Will AI replace healthcare professionals? What role does artificial intelligence have in physiotherapy?
Reading time: 10 minutes
News
Introduction
Artificial intelligence (AI) is now emerging as one of the most widely discussed technological revolutions in the healthcare sector. Between fascination and concern, this question keeps coming up: will AI replace healthcare professionals? In France, where the care system is built on a relationship of trust between caregiver and patient, this question raises technical, ethical, and human issues.
Diagnostic algorithms, prescribing support tools, medical imaging analysis software, and virtual assistants are multiplying in private practices, hospitals, and rehabilitation centers. Faced with this technological wave, physical therapists, physicians, osteopaths, and other practitioners are understandably questioning the future of their profession. Yet far from being a threat, artificial intelligence should be viewed as an assistance tool rather than a substitute for human clinical judgment.
This article offers an in-depth reflection on AI's place in the care pathway, the limits of medical automation, and the need to keep the healthcare professional at the center of therapeutic decision-making. We will successively explore the current role of AI in healthcare, the reasons it cannot replace people, and finally how it can become a valuable ally in improving the quality of care.
Artificial intelligence in healthcare: current state of play
A growing presence across all medical fields
Artificial intelligence has gradually become integrated into many aspects of medical practice. In radiology, algorithms detect certain abnormalities on CT scans and MRIs with remarkable accuracy. In oncology, AI helps personalize treatment protocols by analyzing thousands of genomic data points. In physical therapy, software such as Andrew® makes it possible to prescribe personalized exercises and provide remote therapeutic follow-up through telecare and therapeutic education tools.
In France, commitment to AI in healthcare is reflected in massive investment: the digital health acceleration strategy of the France 2030 plan:
Has already mobilized €500 million, of which 50% is dedicated to projects incorporating AI.
More specifically, €90 million has been invested in the development of AI for medical imaging technologies.
In February 2025, the Minister of Health announced an additional investment of €119 million to train around 500,000 caregivers in the use of AI starting in September 2025, or 140,000 healthcare professionals trained per year over 5 years.
These technologies rely on machine learning and the processing of vast databases. They can identify patterns invisible to the human eye and propose recommendations based on thousands of similar cases. This ability to perform large-scale analysis is their main strength, especially in fields that require the processing of large volumes of information.
Promising but limited applications
In France, the Haute Autorité de Santé (HAS) strictly regulates the use of medical devices incorporating AI. These tools must demonstrate their reliability, safety, and real contribution to patient care. While some applications have proven useful, particularly in early disease screening or optimizing care pathways, they remain decision-support tools rather than autonomous decision-making systems.
To ensure safe and ethical adoption of AI, France has put in place a strict regulatory framework. The European Artificial Intelligence Act (AI Act), adopted in May 2024, classifies most medical devices incorporating AI as high-risk.
In 2023, the HAS published a guide to help health professionals choose digital medical devices. In addition, the Digital Health Ethics unit produced best-practice recommendations as early as 2022, and a comprehensive framework for AI ethics in healthcare was published in 2025, structured around five principles: beneficence, non-maleficence, autonomy, justice-equity, and digital frugality.
The National Council of the Order of Physicians regularly reminds us that all therapeutic decisions must remain the sole responsibility of the healthcare professional. AI can suggest, alert, and analyze, but it can never replace the clinical judgment that incorporates dimensions far broader than objective data alone.
Why AI cannot replace healthcare professionals
The irreplaceable human dimension of care
The therapeutic relationship is the very foundation of medical and paramedical practice. A physical therapist does not simply apply protocols: they observe, listen, and adjust their approach based on the patient's reactions, psychological state, and family and work environment. This ability to adapt to context is completely beyond the reach of algorithms, no matter how sophisticated they are.
Empathy, clinical intuition, and the ability to reassure an anxious patient or motivate a discouraged person are deeply human skills. In France, where medicine retains a strong humanistic dimension, these aspects are not secondary: they are therapeutic in their own right. Studies also show that a good caregiver-patient relationship significantly improves treatment adherence and clinical outcomes.
