The AI adoption gap in Borderplex healthcare
El Paso and Juárez together form one of the largest binational healthcare markets in North America. University Medical Center, Las Palmas Del Sol Healthcare, Providence Health System in El Paso. The IMSS hospital network, ISSSTE facilities, and growing private hospital capacity in Juárez. These institutions collectively serve millions of patients and employ tens of thousands of clinical and administrative staff.
AI procurement is accelerating across this system. Clinical documentation tools, AI-assisted scheduling, EHR navigation support, revenue cycle automation, imaging-adjacent AI. The tools are being bought.
The adoption numbers do not match the procurement numbers.
HIMSS's 2024 Health IT survey found that while 60% of health system leaders reported AI investments in the past 24 months, only 23% reported meaningful clinical productivity gains. The gap is not technology failure. Most of these tools work as described. The gap is capability: clinical and administrative staff who received no structured preparation for the tools they are being asked to use.
This is not a technology problem. It is a training problem.
Where clinical staff are losing time to poor AI tool adoption
The irony of under-adopted clinical AI tools is that they create more work, not less. A staff member who does not trust an AI documentation suggestion reviews and rewrites it manually — spending more time than if the AI had not been involved. A nurse who cannot navigate an AI-enhanced EHR interface efficiently avoids the AI features and reverts to manual workflows. The organization paid for capability it is not using. The staff member carries friction the tool was supposed to remove.
Three workflows where adoption failure is most costly in clinical environments:
Clinical documentation: AI-assisted ambient documentation and note completion tools can reduce documentation burden by 30–50% for physicians and nursing staff. Organizations where staff have not been trained to trust, review, and correct AI-generated notes see adoption rates below 20% — and those who do use the tools use them less efficiently than a trained user.
Scheduling and patient flow: AI-assisted scheduling reduces wait times and no-show rates when staff understand how to use prediction outputs in their decision process. Untrained staff override AI recommendations consistently without evaluating them, eliminating the tool's contribution.
Revenue cycle and coding: AI-assisted coding and billing tools carry significant ROI potential in healthcare. They also carry compliance risk when used by staff who do not understand the review requirements. Under-trained revenue cycle staff either avoid the tools or use them without appropriate human oversight.
What AI training for healthcare teams must cover
Healthcare AI training is not the same as general AI literacy training. The context is different: the stakes are higher, the compliance requirements are specific, and the users are typically working under time pressure with direct patient impact.
Effective AI training for healthcare staff covers four areas:
Tool-specific workflow integration: not a generic introduction to AI, but a session-by-session walkthrough of how the specific tool integrates into the specific workflow this staff member performs. Documentation AI training for a hospitalist looks different from documentation AI training for an emergency department nurse.
Quality review and override judgment: healthcare AI tools produce outputs that require human review and clinical judgment. Training must specifically address when to accept, when to modify, and when to reject AI output — and how to do each quickly without disrupting patient care workflow.
Documentation and compliance practices: AI-generated content that enters a medical record carries the same compliance requirements as any other clinical documentation. Staff must understand what review is required before acceptance and what constitutes appropriate oversight.
Confidence-building with supervised practice: the most common barrier to clinical AI adoption is not confusion — it is distrust. Staff who are unsure whether the AI output is reliable default to ignoring it. Supervised practice in a non-patient-care context — reviewing AI outputs with a trainer, discussing quality signals, building calibration — is the mechanism that converts skepticism into confident use.
HIPAA and NOM compliance considerations in training design
Training design in healthcare cannot ignore the regulatory environment in which the tools are used.
For El Paso institutions subject to HIPAA: training must address what constitutes PHI, what data the AI tool accesses or processes, and what the staff member's obligations are when AI output involves patient information. Training sessions that use real patient data must comply with minimum necessary standards.
For Juárez institutions subject to Mexico's Ley General de Salud and NOM-024-SSA3 (electronic clinical records): staff must understand how AI-generated documentation integrates with electronic record requirements and what manual verification is required before entry.
Compliance training does not need to be a separate session. Integrating compliance context into the workflow training — "when the AI suggests this, your review before accepting must include these specific checks" — keeps it practical and memorable.
What happens when a healthcare team goes through structured AI training
The most consistent outcome reported by healthcare organizations that implement structured AI training is not speed. It is trust.
Staff who go through role-specific, supervised AI training with real tool scenarios report significantly higher confidence in their ability to evaluate AI output quality. That confidence is the precondition for adoption. Staff who distrust the tool do not use it enough to get better at using it. Staff who trust the tool — because they have been trained to evaluate it — use it more, evaluate it more accurately, and compound their proficiency over time.
A clinical operations team at a large regional health system that implemented structured documentation AI training for all physicians and nursing staff reported a shift from 18% tool adoption to 71% within 90 days of training completion. The training program was eight hours per cohort, role-differentiated, and delivered by trainers with clinical workflow knowledge.
Measuring adoption — what to track
Three adoption metrics that give a real picture of training effectiveness in healthcare:
Active use rate: what percentage of the trained population is using the tool at least weekly? Monthly? This can be pulled from tool usage logs. It tells you whether adoption happened, not whether the training was worth attending.
Override rate: for AI-assisted documentation and coding tools, what percentage of AI outputs are accepted, modified, or rejected? A 100% acceptance rate suggests staff are not reviewing. A 100% rejection rate suggests staff are not trusting. The target is a thoughtful middle — acceptance with selective, appropriate modification.
Documentation quality indicators: are AI-assisted notes meeting the same quality standards as manually authored notes? This requires clinical leadership involvement in the measurement but produces the most meaningful signal.
Building a sustainable internal AI training function
One-time training programs do not match the pace at which clinical AI tools evolve. Organizations that want sustained adoption build an internal capability rather than depending on external program delivery for every new tool.
That internal capability requires: one or more designated AI training leads who maintain current knowledge of the tools in use, a structured onboarding pathway for new staff that includes AI tool training from day one, and a feedback loop between operations and training — so that workflow problems that emerge from AI tool use surface back to the training function.
NDA helps Borderplex healthcare organizations design both the initial training program and the internal capability structure that sustains it.
Learn more about NDA's Corporate AI Readiness programs. | AI Consulting and readiness assessment.