Operators, competency, and privileging
The privileging pathway is approved and the three-gate screen is documented for every operator population.
Privilege by application, not by device, and treat expanded operators as a separate program rather than a footnote.
Privileging is essential to a quality program and time-consuming to establish. Beginning it late is the norm and it should not be. The work should start during Phase 3 and complete before production scanning begins, and it should follow the principle that POCUS applications within a clinical specialty are determined by that specialty rather than imposed centrally.
The provider pathways
Three populations require three different routes, and organizations that build only the first route leave most of their eventual scanning workforce unprivileged.
Residency or fellowship trained. Read the ACGME position carefully, because it is more permissive than most program charters assume. Emergency medicine is the clearest case, with an explicit point-of-care ultrasound requirement written into the requirements approved for the 2028 revision, including a minimum dedicated experience. Internal medicine, as of the 2026 requirements, does not mandate POCUS. It requires residents to use or perform point-of-care diagnostic and imaging studies, and names POCUS as one way a program might meet that. Family medicine recommends the experience through a published curriculum guideline rather than requiring it.
The practical consequence is that program completion documentation is strong evidence for emergency medicine graduates and weak evidence almost everywhere else. Build the privileging pathway on demonstrated competency rather than on an assumption about what a specialty required in the year someone trained. For clinicians whose training included it, program completion documentation generally supports privileging directly. Systems commonly add a requirement for clinicians more than five years past training, either documented recent volume or completion of the non-residency pathway.
Non-residency trained. Didactic and hands-on training, typically through a structured CME course, followed by a defined number of proctored studies per application. Volume thresholds vary widely across published programs and the specific number matters less than whether the assessment is competency-based. Attendance at a course is not competency. A portfolio reviewed against defined criteria is.
Externally privileged. Clinicians arriving with privileges from another institution need a verification pathway that is faster than the full non-residency route and more rigorous than automatic reciprocity.
Privilege by application
The unit of privileging is the clinical application, not the device and not POCUS as a category. At least nine specialties have published national-organization recommendations covering cardiac POCUS alone, which means the definition of competent varies by the specialty performing it and by the question being asked. A clinician competent in focused cardiac assessment for pericardial effusion is not thereby competent in volume status assessment or in musculoskeletal imaging.
Application-level privileging is more work to administer and it is the only version that survives an audit.
The expanded operator track
This is where the volume is, and it is the section every existing playbook underweights.
At URMC, transitioning from traditional bladder scanners to ultrasound probes on iPad stands, supported by a dedicated nursing education effort, produced approximately 70,000 bladder examinations from 86 probes in the first six months. For comparison, the same institution reports roughly 22,900 clinical POCUS studies across its entire enterprise in a year, from nearly 1,200 probes. Eighty-six nursing probes generated something on the order of six times the annualized volume of the full physician program, in an application nobody classifies as glamorous.
AI-guided acquisition extends the same logic to more consequential applications. Novice operators using AI guidance have demonstrated diagnostic performance for reduced ejection fraction that approaches expert benchmarks, with sensitivity above 96 percent and specificity above 95 percent in reported series, and with only a small fraction of studies requiring specialist review. The clinical capability is real.
The constraint is not capability. It is that three separate gates must all clear before an expanded operator program is viable, and clearing one or two of them is the most common and most expensive mistake in this category.
The three-gate screen
Gate one, state scope of practice. State boards of nursing, medicine, and allied health determine what each licensed profession may legally perform. AI guidance does not change scope of practice and is not recognized as a substitute for professional training in any state statute. The question is whether a registered nurse, medical assistant, or other professional may perform diagnostic image acquisition under order and supervision in your specific jurisdiction, and the answer varies. For multi-state systems it varies within the same program.
Gate two, institutional credentialing and supervision. Even where state law permits, institutional policy has to define the supervision model with precision. Direct, indirect, general, and remote supervision are different standards with different staffing implications and different liability profiles. The policy has to name which applies, to which applications, performed by which professionals. A program operating on assumed supervision standards is exposed on the specific question an auditor will ask.
