We can’t build the future of healthcare with ‘never-skilling’

As artificial intelligence becomes increasingly integrated into clinical practice, healthcare providers navigate challenging decision-making dilemmas. Consider a clinician reviewing a diagnostic image flagged by an AI system as high risk. Despite high algorithmic confidence, the output appears inconsistent with the clinical context and long-standing professional intuition. In that instant, the clinician must decide whether to accept the recommendation, override it or seek additional confirmation, with stakes that include both potential patient harm and professional liability. This is no longer a futuristic concept. It is the daily reality of modern healthcare. Public debate often swings between fear and overhype, but neither extreme serves the patient: fear stalls progress, while hype invites uncritical reliance. To navigate this transition safely, AI is best understood not as a replacement for clinicians nor as a panacea, but as a tool the value of which depends on thoughtful integration into care, guided by sound human judgment.At its core, AI refers to computer systems designed to perform tasks that typically require human intelligence. In healthcare, AI supports a wide range of applications, from advanced diagnostic support that identifies patterns in imaging and laboratory data to predictive analytics that enable real-time patient monitoring, as well as administrative and documentation automation. At the health system and population level, AI is increasingly integrated to support risk stratification, resource planning, and the analysis of social and environmental determinants of health. These tools are already embedded across multiple specialties, with imaging disciplines such as radiology, dermatology, pathology, cardiology, and ophthalmology increasingly using AI for image interpretation.Emerging applications include early warning systems for conditions such as sepsis, personalized treatment recommendations, clinical trial optimization, population health surveillance and virtual care delivery. When used appropriately, AI can enhance diagnostic accuracy, reduce delays, optimize workflows and support care in low-resource or remote settings through portable diagnostics and telemedicine, ultimately improving clinical decision-making and interventions. These technologies are increasingly evolving from isolated tools into integrated systems that function as clinical decision support across the care continuum.These benefits are however accompanied by risks that warrant careful attention. AI systems are only as reliable as the data used to train them. If datasets are incomplete or biased, algorithms may underperform for certain populations, reinforcing or worsening existing health inequities. Privacy concerns extend beyond routine clinical data to include genomic information and the potential re-identification of de-identified datasets. Questions of liability also remain unresolved. When harm occurs following an AI-assisted decision, responsibility may be unclear, falling between clinicians, developers, and healthcare institutions. Trust is further complicated by the opaque, “black box” nature of many AI models. Effective AI design therefore depends not only on technical performance but also on transparency, usability, and alignment with patient values and clinical context.AI is most effective when it augments rather than replaces clinical judgment. It excels at pattern recognition, predictive analytics and repetitive tasks, with human review remaining indispensable for ethical reasoning, contextual interpretation and complex decision-making. Communication, empathy and professional accountability continue to define clinical care. New competencies are emerging, including the ability to frame appropriate clinical questions for AI tools and critically interpret their outputs. These skills complement core clinical reasoning, critical thinking and ethical judgment without replacing them.The implications of AI extend deeply into medical education. AI-powered simulations, adaptive learning platforms and virtual training environments allow learners to practice complex clinical scenarios safely while receiving individualized, real-time feedback. At the same time, evidence from high-reliability industries, including healthcare and aviation, suggests that over-reliance on automation may lead to cognitive off-loading, where learners defer reasoning to the tool rather than developing independent judgment. This phenomenon, sometimes referred to as “never-skilling,” is particularly concerning for trainees in the early stages of clinical skills development. Automation bias may reduce engagement with complex problem-solving, and poorly integrated automation can erode vigilance and skill over time.Similar concerns apply to practicing clinicians. Deskilling may occur if physicians rely heavily on AI for routine interpretations, particularly when opportunities for independent assessment are limited. Insufficient rigour in evaluating AI outputs risks propagating errors when models are biased, inadequately calibrated, or applied outside their intended context. Maintaining clinical competence in an AI-enabled environment therefore requires deliberate reinforcement of independent clinical reasoning and ongoing training in the critical appraisal of AI-supported recommendations.Addressing these risks requires action at the system level. Standardized evaluation frameworks allow healthcare organizations to assess whether AI tools meaningfully improve patient outcomes rather than focusing solely on efficiency. Governance structures analogous to professional regulatory frameworks support accountability, transparency and ongoing monitoring. Clinicians benefit from structured AI literacy training that clarifies capabilities, limitations and appropriate use. Human review remains central, with clear guidance on situations in which AI outputs warrant confirmation or escalation. Investment in nationally representative and diverse data infrastructures further supports equity and generalizability in AI development and deployment.AI will continue to shape healthcare, but it does not replace clinicians. Its promise lies in augmenting human decision-making while preserving professional judgment, ethical responsibility, and patient-centred care. Clinicians are responsible for mitigating the risks of overeliance through collaborative intelligence, where AI acts as a scaffold to support human thinking. With deliberate, evidence-based integration, AI can improve outcomes, enhance education, and expand access to care without compromising the human touch that defines medicine.Dr. Henrietta Ezegbe MBBS, MPH is a PGY-4, Public Health and Preventive Medicine Resident at the University of Saskatchewan's College of Medicine.

We can’t build the future of healthcare with ‘never-skilling’

Register for free access or login.

Related Topics