4 Whats About Deskilling and AI

During the creation of the AI Guide, deskilling was a topic that arose when discussing the use of AI in clinical practice. Deskilling in its basic form is “the reduction of the skill level required to perform a task.”¹ In health care it is viewed as the “…gradual erosion of independent clinical reasoning skills, together with crucial elements of clinical competence.”¹

What is deskilling?

Deskilling is not a new topic in physiotherapy as it was part of the conversation back when Evidence Informed Practice was just beginning. At the time, some physiotherapy leaders and academics were concerned that “best practice guidelines” and relying on evidence over experience would lead to practitioners losing autonomy and skill required for clinical decision making. However, deskilling related to AI does change the context of the conversation as the influence of AI is being realized. Deskilling is a topic of concern across health care and specifically with physiotherapists using AI scribes and other forms of AI to assist with clinical tasks it is important to raise awareness of this topic. 

Deskilling can be categorized into five categories.²

• Cognitive deskilling is the decline in a clinician’s ability to make good clinical decisions and implement diagnostic reasoning. Things like working through complex client histories, evaluating current research trends and applying them to practice would all fall under this category.
• Semiotic deskilling is the inability to cope with clinical uncertainty. As physiotherapists we are used to dealing in gray areas such as healing timelines, complex diagnosis, and disruptions to treatment plans. However, if we are constantly getting assurance from AI that we are on the right track, there is the risk that we start to ignore uncertainty or choose the easiest pathway when confronted with clinical uncertainty. 
• Technical deskilling is a reduction in the efficiency and effectiveness a clinician may have in performing special tests, goniometry, auscultation or other clinical skills. If you are relying on AI to do these tasks more and more it is inevitable that your skills are going to decline.
• Social deskilling erodes a clinician’s communication skills and empathy. In this profession, would reliance on AI to do your subjective history and intake alter your ability to recognize verbal and non-verbal cues, provide empathy to your client struggling with their injury, or limit your ability to break complex issues down into simple understandable concepts your clients can understand?

There is also the potential for moral deskilling in which clinicians lose their ethical sensitivity and judgment. Combined with multiple reports of AI inducing bias and discrimination within their algorithms this could be quite a concerning trend if it begins to occur.

What are some real-life scenarios in which we could see AI use and risk of deskilling in physiotherapy practice?

Since the majority of research is currently focused elsewhere, we can examine other health-care professions and what they are currently dealing with to get a sense of how it might apply to physiotherapy.

Assessment: AI can be used to take a subjective history, evaluate that history, measure range of motion, force production and other objective measures. It can then analyze huge amounts of available data to compare the subjective and objective measurements to identify patterns and suggest further areas of assessment. 

Diagnosis: After analyzing assessment findings an AI can continue to build on the assessment findings to produce several options for differential diagnosis and prioritize which diagnosis is most likely for your client.

Treatment Planning: An AI can also utilize the assessment and diagnostic information it is provided with, compare that to established best practice guidelines and published research to produce a personalized treatment plan.

Documentation: AI scribes are becoming more and more popular. The scribe can record a conversation, interpret the conversations and interactions between physiotherapists and clients to produce a chart note. There are programs that can now generate a report, referral, or letters to other providers. 

What Are The Concerns?

Based on the current uses that we can find in health care and articles on deskilling, the following are areas of concern for the clinician.³

Over-reliance and trust that what the AI is providing is accurate

It is currently recognized that AI can have issues with reliability, truthfulness, hallucinations, and relative unknowns surrounding the process of how it gets to the answers and suggestions provided. For those using AI scribes, how often are you tempted to not review your AI generated chart note? If you were to use an AI to assist with differential diagnosis or creating treatment plans, how do you know it’s providing these based on peer-reviewed up-to-date research?

Our brains, like our bodies, conform to the adage of “use it or lose it.” So, if you aren’t flexing your clinical decision-making muscles, do you start to lose that cognitive ability to problems solve, draw on research, years of experience, etc.?

