Physical Medicine and Rehabilitation Physicians

29-1229.04
Median wage $265,930/yr342,720 employed (US)Rank #802 of 923 scored · top 87% by substitution

Diagnose and treat disorders requiring physiotherapy to provide physical, mental, and occupational rehabilitation.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution15
Exposure16
Augmentation54

Substitution — the headline: capability discounted by cost, barriers and adoption.

Exposure — technical capability alone, regardless of whether anyone deploys it.

Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.

Tasks on the substitution scale

15 rated tasks, binned by substitution score.

Position among all scored occupations

Distribution of 923 occupation scores; the marker is this occupation.

Tasks with substitution ≥ 70

0%

Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.

Why this score

The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.

Task automatabilityw 35%16

panel mean rating 1.6/5 → substitution pressure 16/100

Technical feasibility todayw 20%17

panel mean rating 1.7/5 → substitution pressure 17/100

Cost vs. human wagew 15%12

panel mean rating 1.5/5 → substitution pressure 12/100

Adoption barriersw 20%inverted — strong barriers lower the score9

panel mean rating 4.7/5 (barrier strength) → substitution pressure 9/100

Sector adoption velocityw 10%20

panel mean rating 1.8/5 → substitution pressure 20/100

Task breakdown (15 tasks)

Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.

Document examination results, treatment plans, and patients' outcomes.

35

CI 25–45 · exposure 38 · augmentation 75 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI documentation tools in rehabilitation medicine remains in early pilots; most physicians still hand-document or use EHR templates with minimal AI augmentation, reflecting organizational conservatism, regulatory caution, and integration complexity in legacy health systems.
Sector adoption velocityclaude-sonnet-53/5Healthcare has moderate AI documentation adoption via ambient scribe pilots and growing production use, but overall sector digitization and regulatory caution keep it behind fast-moving sectors like finance or tech.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by transcribing examination findings, suggesting structured headings, and populating routine sections of notes, materially reducing clerical burden; however, the clinical synthesis of assessment and treatment planning remains physician-driven, so augmentation is meaningful but not transformative.
Augmentation potentialclaude-sonnet-55/5AI-powered ambient documentation and note-generation tools are widely reported to significantly reduce physician documentation burden and time spent on notes while the physician remains the responsible reviewer.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can extract structured data from examination findings and generate template-based documentation drafts, the task requires clinical judgment to synthesize complex patient history, physical exam nuances, and functional status into a coherent, medically sound treatment plan—something current systems cannot reliably do end-to-end without substantial physician oversight and revision.
Task automatabilityclaude-sonnet-53/5AI can draft clinical documentation from dictation or notes (ambient scribes), but final review, accuracy verification, and clinical judgment calls remain necessary, limiting full end-to-end automation.4
Adoption barriersclaude-haiku-4-5-202510014/5Medical documentation is legally and regulatory binding (HIPAA, state board oversight, malpractice liability); the physician must attest to the accuracy and clinical appropriateness of the record, making autonomous or minimally supervised AI generation a liability and compliance risk that deters deployment.
Adoption barriersclaude-sonnet-54/5Medical documentation must be attributable to and legally signed off by the licensed physician, with liability, HIPAA, and EHR compliance requirements creating substantial barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI-assisted documentation (transcription + template filling) can reduce clerical time, but the physician still must review, correct, and clinically validate the output, limiting cost savings to roughly 20–40% of the task; full physician wage loading remains high due to required physician attention.
Cost vs. human wageclaude-sonnet-53/5AI scribe subscriptions are cheaper than added clerical/transcription staff time, but licensing, integration, and required physician oversight keep costs closer to comparable rather than order-of-magnitude cheaper.
Technical feasibility todayclaude-haiku-4-5-202510012/5Medical transcription and note-generation products exist, but they typically operate in narrowly constrained settings (e.g., orthopedic intake forms) and still require physician validation and correction; no deployed system reliably documents examination results and generates treatment plans autonomously in production rehabilitation medicine workflows.
Technical feasibility todayclaude-sonnet-53/5Ambient clinical documentation tools (e.g., Nuance DAX, Abridge) are deployed in real health systems today but still require physician review/edit and have variable accuracy across specialties.

Assess characteristics of patients' pain, such as intensity, location, or duration, using standardized clinical measures.

