For the first time in the United States, a state has cleared a software system to issue prescriptions without a physician approving each one beforehand. The condition being treated is acne. The implications, though, reach well beyond dermatology.
Utah has authorized New York-based health technology company Nolla Health to run a controlled pilot through which its AI system can select and issue initial acne prescriptions autonomously for eligible adults. The program operates through the Nolla Derm app and is available to Utah residents aged 18 and older. Patients complete a 10 to 15 minute intake questionnaire and a facial scan. The system then assesses their skin and selects a treatment from a predefined list of physician-approved topical options. According to Nolla, the whole process can produce a prescription within minutes.
How the autonomy is introduced — carefully
The rollout is deliberately staged. For the first 100 patients, a physician approves every prescription before it goes out. In the next phase, covering up to 500 patients, the system prescribes autonomously but a physician reviews all cases retrospectively at the end of each day. In the third phase, doctors review a sample of prescriptions once a week. So this is not a sudden hand-off to an algorithm. It is a structured transfer of authority with built-in clinical oversight at each step.
The system also has hard limits. It cannot improvise or choose freely among available drugs. It can only select from a short list of approved topical treatments and predefined pathways. And when a patient falls outside those parameters, the system stops and routes the case to a licensed physician. That constraint matters. It means the AI is not making open-ended clinical judgments — it is executing decisions within a tightly bounded space that clinicians have already defined.
The technology behind the decisions
Nolla says its platform uses multimodal machine-learning, including vision transformers, to analyze facial images alongside structured patient data from the intake form. The company reports its broader systems have been trained on more than 3 million labeled clinical cases. For the acne pilot specifically, Nolla says licensed clinicians agree with the system’s treatment choices in more than 96 percent of real-world cases. Disagreements, the company says, have typically involved minor adjustments like changing a topical medication’s strength rather than selecting a fundamentally different treatment. That figure was provided by Nolla and has not been independently verified.
Why this is different from standard telehealth
Most digital health platforms collect symptoms or assist clinicians, but a human still makes the final call. That’s the conventional model. Nolla’s system is designed to eventually make that call itself, within limits. The distinction is significant both clinically and regulatorily. The pilot was developed with Utah’s Office of Artificial Intelligence Policy under the state’s framework for testing new AI applications, which gives it a formal regulatory structure rather than operating in a gray area.
What this means for patients and clinicians
Patients who receive a prescription through the app can choose home delivery or pharmacy pickup. The app also uses daily skin scans to monitor progress, with treatment plans reviewed monthly based on response. Direct access to a licensed physician is available when needed. The service starts at $4.99 per month under the Utah pilot. Nolla says the app has been downloaded more than 175,000 times across 44 states, though autonomous prescribing is currently only available to eligible adults in Utah.
For clinicians, the pilot raises questions worth tracking closely. Acne is a low-risk condition with well-established treatment protocols, which makes it a reasonable place to test this model. But the infrastructure being built here, algorithmic prescribing with retrospective physician review, could be proposed for other conditions over time. Whether the clinical safeguards scale appropriately will depend heavily on the data that emerges from this pilot. That data, and who gets to see it, will matter enormously.