I watched a woman in a waiting room tap through an app, snap three photos of her face, and hand her phone back with a shrug. The screen told her a machine would assess her acne and could prescribe treatment — no clinician on the other end. You feel the shift the moment you realize the person in white might not be needed.
I want to walk you through what Utah has quietly started and why it matters. This is not theater; it’s a one-year pilot that puts Nolla Health’s app, Nolla Derm, into the center of a clinical loop for patients with mild-to-moderate acne. It’s limited now, but the steps are precise and fast-moving.
A patient in Salt Lake City opens Nolla Derm and follows a checklist.
The app asks for photos, medical history, and symptoms. Nolla Derm analyzes oil levels, lesion counts, and then generates an “acne severity score.” For the pilot’s first phase, doctors review the initial 100 prescriptions before the AI has independent authority. Phase two stretches to the first 500 prescriptions: the AI writes them and clinicians review those orders after the fact. After that, the system moves to full automation with spot-check reviews by physicians.
The program is limited to straightforward acne cases; severe or complex patients are routed to human doctors. Nolla Health’s Chief Medical Advisor, Dr. Zaid Fadul, emphasizes the algorithm “can’t improvise” and relies on a short, physician-approved list of topical treatments. The app costs $4.99 per month (≈€5).
Can AI prescribe medication without a doctor?
Short answer: in Utah, sometimes. The state’s pilot frameworks allow a phased handoff where physicians initially validate actions, then step back. Regulatory nuance matters here — the law doesn’t hand carte blanche to every AI product. Bloomberg, The Verge, and the Wall Street Journal have followed the rollout closely because it tests where legal authority and clinical judgment intersect.
A primary-care clinic printed a list of renewals and found some were done by an algorithm.
Earlier in the year Utah launched another pilot where, after a physician’s sign-off, an AI can renew certain prescriptions. That program stirred pushback from clinicians worried about safety and oversight; the JAMA Health Forum and the WSJ reported on the debate. You should know the anger isn’t theater: many doctors see risk where payors and startups see efficiency.
Is AI diagnosis safe for patients?
Safety depends on scope and failure modes. For trivial, repeatable tasks — renewing a stable prescription, or matching acne severity to a narrow list of topical options — an AI can reduce friction. But I’m honest with you: when edge cases appear, or when paperwork hides drug interactions, machines can miss context. That’s why Utah’s model keeps humans in sampling and retroactive review early on.
A startup frames this as a cheaper, faster visit; clinicians frame it as a liability problem.
Nolla Health positions Nolla Derm as an accessibility play: faster access, fewer waits, less clinic overhead. Opponents point to unpredictable failures and the erosion of bedside judgment. Geoffrey Hinton has warned broadly about AI displacing skilled roles in medicine; what Utah shows is a careful test of that warning rather than its fulfillment.
Think of the app like a vending machine for prescriptions: you select, pay, and receive a product vetted by rules. Think of the rollout like a sieve separating uncomplicated cases from the ones that still need a human hand.
Should patients trust AI-written prescriptions?
Trust will come in two forms: demonstrable safety records and transparent oversight. Utah’s pilot is testing both — rolling oversight from direct physician validation to retrospective review. If you use such an app, ask how often clinicians audit outcomes and what happens when the algorithm is wrong.
I’ve followed medical pilots and regulatory skirmishes for years, and I’m not here to sell you fear or comfort. You should watch which companies and platforms get early regulatory approval — in this case, Nolla Health and its Nolla Derm app are the ones to watch — and whether pushback from clinician groups slows or reshapes the program.
Utah’s experiment is small in time but large in precedent: one-year pilot, staged autonomy, and an explicit path from human-in-the-loop to human-on-watch. Bloomberg and The Verge documented the launch, and physicians quoted in press pieces are split between caution and curiosity. The cost of entry is modest at $4.99 per month (≈€5), but the cost of error can be higher.
So where does that leave you, the patient or the clinician? If you’re a patient with mild acne who values speed and low cost, this is a tempting option. If you’re a clinician, this is a technical and ethical stress test on the doctor’s role. Who gets to decide what’s simple enough for a machine — and who pays when the machine errs?
Are we ready to let software write prescriptions and call that medicine?