25

CI 25–25 · exposure 25 · augmentation 63 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While healthcare is digitizing, adoption of AI for direct clinical assessment tasks lags significantly. Most pain assessment tools deployed are decision-support aids used alongside physician evaluation, not replacements. Real production displacement in this clinical specialty remains minimal.
Sector adoption velocityclaude-sonnet-52/5Healthcare, particularly physical exam-based specialties, has been slower to adopt AI for clinical assessment tasks compared to purely administrative or information-based sectors, though digital pain questionnaires are increasingly used.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by summarizing patient-reported pain data, flagging patterns in standardized measures, and surfacing relevant clinical guidelines, thereby augmenting the physician's efficiency in synthesizing information. However, the core task—direct clinical judgment about pain characteristics—remains substantially human-centered.
Augmentation potentialclaude-sonnet-54/5AI-assisted questionnaires, natural language processing of patient-reported symptoms, and standardized digital pain scales can meaningfully speed up data collection and documentation, letting physicians focus on clinical correlation and judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist in interpreting pain scales and documented symptoms, the task requires direct patient interaction, observation of physical responses, and clinical judgment to assess pain characteristics reliably. Current AI cannot fully replace the physician's role in conducting and synthesizing these assessments end-to-end with equal quality.
Task automatabilityclaude-sonnet-52/5AI can help structure pain assessment data and apply standardized scales (e.g., VAS, McGill Pain Questionnaire) but the actual patient interview, physical examination, and clinical judgment about pain characteristics require direct human interaction and interpretation.4Full end-to-end automation with equal quality is not achievable today.5
Adoption barriersclaude-haiku-4-5-202510014/5Regulatory and licensing requirements mandate that a licensed physician perform or directly oversee clinical assessment of pain, especially in rehabilitation contexts where diagnosis and treatment planning depend on the assessment. Liability exposure and the need for human judgment in individualizing care create strong legal and organizational barriers.
Adoption barriersclaude-sonnet-54/5Clinical pain assessment as part of diagnosis is typically restricted to licensed physicians, especially in physical medicine and rehabilitation where physical exam findings must correlate with reported pain; liability and licensing requirements are significant barriers.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI systems for pain assessment require integration with clinical workflows, oversight, and validation by the physician. The total cost of a decision-support system is not yet substantially lower than the incremental cost of the physician performing the assessment directly.
Cost vs. human wageclaude-sonnet-52/5Digital pain-scale intake tools are cheap, but since a licensed physician must still perform the clinical correlation and physical exam, the AI only offsets a small portion of cost, making overall savings limited.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI products can extract pain data from patient-reported outcomes and notes, but no deployed system reliably performs the full clinical assessment (visual inspection, palpation response, functional testing) that defines this task in clinical practice. Most implementations remain narrow and require human validation.
Technical feasibility todayclaude-sonnet-52/5Some digital intake tools and symptom-tracking apps exist that collect patient-reported pain data using standardized measures, but no deployed product independently conducts the full clinical pain assessment with physical exam correlation reliably in production.

Develop comprehensive plans for immediate and long-term rehabilitation, including therapeutic exercise, speech and occupational therapy, counseling, cognitive retraining, patient, family or caregiver education, or community reintegration.