The complexity of clinical reasoning
Diagnosis and therapeutic management are never reduced to a simple equation. An experienced healthcare professional simultaneously mobilizes several levels of reasoning: up-to-date scientific knowledge, experience accumulated through cases encountered, consideration of the patient's individual characteristics, and the integration of contextual factors that are often decisive.
When faced with a patient suffering from chronic low back pain, for example, a physical therapist does not simply identify the location of the pain. They assess overall posture, look for compensations, ask about lifestyle habits, and detect possible aggravating psychosocial factors. This holistic approach goes far beyond the current capabilities of artificial intelligence, which remains limited to analyzing predefined parameters.
Ethical and liability issues
The question of medical responsibility is central to this debate. In France, the healthcare professional incurs personal responsibility for each act of care. They must be able to justify their therapeutic choices and adapt their practice to recommendations while taking into account the specifics of each situation. This individual responsibility is incompatible with delegating decision-making to an algorithmic system.
Moreover, algorithmic bias is a major concern. An AI trained on data that are not representative of population diversity may reproduce and amplify health inequalities. Human judgment remains essential to identify these limitations and ensure equitable care for all patients, regardless of origin, sex, age, or socioeconomic status.
The need for informed consent
The medical code of ethics requires practitioners to obtain the patient's informed consent before any care is provided. This process involves a personalized explanation, adapted to the patient's level of understanding, of the therapeutic options, expected benefits, and potential risks. This dialogue-based approach, essential to respecting patient autonomy, cannot be delegated to a machine.
AI as an assistant: a partnership in the service of the patient
Optimizing clinical time to strengthen the relationship
Paradoxically, artificial intelligence can help humanize care by freeing healthcare professionals from time-consuming, low-value administrative tasks. Administrative management, data entry, scheduling, and literature searches can be partially automated, allowing practitioners to devote more time to listening to and supporting their patients.
Physical therapy tools such as Andrew® illustrate this virtuous complementarity. By making it easier to prescribe personalized exercises and provide remote follow-up, the software allows the physical therapist to maintain the therapeutic connection between sessions without increasing their workload. The practitioner remains the decision-maker for the therapeutic strategy, while the technology ensures the transmission and follow-up of exercises, strengthening the patient's self-management.
There are also physical therapy software solutions such as Markus, which is developing an AI agent that allows physical therapists to be relieved of all time-consuming tasks (writing assessments or letters, transcribing sessions, document analysis, literature searches, decision support, etc.).
Decision support and practice safety
AI excels at identifying warning signals that might escape human attention, especially in situations of fatigue or heavy workload. A prescribing support system can, for example, detect a potentially dangerous drug interaction or alert the user to a contraindication. These features increase patient safety without depriving the professional of decision-making power.
In physical therapy, AI analysis of biomechanical parameters can enrich the clinical assessment by objectifying certain measurements that are difficult to evaluate with the naked eye. Combined with the practitioner's expertise, this additional information refines the diagnosis and allows for a more precise adjustment of the treatment plan.
At Andrew®, we are currently working on training a sovereign AI exercise recommendation model to enable physical therapists to prescribe better and faster, in an evidence-based way (with the latest recommendations). Once again, the goal of this model is to assist the therapist in clinical decision-making, not to make decisions in their place.
Continuing education and knowledge updates
The volume of scientific publications in healthcare is such that no professional can claim to master all up-to-date knowledge in their field of practice. Artificial intelligence can assist practitioners by synthesizing the latest recommendations, identifying studies relevant to a specific clinical case, or proposing scientifically validated protocols.
To support this transformation, France has deployed an ambitious training program. As of the 2024-2025 academic year:
70,000 healthcare and social care professionals have been trained in the uses of AI. The goal is to reach 500,000 trained professionals over 5 years.
At the same time, 15 new master's tracks are being created to train digital health experts, particularly in AI (engineers, computer scientists), as well as 2 master's tracks for legal professionals and 5 master's programs for regulatory affairs officers. These programs are intended to enable professionals to use these tools in their daily work and optimize patient care.
