Physician Perspectives
I am an enthusiastic adopter of AI in my daily practice. Without a doubt it is and will change the world for the better. I build and work with these tools to enhance the care I provide. Earlier this spring I presented a pilot study on using artificial intelligence to help draft the operative notes I write after Mohs surgery at the annual ACMS meeting. Despite my love and enthusiasm for AI, a recent development in medical AI concerns me deeply.
This past January, the state of Utah did something no state had done before. It granted a private company permission to let an artificial intelligence system renew patients' prescriptions on its own, without a physician necessarily making the decision.
The company is Doctronic, a New York health-technology startup. Through a Utah program designed to fast-track promising new ideas, sometimes called a "regulatory sandbox," Doctronic became, in its own words, the first AI system legally authorized to practice medicine in the United States. Its platform is approved to renew roughly 191 commonly prescribed medications for chronic conditions, including statins, blood pressure drugs, antidepressants, and blood thinners.
To be clear about where things actually stand: the pilot is still running, and it is not the free-for-all the headlines might suggest. In its current first phase, a licensed physician reviews every single AI-generated renewal before it is sent to a pharmacy. When Utah's own Medical Licensing Board, which has said it learned about the program only after it launched, asked the state to suspend it, the Department of Commerce declined, pointing to exactly that physician review. The friction is real: according to news reports, eleven of the fourteen physicians on that board signed the letter calling for a halt.
But the plan does not end there. A proposed second phase would flip the arrangement. The AI would make the renewal decision, and physicians would audit only a sample of cases after the fact. That single change, from a doctor reviewing every decision to a doctor spot-checking some of them, is the whole ballgame. It is the line between a tool and a replacement.
Here is the distinction I keep coming back to, and it is the same one I live every day in my own practice.
When an AI helps me draft a surgical note, I read it, I correct it, and I sign it. My name is on it. My judgment governs it. The technology made me faster; it did not make the decision. That is augmentation, and I am all for it.
What Utah is piloting, at least in its proposed second phase, is something different. It is AI rendering the medical decision itself, with the physician moved to the margins or removed entirely. That is replacement. And replacement is where I get off the train.
There is a world of difference between AI that helps the doctor and AI that becomes the doctor.
Every physician takes an oath. The words vary, but the heart of it does not: a duty to the patient in front of you, to act in their interest, to first do no harm, to exercise judgment and conscience on their behalf. That oath is not a formality we recite and forget. It is the ethical covenant that makes medicine a profession rather than a transaction.
You cannot delegate an oath to software. An algorithm has no conscience to answer to, no license to lose, and no patient it has looked in the eye. When you remove the physician from the decision, you do not simply remove a step in a workflow. You remove the person who carries that covenant. What remains may look like medicine, but the thing that made it medicine, a human being personally accountable to another human being, is gone.
This is what I mean by keeping the physician in medicine. Not protecting doctors' jobs for their own sake, but keeping the oath, the judgment, and the human accountability at the center of care, where patients have always been entitled to find them.
There is a second reason this matters, and it comes down to a simple question: to whom does the decision-maker answer?
I answer to my patients. I answer to the Minnesota board that licenses me, to the colleagues who would judge my work, and to an ethical tradition older than any company. I answer to the House of Medicine. If I put profit ahead of a patient's need, I can lose the privilege of practicing medicine. That accountability is not optional, and it is not something I can outsource.
A for-profit AI company answers, by law, to its shareholders. Doctronic has raised roughly $65 million from venture investors who expect a return on that investment. I do not say that to vilify anyone. Building a company that way is simply how the venture model works. But it means that when the duty to the patient and the duty to investors point in different directions, the legal and structural pressure runs toward the investors, not the patient. The legal scholars who examined the Utah deal in a Viewpoint published in the Journal of the American Medical Association this past May put it plainly: an AI prescriber faces no equivalent gatekeeper to the training, licensing, and accountability we require of human physicians. A physician's first obligation is built into the profession. A company's first obligation is built into its corporate structure. Those are not the same thing, and patients deserve to know the difference.
None of this means the problem Doctronic is chasing is imaginary. It is very real. The country faces a serious and worsening shortage of primary care physicians. Refilling a routine, stable medication can mean a wait, a phone call, or a portal message that goes unanswered for days, and some patients simply give up. One widely cited analysis estimated that non-adherence and non-optimized medication regimens cost the United States more than $500 billion a year. If a well-built tool can get a patient their stable blood pressure refill at eleven at night without a needless delay, that is a genuine good, and I will be the first to say so.
The early Utah numbers are promising. According to data reported by the health-news outlet STAT, the AI renewed prescriptions on its own about 72 percent of the time, escalating the rest to a human, and when it did grant a renewal, the reviewing physicians agreed with it roughly 91 percent of the time. That is genuinely impressive. It is also worth sitting with the other side of that figure: in about one case out of every eleven, a physician would have done something different. The company's own marketing cites a 99.2 percent agreement figure from an internal study, but as the JAMA authors point out, that study was conducted by company-affiliated authors and has not been peer reviewed. The real-world 91 percent is a more honest place to begin the conversation, and that remaining gap is precisely where a patient can be harmed.
So my objection is not to the goal. It is to removing the physician in order to reach it.
There is one more question that no one has fully answered: when an autonomous system makes a mistake, who is accountable? The scholars writing in JAMA argue that an AI which prescribes without a physician most likely should be regulated by the U.S. Food and Drug Administration as a medical device, and that "black box" systems are difficult to hold to the same standard of care we apply to doctors. These are unsettled legal questions. I am a doctor, not a lawyer, so I will leave the statutes to the people who write them. But as a patient, you should want those questions answered before the technology becomes too entrenched to regulate, not after.
I will keep using artificial intelligence. I will keep looking for ways it can streamline clinical processes to enhance my productivity, because every minute it saves me is a minute I can give back to the person in the exam room. That is the real promise of this technology, and this is the way to solve the physician shortage. Enhance physician productivity and workflow with intelligent tools. Do not use the tools to replace the physician. Keep the thinking machines and their corporate owners out of medical decision-making so we can keep the physician in medicine. The oath is worth protecting, and so are the patients it was written for.
Dr. Kevin Christensen
The opinions in this post are my own and are offered as commentary on a developing public policy question, not as medical or legal advice. Details about the Utah pilot reflect news reporting and public records available as of June 2026 and may change as the program evolves.
Looking for a physician-led dermatology practice in Rochester, MN?
Schedule Your Appointment Call (507) 206-3211