23

CI 20–25 · exposure 25 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption in rehabilitation medicine lags other clinical domains; most practices still rely on physician-led planning with templates and consultation models rather than AI-assisted planning. Even where digital tools exist, they serve as references rather than autonomous plan generators.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially physician-level clinical decision-making, has been slow to adopt autonomous AI tools due to regulatory, liability, and workflow integration challenges, with pilots more common than production use.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully assist by surfacing evidence-based therapy options, suggesting standard protocol components for a diagnosis, or organizing therapy sequences, which helps a physician work more efficiently. However, the augmentation is modest because physician judgment and synthesis remain essential for comprehensive personalized planning.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist by drafting plan templates, summarizing patient history, suggesting evidence-based interventions, and generating patient education materials, substantially speeding the physician's planning process.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist in generating template rehabilitation plans and evidence-based therapeutic recommendations, the task requires integration of complex patient assessment, clinical judgment about sequencing therapies, and adaptation to individual patient presentation and goals. No current system can reliably develop a full comprehensive plan end-to-end that meets the 50% time-saving bar at equal clinical quality.
Task automatabilityclaude-sonnet-52/5Developing a comprehensive rehabilitation plan requires integrating clinical exam findings, prognosis, patient goals, and multidisciplinary input into individualized decisions that current AI cannot reliably perform end-to-end.dev AI can draft components but cannot autonomously synthesize the full clinical judgment required.
Adoption barriersclaude-haiku-4-5-202510014/5Rehabilitation planning is a core scope-of-practice task for licensed physicians; liability for poor outcomes rests with the responsible physician who must personally sign off on the plan. Regulatory and professional standards require physician judgment on complex multi-modal therapy selection, making substitution legally and ethically problematic.
Adoption barriersclaude-sonnet-55/5This is a licensed medical act requiring physician diagnosis, judgment, and legal accountability; regulations and malpractice liability firmly require a physician to create and sign the plan.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of AI system inference and integration for plan drafting, plus required physician review and revision of the output, remains comparable to or exceeds the time cost of a physician developing a plan with standard clinical templates and guidelines. The oversight burden is substantial.
Cost vs. human wageclaude-sonnet-52/5AI could cheaply draft template plans, but the human physician's oversight, exam, and liability-bearing sign-off dominate the cost, making all-in cost comparable to or only modestly cheaper than the physician's time.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI tools exist for suggesting rehabilitation protocols and therapy components based on diagnosis, but no deployed product reliably produces clinically adequate comprehensive rehabilitation plans without substantial physician revision and oversight. Current systems lack the contextual understanding of patient-specific factors (severity, comorbidities, social support, prognosis) needed for safe plan development.
Technical feasibility todayclaude-sonnet-52/5Some clinical decision-support tools exist for suggesting therapy protocols, but no deployed product independently generates full rehabilitation plans used without physician authorship in production settings.

Prescribe physical therapy to relax the muscles and improve strength.

23

CI 20–25 · exposure 25 · augmentation 63 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Despite digital health adoption in clinical settings, AI prescription of physical therapy remains rare in production; most adoption is limited to decision-support tools that augment rather than replace physician judgment. Healthcare's conservative posture on liability and licensing slows true automation of prescribing tasks.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially physical medicine, adopts AI tools cautiously due to regulatory, liability, and clinical validation requirements, with pilots more common than production deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by synthesizing clinical evidence, suggesting evidence-based protocols, and flagging contraindications, improving a physician's efficiency in drafting therapy plans. However, the core clinical reasoning—integrating patient history, examination findings, and risk assessment—still requires the human physician's direct involvement.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist physicians by suggesting evidence-based exercise protocols, summarizing patient history, and flagging relevant guidelines, improving efficiency while the physician retains decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can retrieve clinical evidence and draft therapy recommendations, prescribing requires individualized assessment of patient pathology, contraindications, and functional goals that demand licensed clinical judgment. AI cannot reliably perform the full diagnostic-to-prescription workflow end-to-end at the quality threshold needed for medical safety.
Task automatabilityclaude-sonnet-52/5AI can help draft physical therapy plans or suggest evidence-based protocols, but the clinical decision to prescribe therapy requires physical examination, diagnosis integration, and legal accountability that current AI cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Strong regulatory and professional barriers exist: only licensed physicians can legally prescribe physical therapy in most jurisdictions, and malpractice liability attaches to the prescription itself. Legal authority cannot be fully delegated to AI regardless of accuracy.
Adoption barriersclaude-sonnet-55/5Prescribing medical treatment is a licensed physician activity with strict legal and regulatory requirements; a licensed physician must authorize and sign off on this task.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI inference and integration costs are low, but the task requires human physician oversight and liability, so the total cost per prescription remains dominated by the physician's loaded wage. No meaningful cost advantage emerges when legal and professional responsibility remain with the human.
Cost vs. human wageclaude-sonnet-52/5While AI-assisted drafting is cheap, the overall cost includes mandatory physician review, examination, and liability oversight, keeping costs comparable to physician-led care rather than dramatically cheaper.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably prescribes physical therapy independently; AI tools exist for literature retrieval and decision support but all production systems require physician review and final sign-off. The task involves complex patient-specific risk stratification that current systems cannot handle autonomously at acceptable error rates.
Technical feasibility todayclaude-sonnet-52/5Clinical decision-support tools exist that suggest treatment options, but no deployed product autonomously prescribes physical therapy regimens in production without physician oversight.

