When Algorithms Start Prescribing: Who Wins, Who Loses, and What’s at Risk

Utah’s AI-driven prescription refill pilots are designed to improve access and efficiency, but they raise significant questions about adherence, clinical decision-making, healthcare quality, and drug utilization patterns

Why does AI prescribing matter now?

In January 2026, the Utah Department of Commerce signed a one-year pilot agreement with Doctronic for AI-supported prescription renewals in chronic conditions (e.g., diabetes and hypertension) covering approximately 200 medications. Although still evolving, after an initial launch period, physicians will oversee prescriptions retrospectively rather than at each refill decision. Subsequently, in March 2026, Utah entered an agreement with Legion Health for AI-supported behavioral health prescription renewals, covering 15 medications under a similar monitoring model.

These programs are ground-breaking, large-scale efforts to use AI in a prescribing function rather than simply as a decision-support tool. Although Utah has emphasized safety protocols, the pilots have generated significant controversy. Much of the public discussion has focused on patient safety and legal accountability. Yet the implications for payers and pharmaceutical manufacturers are more nuanced and deserve equal attention, particularly if other states follow Utah’s lead.

Why has AI prescribing been developed?

Utah estimates that prescription renewals account for roughly 80% of all medication-related activity, contributing to delays that affect patient experience, provider workload, and downstream quality and cost outcomes. From the state’s perspective, automating routine renewals could free clinician capacity for more complex care while improving continuity for stable patients.

The pilots are expected to be evaluated on their impact on several core measures, including patient adherence and safety, medication utilization, hospitalization rates, and total cost of care. This new use case for AI extends beyond conventional clinical decision support and operational efficiency to care delivery with limited human oversight.

What does it mean to have AI prescribing?

AI has been used to support clinical decision-making, from diagnosis to treatment selection, through tools embedded in the electronic health record or used directly by clinicians. In those settings, however, the clinician has remained the prescriber.

AI-supported refills represent a different model. While the technology does not replace clinicians in initiating therapy, it does shift refill authorization away from the traditional prescriber encounter and toward an autonomous system with limited physician oversight. That distinction matters. A refill is not always an administrative event; in many cases, it is also a clinical checkpoint for reassessing response, tolerability, adherence, the need for dose adjustment or therapy change, and closing care gaps. It offers opportunities for patient engagement and education.

This shift creates broader policy questions as well. The FDA has not yet clearly defined how autonomous prescribing platforms should be regulated, including whether some may qualify as medical devices. At the same time, other states (e.g., Texas and Missouri) are reportedly exploring similar arrangements. If adoption accelerates before standards are established, the United States could quickly develop a patchwork of autonomous AI prescribing models with uneven expectations for governance, safety, accountability, and data integration.

What should payers and pharmaceutical manufacturers consider?

Payer Considerations

For payers, autonomous AI refill programs intersect directly with their existing accountability for member experience, quality, safety, and cost management. Health plans do not simply pay claims; they manage networks, formularies, utilization, care coordination, and risk. As a result, independent AI refill models cannot be evaluated in isolation.

Governance is an immediate concern. Health plan leadership may need to assess howindependent AI refill processes align with existing medical and pharmacy policy, privacy practices, peer review processes, and enterprise risk frameworks. Many plans are still formalizing governance for internal AI use and external vendor-driven autonomous prescribing adds another layer of complexity.

Network alignment is crucial. AI vendors describe escalation pathways involving clinicians, including pharmacists and telehealth providers. Plans will need to decide whether those clinicians fit within current credentialing, contracting, reimbursement, and telehealth coverage policies. 

Drug utilization and safety oversight become even more important when physician oversight is limited. Health plans remain responsible for medication safety in accordance with accreditation quality standards pertinent to safe prescribing and care coordination. As Utah’s approach to clinical data integration evolves, plans may wish to monitor, and potentially help shape, protocols for HIPAA-compliant data exchange to reduce drug interaction risk, promote member safety, preserve continuity of care, and maintain compliance. They should also watch for shifts in medication utilization that could affect budget assumptions, contracting arrangements, and rebate performance.

Finally, payers will need to monitor whether AI-supported renewals improve adherence and access without worsening therapeutic inertia. Fewer prescriber touchpoints may streamline care for stable patients, but they may also delay necessary reassessment, dose escalation, or switching, potentially affecting outcomes, avoidable resource utilization, and patient experience.

Manufacturer Considerations

For manufacturers, the implications to brand strategy may be significant. Autonomous AI prescribing has the potential to influence refill behavior, brand progression, competitive dynamics, and the conditions under which patients move from one therapy to another.

 Treatment inertia is a key concern. AI refill systems may be more efficient at maintaining patients on existing therapy, but less effective at recognizing when treatment should be advanced or optimized. That risk is particularly relevant in therapeutic areas such as behavioral health, where partial response, fluctuating symptoms, tolerability concerns, and functional decline often require active clinical reassessment rather than automatic continuation.

Manufacturers should therefore monitor whether AI-driven renewal pathways change utilization patterns favoring maintenance over optimization. Rates to track include refill persistence, discontinuation, switching, dose escalation, time to therapy advancement, and differences across brands, generics, payer segments, and geographies. A smoother refill process may improve adherence; however, it may also reinforce the status quo in ways that deter uptake of more beneficial or advanced therapies.

Patient support strategies may also need to evolve. Support programs are typically designed for providers, patients, and specialty hubs, and not for autonomous refill tools. Manufacturers may need to adapt educational and support content to also function in an AI-enabled environment. This would include offering clearer triggers for escalation, promptsaround adherence and tolerability, and patient-facing guidance on when to seek live clinical review. Over time, compliant partnerships or integrations may become important if AI tools begin directing patients toward support resources.

Field strategy merits additional attention. Account teams may need to engage plans and providers on how autonomous refill programs impact therapeutic optimization, outcomes metrics, and care pathways. In parallel, manufacturers should invest early in HEOR and real-world evidence generation to assess whether AI-supported renewals improve persistence while also preserving appropriate care advancement. Those manufacturers that demonstrate impact on outcomes, utilization, and total cost of care will be better positioned in both payer discussions and future policy deliberations.

What comes next?

AI prescribing has the potential to reshape how routine care is delivered. Utah’s pilots may improve efficiency and continuity, but they also introduce important considerations regarding safety, oversight, and therapeutic decision-making. Payers and manufacturers should act now to define governance, track outcomes, and build evidence. If autonomous AI prescribing expands, payers and manufacturers must proactively influence its evolution leveraging real-world evidence and operational experience.