This role of documentary monitoring and support for continuing education strengthens the expertise of healthcare professionals rather than replacing them. It ensures evidence-based practice while leaving the practitioner to adapt these general recommendations to the unique situation of each patient.
Improving treatment adherence
One of the major challenges in rehabilitation remains adherence to home exercise programs. Applications incorporating AI can personalize reminders, adjust exercise difficulty in real time, provide targeted educational content, and maintain patient motivation between sessions. This therapeutic continuity, guided by the professional but supported by technology, significantly improves clinical outcomes.
Andrew® embodies this approach by placing the patient at the center of the rehabilitation process while keeping the practitioner in the role of prescriber and supervisor of the care pathway. The software facilitates therapeutic education, strengthens patient autonomy, and enables regular follow-up without requiring additional appointments.
The French model: regulation and ethics
A protective legal framework
France has chosen to strictly regulate the use of AI in healthcare. The European Medical Device Regulation (MDR), the Data Protection Act, the GDPR, and HAS recommendations form a regulatory arsenal aimed at ensuring patient safety and the clinical relevance of digital tools.
Any medical device incorporating artificial intelligence must obtain CE marking after assessment of its compliance with essential health and safety requirements. This regulation, though burdensome for software publishers, protects professionals and patients from unvalidated or potentially dangerous tools.
To further structure this development, the government announced in February 2025 the launch of two major calls for projects.
The first, called IMPACT IA, aims to test digital medical devices incorporating AI in real-world conditions and to assess their impact on organization and usage in healthcare and social care institutions.
The second will focus on evaluating new methodological approaches in clinical research, including synthetic cohorts and simulated trials, which are particularly useful for rare diseases where the number of patients available for traditional studies is limited.
An observatory of AI use in healthcare has also been created, in connection with the DGOS and ANAP, to identify barriers and drivers and build a structured public support policy.
The position of professional bodies
The National Council of the Order of Physical Therapists, like the other health professional bodies, has clearly stated that AI must remain a tool at the service of the professional and never a substitute. Ethical codes remind us that the care relationship necessarily involves human interaction and that therapeutic responsibility cannot be transferred to an algorithm.
This position reflects a humanistic vision of medicine that places the patient, in their wholeness and uniqueness, at the heart of the care process. It recognizes the potential contribution of technologies while reaffirming the irreplaceability of human clinical judgment.
This vision is aligned with the 2023-2027 digital health roadmap, which structures the deployment of AI around four strategic pillars:
prevention to make everyone an active participant in their own health
care delivery to give professionals more time and improve the patient journey
access to healthcare by integrating AI tools into care pathways
the development of an enabling framework with a sustainable economic model.
A specific roadmap for AI in healthcare was published in 2025, developed in consultation with all stakeholders (patients, caregivers, industry, researchers, citizens) as part of a dedicated task force within the Digital Health Council.
Conclusion
No, artificial intelligence will not replace healthcare professionals.
This conclusion is supported by a triple reality:
the irreducible complexity of care, which integrates human, contextual, and ethical dimensions inaccessible to algorithms
the need for personal accountability and nuanced clinical judgment in every unique situation
the fundamental importance of the therapeutic relationship in the effectiveness of treatment itself.
The future of healthcare is not being shaped by a sterile opposition between human and machine, but by an intelligent partnership in which technology amplifies the professional's capabilities without ever replacing them. AI can analyze, suggest, alert, and optimize, but only the practitioner can decide, adapt, reassure, support, and assume responsibility for their therapeutic choices.
In France, the regulatory and ethical framework ensures that this complementarity develops in respect of the patient and medical ethics. Tools such as Andrew® illustrate this virtuous path: they enrich professional practice, strengthen patient autonomy, and optimize the care pathway, while keeping the practitioner in their central decision-making role.
The challenge for healthcare professionals is therefore not to resist AI, but to train in its prudent use, understand its possibilities and limits, and integrate it as a high-performance assistant in the service of a medicine that will always remain, fundamentally, a human practice serving other humans.
Article Sources