Prescribe therapy services, such as electrotherapy, ultrasonography, heat or cold therapy, hydrotherapy, debridement, short-wave or microwave diathermy, and infrared or ultraviolet radiation, to enhance rehabilitation.

18

CI 16–20 · exposure 25 · augmentation 50 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Rehabilitation medicine lags in AI adoption relative to radiology or pathology. Pilot decision-support tools exist but production deployment of autonomous or semi-autonomous therapy prescription remains minimal; organizational and clinical culture still emphasize physician-led individualized assessment.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially specialized physical medicine practice, has historically been slow to adopt AI for clinical decision-making due to regulatory and liability constraints, though some diagnostic support pilots exist.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully assist by summarizing clinical evidence, flagging contraindications, and proposing therapy options based on diagnosis and patient history, thereby streamlining the physician's deliberation. However, the task's requirement for real-time patient response evaluation limits augmentation impact.
Augmentation potentialclaude-sonnet-53/5AI can help synthesize evidence, suggest modality options based on diagnosis, and streamline documentation, offering moderate assistance while the physician retains prescribing authority.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist in identifying candidate therapies based on clinical guidelines and imaging, the task critically depends on real-time assessment of patient response, contraindications, and individualized clinical judgment. Current systems cannot reliably perform the full diagnostic-to-prescription workflow end-to-end with 50% time savings and equal safety.
Task automatabilityclaude-sonnet-52/5Prescribing therapy requires clinical judgment integrating exam findings, diagnosis, and patient-specific factors; AI can suggest options but cannot independently generate a defensible, personalized prescription end-to-end today.
Adoption barriersclaude-haiku-4-5-202510015/5Prescribing therapy services is a licensed medical act; regulatory frameworks (FDA, state medical boards) require a licensed physician to take professional responsibility for the prescription. Liability for adverse effects and patient safety requirements create hard legal barriers to autonomous AI substitution.
Adoption barriersclaude-sonnet-55/5Prescribing medical therapy is a licensed physician act with direct liability and regulatory requirements; a qualified physician must legally authorize the treatment plan.
Cost vs. human wageclaude-haiku-4-5-202510011/5The oversight, liability review, and clinical validation costs for AI-assisted prescription in rehabilitation exceed the physician time saved, since the physician must still assess contraindications, monitor outcomes, and take legal responsibility for therapy choice.
Cost vs. human wageclaude-sonnet-52/5AI inference is cheap, but the physician's licensed judgment and liability exposure remain necessary, so overall cost savings versus physician time are modest once oversight is included.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed clinical product reliably prescribes complex rehabilitation therapy regimens independently. Diagnostic decision-support tools exist but require physician interpretation and sign-off; they do not constitute autonomous prescription at scale in production environments.
Technical feasibility todayclaude-sonnet-52/5Clinical decision-support tools exist but are not deployed to autonomously prescribe modality-specific rehab therapies in production; use is limited to reference/suggestion aids reviewed by physicians.

Monitor effectiveness of pain management interventions, such as medication or spinal injections.

14

CI 3–25 · exposure 13 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Pain management monitoring is embedded in clinical workflows that rely on physician judgment and patient contact. Adoption of AI-assisted tools in this space remains slow, mostly in early-stage pilots in academic centers, not mature production deployment across healthcare systems.
Sector adoption velocityclaude-sonnet-52/5Healthcare clinical decision-making tasks, especially those involving controlled substances and interventional pain management, show slow AI adoption due to regulatory and liability constraints.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by aggregating pain scores, detecting trends, flagging outliers, and summarizing patient-reported outcomes, improving the physician's ability to review effectiveness quickly. However, the core interpretive and decision-making work remains the physician's responsibility.
Augmentation potentialclaude-sonnet-53/5AI can assist by aggregating patient-reported outcome data, flagging trends in pain scores, or summarizing EHR data, but the clinical interpretation and decision remain human-led.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can process clinical data and patient-reported pain scores to flag trends, the task requires nuanced clinical judgment about causation, symptom progression, and intervention adjustment that demands human expertise. Meaningful automation would require AI to independently modify treatment plans, which is not feasible today.
Task automatabilityclaude-sonnet-51/5Monitoring pain intervention effectiveness requires physical examination, patient interview nuance, and clinical judgment integrating subjective and objective data that AI cannot perform end-to-end today.
Adoption barriersclaude-haiku-4-5-202510014/5Regulatory barriers are high: physicians must legally interpret clinical data, assess causation, and authorize treatment changes. Liability asymmetry is steep—AI missing or misinterpreting pain management failure could lead to serious patient harm. Patient safety and professional duty create strong legal and organizational friction against full automation.
Adoption barriersclaude-sonnet-55/5This is a licensed medical task requiring physician judgment, controlled substance management, and legal liability for treatment decisions, representing a hard regulatory barrier.
Cost vs. human wageclaude-haiku-4-5-202510012/5Implementation requires specialized EHR integration, clinician oversight, and validation of AI recommendations. The all-in cost per patient monitoring episode likely remains comparable to or exceeds the time a physician spends reviewing data and making clinical decisions.
Cost vs. human wageclaude-sonnet-51/5Human physician evaluation is required for safety and billing purposes; AI cannot substitute for the actual monitoring visit, so no cost savings accrue from replacement.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed products can track and visualize pain metrics from EHRs and wearables, but no reliable production system exists that comprehensively monitors intervention effectiveness and recommends changes without significant physician oversight. Clinical decision support tools exist but are narrow and require substantial human validation.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously monitors and clinically judges effectiveness of pain interventions like spinal injections; this remains a physician-performed clinical task.

