren song: Join us, and we will relieve you of the financial and administrative burden and set you free to practice medicine however you choose. As of January 2026, 82 percent of practicing physicians were employed by hospitals or corporate entities and 18 percent remained in physician-owned settings, against 52 percent employed in 2018.
Note what was actually traded. The pitch was that someone else would handle the parts of practice you never wanted to do. What you handed over was not the paperwork. It was the authority to decide anything at all.
Act III: intellectual sovereignty. An employee decides nothing: not whom he hires or fires, not where he practices, not what EMR he uses. He is told his schedule and his RVU targets. And now he is told to use AI.
There is an attractive side, and pretending otherwise is dishonest. Ambient scribes do reduce the burden. Physicians get their evenings back, and no wonder surveys show they like them. The trade-off is not visible on the day it is made. What you are handing over is the thinking.
None of this follows from the technology. It follows from the design, and the design follows from the buyer. These tools could be built to transcribe what the physician dictates and stop there. Mine is. They are not, because the purchaser is the health system, and its interest is throughput and coding capture, not the preservation of your clinical reasoning. Lose the money and you lose the practice. Lose the practice and you lose the veto over the tool. Lose the veto and the tool decides what your thinking looks like.
Will there be an Act IV, complete replacement? Not immediately. A physician’s signature is still required to prescribe and to order laboratory tests and imaging, and most people still prefer human contact. That is a regulatory moat, not a permanent one. As AI improves and cost pressure intensifies, a midlevel practitioner with an AI becomes a plausible substitute for a physician with one, and a moat written into statute can be amended.
So is it hopeless? For most, structurally, yes, and I am not writing to most. Of practicing physicians, 82 percent are employed, and the three acts explain why they will stay employed. I am writing to the minority still independent, and to the smaller minority thinking about a jailbreak.
I am an early adopter of AI. With no computer science training, I architected and built a HIPAA-compliant AI-assisted clinical dictation system, and I use it every day. It does not think for me. It handles the boring generative work, the body of the note, and it will pull up a guideline or answer a clinical question when I ask. I remain the author of the HPI and the plan. That rule is written into the application and it is non-negotiable for one reason: I wrote it. Nobody sold it to me and nobody can change it over my objection.
I have since built other tools for my practice and my patients, sharpening old skills and developing new ones. I am more excited and optimistic than ever thanks to AI, not “deskilled.” I am building a cash-based functional endocrinology practice and shedding insurance plans in the traditional one, one at a time.
The exit is narrow and most will not take it. That is not an argument against taking it. It is a description of who is left to make the choice.
Reclaim intellectual sovereignty. Own the practice, and own the tool.
Michael Duben is a board-certified endocrinologist in solo private practice in Fairfield, Connecticut, and is affiliated with Bridgeport Hospital. He runs both Endocrinology of Fairfield County, a traditional insurance-based practice, and Restore Health, a cash-based functional endocrinology practice. He completed the Institute for Functional Medicine’s training program.
He did his internal medicine residency at Mount Sinai and his endocrinology fellowship at Montefiore/Einstein, and was a clinical instructor in the Department of Endocrinology at Albert Einstein College of Medicine. He has practiced since 2003, with clinical interests in thyroid disease and thyroid cancer, osteoporosis and bone metabolism, parathyroid disease, pituitary disorders, and obesity and metabolic disease. He has prescribed glucagon-like peptide-1 (GLP-1) receptor agonists since the first one reached the market in 2005, and has lectured to practicing physicians on diabetes and thyroid management.
He designed and built the HIPAA-compliant AI-assisted clinical documentation system he uses in daily practice, without formal training in computer science, along with a range of other tools for his patients and his office. He shares updates on LinkedIn.