Gate three, payer coverage. Permission and payment are separate systems. Payers maintain requirements tied to the credentials of the performing professional, the supervision level, and documentation, and those requirements condition payment independently of whether the service was legal to perform. No CPT code currently describes AI-assisted acquisition performed by a non-credentialed operator, which means these studies are billed, when they are billed at all, using existing ultrasound codes whose descriptors assume a different performer. That gap is a reimbursement problem and an audit exposure at the same time.
Run all three gates in writing before training a single expanded operator. Document the answers. Where a gate does not clear, the honest options are to defer that operator population, to restrict the applications, or to proceed with the program as a cost center rather than a revenue line and say so in the business case.
The market access analysis behind this screen is developed at greater length in the companion research, AI-Guided POCUS Has a Market Access Problem, including a fuller obstacle matrix mapping each barrier to reimbursement and compliance exposure.
In the practice setting
There is no medical staff office, which removes a gate and replaces it with two others.
Payer credentialing rather than privileging. The practice still needs a documented competency standard, and the audience for it is payers and malpractice carriers rather than a credentials committee. Document the training completed, the assessment applied, the pass standard, and the ongoing validation, and keep it retrievable. A practice that cannot produce a competency record on request is exposed in exactly the situations where the record matters most.
A written internal policy that does the work a privileging pathway would. Which clinicians may perform which applications, what training preceded it, who supervises expanded operators if any, and what triggers a review. One page is sufficient. Its absence is not.
The three-gate screen applies unchanged, and gate three frequently resolves differently in an office setting than in a hospital, which is why it should be run against the practice's own payer mix rather than inherited from published guidance.
Two training problems, not one
Most training, and most published competency frameworks, assume one operator who both acquires and interprets, with an over-read where the stakes or the operator's experience require it. AI-guided acquisition splits that role. The person holding the probe may be a nurse, a medical assistant, or a resident early in training, acquiring under guidance while a credentialed clinician interprets.
Three uses follow, and each needs its own competency and quality design. As a teaching aid inside graduate medical education, where guidance scaffolds a learner toward independent skill and the goal is eventually to remove the scaffold. As support for credentialing, where documented guided acquisitions build the portfolio a privileging pathway reviews. And as a way to open image acquisition to non-traditional roles at scale, where the operator will never interpret and does not need to.
What does not change is quality assurance. Guided acquisition raises image quality for a novice. It does not remove the need to sample that novice's studies, feed back on technique, and retrain when acquisition drifts. Build the QA and retraining loop for these operators before the volume arrives, because in the bladder-scanning example above it arrives fast. Treat a rising rate of inadequate captures as a retraining trigger rather than a tolerance to absorb.
Training design
Tailor training to each department and each device model, since interface, controls, and configuration differ enough to matter. Combine clinical and technical content, deliver it through experienced POCUS practitioners alongside IT staff, and include hands-on practice on the department's own devices rather than on a demonstration unit.
Support after training is what determines whether the training holds. Designated super users within each department providing just-in-time assistance, a dedicated help desk with a path to the enterprise imaging team, reference materials, and scheduled refresher sessions. Budget for user-interface testing before go-live and for the changes that testing surfaces.
What to require from your vendor
Training that is competency-validated rather than attendance-based, with a defined assessment and a pass standard. Acquisition performance data exportable into your quality system rather than viewable only in the vendor's portal, because a competency program that depends on a vendor dashboard is a competency program you do not own.
Evidence of performance in the hands of the operator population you actually intend to deploy. A validation study conducted with emergency physicians tells you very little about performance with medical assistants in a primary care clinic. Ask specifically. The answer is frequently that the study does not exist, which is itself decision-relevant.
For AI-guided acquisition, ask what the clearance actually covers, which operators are within the cleared intended use, and what postmarket performance data exists across the populations your organization serves.
Failure mode
Training goes to the enthusiasts. Privileging is completed for the physicians who least needed it. The nursing and allied health staff generating most of the scan volume operate under no documented competency framework at all, and the gap surfaces during an audit or after an adverse event rather than during program design.
Gate criteria
This workstream is done when the privileging pathway is approved by medical staff services and covers all three provider populations, applications are defined by the specialties performing them, the three-gate screen is documented in writing for every expanded operator population in scope, and the competency validation method is defined before training begins rather than after. It runs in parallel with infrastructure and capture, and it takes longer than either, so start it early.