Provide inpatient or outpatient medical management of neuromuscular disorders, musculoskeletal trauma, acute and chronic pain, deformity or amputation, cardiac or pulmonary disease, or other disabling conditions.

11

CI 3–20 · exposure 13 · augmentation 63 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Despite heavy digitization in healthcare, adoption of AI for autonomous clinical decision-making in PM&R remains in pilot/early stages. Regulatory caution, liability concerns, and the complexity of managing multiple disabling conditions limit rapid displacement; most adoption is assistive rather than substitutive.
Sector adoption velocityclaude-sonnet-52/5Healthcare delivery, especially hands-on physiatry, has historically lagged in AI adoption for core clinical management despite growing use of AI in documentation and diagnostics support.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can significantly augment physician productivity through automated literature synthesis, imaging analysis, treatment protocol suggestions, documentation support, and outcome tracking. These tools help physicians work more efficiently while retaining clinical judgment, transforming the workflow even if the physician remains central to care decisions.
Augmentation potentialclaude-sonnet-53/5AI can assist with clinical documentation, treatment plan suggestions, literature synthesis, and monitoring data analysis, improving physician efficiency without replacing core clinical judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with diagnosis support, treatment planning, and documentation, the task fundamentally requires real-time clinical judgment, physical examination, patient interaction, and adjustment of complex management plans across multiple organ systems. Current AI systems cannot safely manage the full scope without continuous human oversight, and the 50% time-saving threshold is not met for end-to-end task completion.
Task automatabilityclaude-sonnet-51/5This is comprehensive, hands-on medical management requiring physical examination, procedural interventions, and ongoing clinical decision-making across complex conditions; no AI system can perform this end-to-end today.
Adoption barriersclaude-haiku-4-5-202510015/5Medical licensure laws require a licensed physician to directly manage patient care, make diagnostic decisions, and authorize treatments. Liability, malpractice risk, and regulatory requirements (state medical boards, CMS) create hard barriers: a physician must legally perform or sign off on all clinical decisions for these conditions.
Adoption barriersclaude-sonnet-55/5Medical licensure, malpractice liability, and legal requirements for physician-directed care create hard barriers preventing AI from performing this task independently.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of deploying AI for medical decision-support, verification systems, and integration into clinical workflows, combined with required physician oversight and liability coverage, remains comparable to or exceeds the cost of direct physician management, especially given the need for real-time human validation.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the physician's role here, so there is no meaningful cost comparison for the task as a whole—human physician cost is unavoidable.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed clinical product reliably performs inpatient/outpatient medical management of complex neuromuscular or musculoskeletal conditions independently. AI tools exist for narrow subtasks (imaging analysis, documentation), but production systems do not demonstrate reliable end-to-end management of acute and chronic disabling conditions across diverse patient populations.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs inpatient/outpatient medical management of these conditions; AI is used only for narrow decision-support or documentation tasks within this broader task.

Conduct physical tests, such as functional capacity evaluations, to determine injured workers' capabilities to perform the physical demands of their jobs.