Convenience and efficiency

AI has been heralded as a fix-all for many of the issues in health care. Things like professional burn-out, long wait times, inefficiencies in workflows, length of time to process and interpret results, etc. have all been identified as potential issues to target using AI.^2,3 However, research shows that even with positive results in outcomes there are still concerning trends. Studies that showed improvements in diagnostic accuracy and efficiency have noted concerns in the type of errors that were occurring due to automation bias, cognitive offloading, and a failure to participate in deeper, complex reasoning.³

Those that are just beginning to develop their clinical skills can have those skills stunted by the use of AI. Research looking at newer graduates in radiology and oncology are finding those new grads who are using AI to assist with detection of risk factors are showing diminished ability to develop these recognition skills.³ If we apply that to newer physiotherapists, what are the potential draw backs to having an AI assist you with tasks from the start of your career?

Increased risk

The safety of clients is the number one priority of clinicians across practice sites. There have been issues with accuracy of diagnosis, interpretation of images, inherent bias and discrimination, among other things that all can lead to an increase in risk to our clients. Recent studies have shown gastroenterologists and radiologist can significantly regress in their skills to spot abnormalities in diagnostic imaging.^2,3 Other studies have shown clinicians accepting AI results without further investigation or deeper thought resulting in misdiagnosis and failure to schedule appropriate follow-ups.³ These findings have to be considered in the scope of physiotherapy practice and what risks are we bringing to our practice with the integration of AI?

What are potential ways to address those concerns?

Research and opinion have produced a few recommendations for how physiotherapists can avoid deskilling and work with AI to create better outcomes for our clients.^2,4

1. Become AI literate: Educate yourself on the ins and outs of AI, things to look out for, areas of concern, basic nomenclature, how does it work, where does it draw its information from, how can you best implement it into practice?
2. Select the right tool for the right job: Automation should supplement not replace your clinical skills and knowledge base. Be mindful of how you are using AI. Are you using it for redundant tasks or to replace the key components that make you a physiotherapist (clinical reasoning, judgment, client interactions)?
3. Restrict AI usage to tasks that do not involve clinical reasoning. Keep the AI tools focused on data gathering or tasks like an AI scribe. Things that fit into an AI’s strengths such as high-volume/low-risk tasks. Or when sufficient testing of an AI has occurred and is repeatedly able to demonstrate efficacy and safety in real world clinical situations. 
4. Look for AI systems that create a hybrid model of assistance. Tools that would provide information to clinicians in a way that still requires clinical judgments to occur or allows the provider to question, refine, and still flex their clinical reasoning skills. The AI can also work as the critical voice in the discussion, questioning the practitioner on what they may have missed or what other options they should consider.
5. Use a framework to accomplish tasks in which AI contributes but physiotherapists still rely on their knowledge and experience to make clinical judgments. Build in deliberate pauses in the workflow to force you to review, consider, and think about what is occurring. You can also set it up to allow your brain to do the work first, then incorporate AI to assist you. Did you miss something? Is there an assessment finding you forgot to take into consideration or a diagnosis that you didn’t consider?
6. Work with your organization and others in your multidisciplinary teams to prevent deskilling. Organizational changes should preserve the clinician’s role as someone who evaluates information, makes decisions and safeguards their autonomy to work independently from AI. The organization should be looking at hybrid models and frameworks that increase productivity, efficiency, accuracy, and support better health outcomes. Organizations should balance automation with skill rehearsal, practice, and implementation. Many industries have purposefully forced users to shut off the AI and spend a certain number of hours continuing to preform tasks that AI does or assists with so the user retains their skills.

As a physiotherapist you are responsible for your practice and maintaining your competence. The Standards of Practice make it clear that physiotherapists are accountable for any documentation, action, or decision they make. This includes any actions that involve the use of AI systems or tools, and responsibility for any errors made whether due to the use of AI or arising from deskilling due to use of AI in practice. Deskilling is a real and measurable concern in health care that can be addressed through several pathways. Recognizing deskilling is the first step towards evaluating how best you will manage it as you implement AI systems into your clinical practice. There are several ways in which you can proactively combat deskilling and be mindful of how you utilize AI in your practice. 

Actionable Items

1. Be aware of the risks to deskilling in your clinical practice (Cognitive, Semiotic, Technical, Social, and Moral). 
2. Proactively engage in ways to reduce the risks of deskilling.
3. Maintain competence in all aspects of your practice.
4. Continue to engage in deeper and more mindful clinical reasoning focused on diagnostic accuracy.
5. If using an AI Scribe ensure you continue to review and edit your notes prior to signing off. It is unacceptable to assume the note is accurate. 

Sean FitzGerald, PT, Practice Advisor

Page updated: 12/06/2026