Healthcare’s hidden workforce left behind in AI revolution

Despite the increasing presence of artificial intelligence (AI) in medical settings, a significant portion of the healthcare workforce remains overlooked in digital innovation strategies. A new study, titled “AI in Healthcare: Do Not Forget About Allied Healthcare”, published in the journal AI (2025), warns that the current AI revolution in healthcare disproportionately benefits high-profile medical specialists, leaving behind the majority - nurses, physiotherapists, midwives, radiographers, and other allied health professionals (AHPs).

The study, authored by researchers from Philips and Rotterdam University of Applied Sciences, identifies the massive untapped potential of AI in supporting allied healthcare and offers concrete examples across seven core professions. It also addresses the barriers limiting AI adoption in these roles and provides a strategic roadmap for equitable and effective AI integration across the broader healthcare ecosystem.

Why is allied healthcare largely ignored in AI innovation?

While AI is increasingly used to assist surgeons, physicians, and radiologists in decision-making and diagnostics, more than 80% of the global healthcare workforce comprises AHPs, who receive comparatively little attention from developers or investors. These roles involve direct patient care and essential administrative responsibilities, yet many AI applications continue to focus on specialized diagnostics rather than operational efficiency and workload reduction for AHPs.

The authors argue that the low financial incentive for private AI companies, coupled with a lack of engagement with allied health professional organizations, has contributed to this imbalance. As aging populations and chronic disease burdens stretch global health systems, ignoring AHPs risks widening gaps in care delivery and professional burnout, especially after the COVID-19 pandemic exposed the fragility of workforce sustainability in healthcare.

How can AI support allied health professions?

The study details AI applications across seven AHP domains, showing their relevance in clinical care, workflow management, and education.

Nurses can benefit from AI in patient monitoring through wearable sensors, automated medication management, predictive fall detection, and administrative tasks like speech-to-text documentation using generative AI (GenAI). Projects like BETerZO demonstrate how multi-agent GenAI systems integrated with nursing taxonomies (e.g., NANDA) can reduce registration burdens and support clinical decisions.

Physiotherapists are aided by AI-driven rehabilitation tools, including motion sensors that correct posture, wearable robotic exoskeletons, and tele-rehabilitation platforms that reduce the need for in-person visits. AI also supports early disease detection and fatigue monitoring in youth with chronic conditions, as evidenced by the ACT4FATIGUE project.

Midwives can employ AI for risk assessment (e.g., preeclampsia prediction), fetal monitoring, and prenatal ultrasound analysis. Tools developed under the PregnaDigit initiative help high-risk pregnancies through at-home monitoring combined with AI-powered alerts, reducing complications and hospital stays.

Radiographers benefit from AI in image interpretation, patient positioning, radiation dose optimization, and structured report generation. GenAI can even convert handwritten notes into digital radiology reports, improving efficiency and data accessibility.

Occupational therapists use AI for personalized prosthetics, activity tracking through sensors, and cognitive rehabilitation via games and apps. These innovations support patient autonomy and therapy optimization, especially for neurological and cancer patients.

Dietitians utilize AI for personalized meal planning based on medical data, real-time nutritional intake assessment via food image analysis, and long-term health outcome prediction using dietary data.

Speech therapists apply AI to early detection of speech disorders and the creation of affordable, accessible therapy tools. Systems like Voiceitt and Google’s Project Euphonia are examples of AI aiding communication for individuals with speech impairments.

What challenges must be addressed for broader AI adoption?

Despite the potential, several barriers impede AI deployment in allied healthcare. One major concern is bias in datasets, which can lead to inequitable treatment if models are trained on unrepresentative populations. The authors advocate for adopting principles of Responsible AI, emphasizing transparency, fairness, and inclusivity in both algorithm design and implementation.

Privacy and data security are also crucial, given the sensitivity of healthcare records. Techniques like federated learning, pseudonymization, multiparty computation, and synthetic data generation offer technical solutions to protect patient information while enabling AI training across institutions.

Workforce readiness remains another obstacle. Many AHPs lack AI-specific training, leading to resistance or improper use. The paper calls for targeted educational programs, including EU-compliant human oversight training for high-risk AI systems. Courses like Stanford’s “AI in Healthcare” and MIT xPRO’s “AI Fundamentals in Healthcare” are highlighted as foundational but must be complemented with profession-specific instruction.

Furthermore, financial disinterest from private sector developers remains a bottleneck. The study suggests that professional bodies, such as the American Nurses Association, APTA, ACNM, and the World Confederation for Physical Therapy, should partner with tech firms and public agencies to stimulate development and deployment. Public–private partnerships and legislative reforms are recommended to facilitate this process.