11

CI 0–23 · exposure 13 · augmentation 50 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Physical medicine and rehabilitation is a clinical specialty with slower digitization and AI adoption compared to information-based professions. The physical, hands-on nature of the work and regulatory/liability constraints mean adoption of AI for conducting these evaluations remains minimal in practice.
Sector adoption velocityclaude-sonnet-51/5Physical rehabilitation medicine involves hands-on clinical exams in a highly regulated, low-digitization physical care setting with minimal AI displacement of the core exam task.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by analyzing video recordings of functional tests, suggesting quantitative metrics, or flagging movement patterns for physician review, supporting documentation and result interpretation. However, the core assessment still requires the physician's direct observation and clinical judgment, limiting the transformative potential of AI assistance here.
Augmentation potentialclaude-sonnet-53/5AI can assist with documentation, scoring standardization, and report generation from FCE data, but does not touch the physical testing itself.
Task automatabilityclaude-haiku-4-5-202510012/5Physical tests require direct hands-on assessment and real-time observation of patient movement, which current AI cannot perform end-to-end. While AI can analyze recorded video or assist in interpretation of test results, the core task of conducting functional capacity evaluations demands in-person physical interaction and clinical judgment that remains outside current automation capabilities.
Task automatabilityclaude-sonnet-51/5Functional capacity evaluations require hands-on physical testing, direct observation of movement, strength, and pain response, which current AI cannot perform without a human physically administering and interpreting the exam.
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: workers' compensation law typically requires a licensed physician to conduct or sign off on functional capacity evaluations, direct patient contact is legally and clinically necessary, and high liability exposure for incorrect capability determinations creates strong organizational and regulatory friction against full automation.
Adoption barriersclaude-sonnet-55/5This is a licensed medical/legal evaluation used for workers' compensation and disability determinations, requiring a credentialed physician's direct examination and legal sign-off.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of deploying AI systems for physical assessment (cameras, sensors, software infrastructure, oversight) combined with the need for human supervision and validation likely exceeds or approaches the cost of a physician's time, especially given liability concerns and the high stakes of disability determinations.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the physical examiner role at all, so there is no viable AI cost comparison—human labor is required regardless of cost.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed AI system conducts functional capacity evaluations independently. While computer vision systems can analyze movement from video and machine learning can support result interpretation, these represent narrow components rather than reliable end-to-end performance of the full clinical assessment task in production settings.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product conducts or physically administers functional capacity evaluations; this remains an entirely clinician-performed physical task.

Perform electrodiagnosis, including electromyography, nerve conduction studies, or somatosensory evoked potentials of neuromuscular disorders or damage.

11

CI 0–21 · exposure 13 · augmentation 38 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI in electrodiagnostic interpretation is slow and scattered, limited mostly to academic medical centers and research partnerships. Most clinical practices rely on traditional physician-performed or technician-assisted workflows; production deployment of autonomous or semi-autonomous systems is rare and nascent.
Sector adoption velocityclaude-sonnet-51/5Clinical neuromuscular diagnostics is a highly manual, physically-mediated specialty with minimal AI deployment in the procedural workflow itself.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist physicians in pattern recognition on recorded waveforms, flagging abnormalities and suggesting diagnostic categories, thereby accelerating report generation and reducing manual review time. However, the real-time procedural demands and need for clinical judgment limit the scope of augmentation to post-acquisition analysis rather than the full task.
Augmentation potentialclaude-sonnet-52/5AI can assist with waveform pattern analysis or data logging support, but the core hands-on procedure and diagnostic judgment remain largely unaided by current tools.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist in interpreting electrophysiological waveforms and data, the full task requires real-time technical execution (electrode placement, stimulus calibration, patient positioning), clinical judgment about abnormal findings, and adaptation based on patient response. Current AI systems cannot autonomously perform the procedural components that comprise a substantial portion of the work.
Task automatabilityclaude-sonnet-51/5This requires hands-on electrode placement, needle EMG insertion, real-time patient interaction, and clinical judgment during the procedure itself; no AI system today can perform the physical exam or acquisition process.
Adoption barriersclaude-haiku-4-5-202510014/5Strong regulatory and liability barriers exist: electrodiagnosis is typically licensed-physician work (or certified technician under physician supervision) in most jurisdictions, requires direct patient contact and safety monitoring, carries malpractice risk for misinterpretation of neuromuscular findings, and involves clinical judgment that healthcare systems and regulators expect a credentialed human to authorize.
Adoption barriersclaude-sonnet-55/5Electrodiagnostic testing requires a licensed physician (or supervised technologist) to perform and interpret, with direct patient contact, needle insertion, and legal responsibility for diagnosis—hard regulatory and licensing barriers apply.
Cost vs. human wageclaude-haiku-4-5-202510011/5The equipment, technical setup, and skilled physician/technician labor required for electrodiagnosis remain expensive. AI interpretation assistance, where deployed, reduces some analysis time but does not eliminate the high cost of the procedure itself, making total cost-per-study comparable to or exceeding the human-delivered workflow.
Cost vs. human wageclaude-sonnet-51/5AI cannot perform the physical procedure at all, so there is no viable AI-only cost comparison; a specialist physician remains required for acquisition.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI-based interpretation tools for EMG and NCS exist in research and limited clinical settings, but no deployed products reliably perform end-to-end electrodiagnosis without human technician execution and physician oversight. The procedural and clinical adaptation requirements mean this remains largely human-performed in production.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs electrodiagnostic testing autonomously; AI is at most used for signal interpretation assistance in research settings, not the procedural task itself.

