When AI Writes the Prescription, Who Owns the Trust?

Utah’s Nolla Health pilot for Derma

By SMO Nawed · Healthcare Brand Strategy & Communication

A patient in Utah opens an app, verifies her identity, answers a short health questionnaire, and scans her face from five angles. An algorithm scores her acne, selects a treatment from eight approved topical medications, and sends a prescription to her pharmacy.

No doctor examined her. No doctor wrote that prescription.

Utah is now the first US state to authorise AI to issue initial prescriptions, not refills, through Nolla Health’s Nolla Derm app, cleared under a narrow exception to the licensed-prescriber law. The guardrails are real. Two physicians review the first 100 prescriptions before issue, then weekly retrospective review, then a 10% monthly sample after 500 patients. Malpractice coverage is required.

  • Stage 1 (First 100 Patients): Two licensed Utah-licensed physicians independently review and approve every AI-generated prescription before it goes to a pharmacy.
  • Stage 2 (Patients 101–500): The AI issues prescriptions independently, followed by a weekly retrospective review by physicians.
  • Stage 3 (After 500 Patients): Physicians audit a minimum of 10% of prescriptions each month, plus all cases involving side effects or symptom escalations.

When news of the pilot hit Facebook, public outrage flared in the comments. That immediate backlash reveals a clear divide. The issue isn’t the safety of a phased rollout, but a public convinced a machine is taking over their care. It is a trust problem, built or leaked across all pillars.

When you jump straight from repeat refills to autonomous prescribing, patients panic. Research shows nearly seven in ten consumers are fine letting an algorithm handle their medication renewals, but that comfort drops off a cliff the moment a computer tries to diagnose a new condition. Nolla stepped right over that invisible line without building a bridge first.

To keep people from pulling back, patient communication needs to make the boundary line completely visible before anyone ever opens the app. People need to know exactly where the software stops, who is reviewing the diagnosis behind the scenes, and how fast a real physician will step in if something feels off. Clear rules about legal liability and malpractice coverage aren’t legal fine print here. They are the foundational proof that someone is holding the safety net.

The exact same thing happens internally when leadership forgets to prep their own medical staff. Local dermatologists found out about Utah’s prescribing pilot through news headlines rather than internal memos, instantly turning potential advocates into vocal critics. When medical associations publicly warn that final care decisions belong with a physician, they aren’t just arguing over policy; they are responding to the sudden shock of being left out of the loop.

Giving clinicians complete clarity on their veto power upfront releases that tension. Once doctors know their authority remains intact and patients see the exact limits of the tool, the anxiety surrounding autonomous care vanishes.

Before any external launch of AI triage, scribes, or diagnostics, three questions need answers. Which decisions does the AI own? Which remain unmistakably human? Is the clinician’s veto real?

Culture is not the soft pillar. In Utah it is the difference between a supervised pilot and a profession-wide backlash. Unlike, Aava, the one testing their acceptance for over a year across various medical practitioners.

Healthtech Adoption: autonomy must be earned in phases. 100% review before issue, then 100% review after issue, then sampled review, each unlocked by safety evidence. Narrow indication, human-in-the-loop first, drift monitored, a route back to human control defined. The failure mode to avoid is launch wide, market loud, govern later. In clinical settings, “later” arrives as an adverse event.

Across the sectors: Hospitals should make Utah’s rule their own: no AI in a clinical pathway without an algorithm oversight committee, local validation, a human veto, and board reporting. Clinics win by visibly reserving human time for judgement and exceptions. Specialists need a public position on where AI assistance ends.

Pharmacists become the last human checkpoint, an elevation if they are given protocol and authority. Pharma’s role is truth-based evidence on indications and limits. And healthtech founders should note that the sandbox rewards phased oversight and malpractice cover already built. It punishes those arriving with only a demo.

Utah has not proven AI should prescribe. It has proven autonomy can be tested safely when it is narrow, phased, supervised, insured, and reversible. Sooner than later, prescription generations will be lead by Ai, medicines refilled and tests performed – but for now, let the human share the emotion.

A pilot can pass every clinical benchmark and still fail in public if the trust architecture is missing.

The prescription is now written by an algorithm. Trust still has to be written by leadership.


SMO Nawed’s 4 Pillars (Healthcare Brand Strategy, Patient Communication, Internal Culture Communication, Healthtech Adoption) applied across 6 Healthcare Sectors: Hospitals, Clinics, Private Practice, Pharmacies, Pharma, Healthtech. Explore the framework at smonawed.com.

Sources: WGME CBS 13 News (Facebook, 18 Oct 2026); GovTech (6 Oct 2026); EMARKETER (Oct 2026); The BMJ (Oct 2026).

“Trust isn’t given. It’s engineered.”

Shahriar Nawed

Founder, dnahealth.co

Executive Strategy Briefing

Authored by Shahriar Nawed (SMO Nawed), Healthcare Brand Strategy & Hospital Branding Consultant. Translating clinical excellence into patient trust, OPD footfall, and referral pipeline growth globally.

Trust isn’t given. Let’s engineer it›

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *