In January 2026, Utah quietly launched something that has no precedent in American medicine: a state-approved pilot program that allows an artificial intelligence system to renew prescriptions autonomously - without a doctor reviewing each case.
Run by the Utah Office of Artificial Intelligence Policy in partnership with a health technology company called Doctronic, the pilot covers nearly 200 medications for common chronic conditions: high blood pressure, diabetes, depression. Right now, human physicians still review the AI's renewal recommendations. The plan is to eventually remove that requirement after the AI has processed 250 cases per drug category.
The Trump administration has signaled its support. Federal officials have described the approach as analogous to self-driving cars - a staged introduction of AI into a function currently performed by licensed humans. The administration is exploring a formal regulatory pathway for what it calls "independent AI physicians."
The Utah Medical Licensing Board has called for the pilot to be halted.
If you run a small independent medical practice, you are watching this from the sidelines - and you should not be.
What's Actually Happening
The Utah pilot is narrow. It covers prescription renewals for stable chronic patients - the clinical equivalent of a refill with no active changes. This isn't AI diagnosing a new condition. It's AI doing the administrative renewal step that most experienced physicians would describe as low-complexity.
But the precedent it sets is not narrow.
If AI can renew a prescription without physician oversight, the question that follows is: what else can it do without physician oversight?
The administration's framing - regulatory pathway for "independent AI physicians" - suggests the ambition goes beyond refills. Stanford researchers are studying AI-generated hospital discharge summaries. The VA is expanding ambient AI scribe technology nationwide in 2026 to reduce documentation burden. The Philips Future Health Index found that nearly two-thirds of clinicians have increased their AI use this year, citing time savings and capacity to see more patients.
The direction is clear. The speed and scope are what's contested.
Why the Medical Board Is Pushing Back
The Utah Medical Licensing Board's concern isn't really about the technology. It's about accountability.
When a physician renews a prescription and something goes wrong - a drug interaction missed, a dosage error, an edge case the algorithm didn't flag - the physician is responsible. Their license is on the line. They carry malpractice liability.
When an AI system does the same and something goes wrong, who is responsible? The software company? The state agency that approved the pilot? The clinic that plugged in the system?
That liability gap is unresolved. There is no established legal framework for "independent AI physician" liability in American tort law. Until there is, every practice that deploys AI for clinical decision-making is operating in uncertain territory.
What This Means for Your Practice Right Now
Short-term: nothing has changed clinically. The Utah pilot does not apply to your practice unless you are in Utah and enrolled in it. You still control all prescription decisions. Federal law and state medical board rules still govern what requires physician oversight.
Medium-term: watch the regulatory pathway closely. If the Trump administration succeeds in creating a streamlined approval process for AI clinical tools - and if that process lowers the bar for what AI can do without physician sign-off - the market will respond fast. Telehealth platforms with lower overhead and AI-assisted care delivery could put meaningful pressure on independent practices competing for the same chronic care patients.
Now: there is genuine low-risk AI ROI available that has nothing to do with clinical decisions.
Healthcare practice management consultants, including analysis from Treetop Growth Strategy's 2026 playbook, consistently point to three proven workflows for small practices:
- Patient communications and recall - AI-automated appointment reminders, recall outreach, and patient portal triage. Measurable ROI, low risk.
- Visit note structuring - Ambient AI scribes that listen to appointments and draft clinical notes for physician review. The VA's nationwide rollout in 2026 is based on evidence from pilots showing significant burnout reduction.
- Eligibility verification and prior auth tracking - Automating the part of revenue cycle management that eats the most staff time with the least clinical judgment required.
The Philips Future Health Index 2026 found that among clinicians already using AI, the most commonly reported benefit was time savings - specifically time freed from administrative tasks that could go toward patient care.
The Compliance Layer You Cannot Skip
Whatever AI you bring into your practice, HIPAA still applies. Any vendor that touches Protected Health Information (PHI) requires a signed Business Associate Agreement (BAA). If a vendor won't sign one, do not use them for anything touching patient data.
This is not a technicality. It's your liability exposure.
The practical rule: use AI for patient communications, documentation assistance, and administrative workflows. Keep it out of direct clinical decision loops until federal liability frameworks catch up to the technology.
The Utah pilot will continue producing data. The administration will continue pushing its regulatory agenda. The medical boards will continue pushing back. That fight will play out over months or years.
What independent practices should not do is wait for the outcome before asking what AI can already do for them - within the compliance boundaries that already exist.
For background on the Utah pilot, see Stanford's analysis and the Philips Future Health Index 2026. The Utah pilot program details are at commerce.utah.gov.
Dr. Renee Carter covers healthcare small business and independent medical practice for The Useful Daily.