Prescribe orthotic and prosthetic applications and adaptive equipment, such as wheelchairs, bracing, or communication devices, to maximize patient function and self-sufficiency.

10

CI 9–11 · exposure 16 · augmentation 50 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Rehabilitation medicine is a clinical specialty with limited digital automation; adoption of AI agents in this workflow remains minimal because patient outcomes depend on individualized assessment and the profession has not integrated autonomous recommendation systems into practice.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially physical medicine and rehabilitation, has been slower to adopt AI-driven clinical decision tools compared to information/finance sectors, with adoption concentrated in imaging and documentation rather than device prescription.
Augmentation potentialclaude-haiku-4-5-202510013/5AI could usefully assist physicians by summarizing device options, comparing patient criteria against database-matched recommendations, or flagging contraindications in literature, but the physician must retain decision authority and perform clinical fitting.
Augmentation potentialclaude-sonnet-53/5AI can help by analyzing gait data, imaging, or functional assessments and suggesting equipment options, and can streamline documentation and insurance justification, improving physician efficiency without replacing judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI could assist in identifying candidate devices or reviewing clinical literature, prescribing requires individualized biomechanical assessment, patient tolerance testing, and iterative adjustment that demand direct clinical evaluation. Current systems cannot perform end-to-end assessment and fitting decisions meeting the 50% time-saving threshold.
Task automatabilityclaude-sonnet-52/5Prescribing orthotic/prosthetic devices requires integrating physical exam findings, functional assessment, and patient goals into a clinical judgment call that current AI cannot perform end-to-end reliably; AI can assist with documentation or option generation but not the full prescriptive decision.
Adoption barriersclaude-haiku-4-5-202510015/5Prescribing orthotic and prosthetic devices is a legally licensed medical function requiring physician or licensed prosthetist sign-off; liability, medical-device regulation, and the requirement for clinical judgment mean this task is heavily protected by professional scope-of-practice law.
Adoption barriersclaude-sonnet-55/5Prescribing adaptive medical equipment requires a licensed physician's signature and is tied to medical necessity documentation, insurance/regulatory requirements, and liability for patient safety, creating hard legal barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI tools that might support literature review or device database searching remain niche and low-cost but cannot replace the physician's specialized knowledge; the integrated cost of oversight and validation would exceed the marginal time savings.
Cost vs. human wageclaude-sonnet-51/5Because AI cannot independently perform this licensed clinical task, there is no valid cost comparison—human physician involvement remains mandatory, so AI does not reduce cost for the core task.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs orthotic/prosthetic prescription independently; this remains a hands-on clinical specialty requiring physical examination, movement analysis, and real-time patient feedback that AI systems do not do in production.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously prescribes durable medical equipment or orthotics/prosthetics; this remains a physician-driven clinical decision with no production-scale AI substitute.

Coordinate physical medicine and rehabilitation services with other medical activities.

