JMIR Publications

ABSTRACT

Generative artificial intelligence (GenAI) holds transformative potential for healthcare delivery, yet its integration into active clinical workflows remains limited. The dominant discourse focuses on risk mitigation — hallucinations, liability, regulatory uncertainty, and algorithmic bias — while a more proximate constraint receives comparatively little attention: the persistent failure to engineer GenAI tools that genuinely empower physicians in daily practice. This perspective article argues that the primary bottleneck impeding GenAI adoption in healthcare is not technical risk but clinical usability, a phenomenon we term the "empowerment gap." We propose a seven-level Empowerment Maturity Framework (EMF) to characterize current deployment readiness and identify the structural dimensions along which progress is needed. Recent evidence indicates that ambient AI documentation has achieved meaningful adoption, with randomized trials demonstrating 21–31% reductions in physician burnout prevalence and up to 30 minutes saved per provider per day. By contrast, higher-order GenAI capabilities — diagnostic reasoning, dynamic treatment planning, and care coordination — remain largely aspirational. Structural barriers including EHR vendor lock-in, heterogeneous clinical environments, and misaligned incentives compound this implementation gap. We conclude that healthcare stakeholders must pivot from a risk-centric to an empowerment-centric paradigm. Engineering GenAI into seamless, reliable, and context-aware clinical workflows is the binding constraint on adoption. Solving usability will generate the real-world use, data, and feedback cycles necessary for trustworthy AI integration — and empowerment engineering must precede, not follow, risk management.