9

CI 3–16 · exposure 5 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Healthcare adoption of AI coordination tools remains limited; most healthcare IT focus is on EHR interoperability rather than autonomous service coordination. Physicians typically retain manual coordination or use basic scheduling tools rather than advanced AI orchestration.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially clinical coordination and physical medicine, adopts AI slowly due to regulation, liability, and integration complexity, with pilots more common than production deployment for this specific task.
Augmentation potentialclaude-haiku-4-5-202510013/5AI could assist by suggesting available providers, flagging scheduling conflicts, drafting coordination summaries, or tracking referral status—useful productivity aids that keep the physician in control of actual coordination decisions and accountability.
Augmentation potentialclaude-sonnet-53/5AI can assist with documentation, summarizing patient records, flagging care gaps, and drafting communications between providers, improving efficiency while the physician retains coordination authority.
Task automatabilityclaude-haiku-4-5-202510011/5Coordinating services across multiple providers requires complex judgment about patient needs, specialist availability, timing, and interdependencies that demand human decision-making and accountability. Current AI cannot reliably manage this multidisciplinary orchestration end-to-end.
Task automatabilityclaude-sonnet-51/5Coordinating rehab care with other medical activities requires synthesizing clinical judgment, interdisciplinary communication, and real-time decision-making across a care team, which current AI cannot execute end-to-end at equal quality.
Adoption barriersclaude-haiku-4-5-202510014/5Physicians must legally direct and sign off on rehabilitation services and coordinate care decisions under medical licensure and liability requirements. Coordination involves clinical judgment and accountability that cannot be delegated to an automated system without physician oversight.
Adoption barriersclaude-sonnet-55/5This is a licensed-physician responsibility involving legal accountability for patient care coordination, medical liability, and required physician sign-off, creating hard regulatory and liability barriers.
Cost vs. human wageclaude-haiku-4-5-202510012/5Automation would require substantial integration with multiple EHR and provider systems, plus ongoing clinical review to ensure appropriateness. The all-in cost approaches or exceeds the value of the human coordination time saved, especially given liability concerns.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI system performing this coordination task, so cost comparison favors the human physician entirely; any AI support adds cost as a supplement, not a replacement.
Technical feasibility todayclaude-haiku-4-5-202510012/5While scheduling and basic communication tools exist, no deployed product reliably coordinates complex rehabilitation services across diverse medical activities with the clinical judgment required. Most existing tools handle only narrow coordination tasks (e.g., appointment scheduling) rather than holistic service coordination.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously coordinates a patient's rehabilitation plan with other medical services; this remains a physician-led coordination task with no production AI substitute.

Diagnose or treat performance-related conditions, such as sports injuries or repetitive-motion injuries.

9

CI 3–16 · exposure 13 · augmentation 63 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While healthcare is digitizing, adoption of AI for autonomous diagnosis and treatment remains slow and confined to research/pilot settings; regulatory caution, liability concerns, and physician skepticism limit production deployment in sports medicine.
Sector adoption velocityclaude-sonnet-52/5Clinical medicine, especially procedural specialties like PM&R, adopts AI slowly due to regulatory, liability, and workflow integration constraints compared to purely digital sectors.
Augmentation potentialclaude-haiku-4-5-202510014/5AI tools demonstrably assist physicians by analyzing imaging, generating differential diagnoses, and organizing injury databases, substantially raising productivity in evidence gathering and documentation while the physician retains clinical decision authority.
Augmentation potentialclaude-sonnet-53/5AI can assist with imaging analysis, literature review, and treatment plan documentation, improving efficiency, but the core diagnostic and hands-on treatment work remains human-driven.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with imaging analysis and literature review for sports injuries, the task requires physical examination, patient history synthesis, and real-time diagnostic judgment that current systems cannot reliably perform end-to-end without substantial human oversight and clinical decision-making.
Task automatabilityclaude-sonnet-51/5Diagnosing and treating sports/repetitive-motion injuries requires physical examination, palpation, imaging interpretation in context, and hands-on treatment planning that AI cannot perform end-to-end today.
Adoption barriersclaude-haiku-4-5-202510015/5Diagnosis and treatment of medical conditions are legally restricted to licensed physicians in nearly all jurisdictions; malpractice liability, regulatory medical boards, and the requirement for physician licensure and clinical judgment create hard barriers to full automation.
Adoption barriersclaude-sonnet-55/5Diagnosis and treatment of injuries legally requires a licensed physician, with direct liability for misdiagnosis and physical intervention, making this a hard-barrier medical task.
Cost vs. human wageclaude-haiku-4-5-202510011/5Current AI tools require significant integration, validation, and physician oversight costs that approach or exceed the marginal cost of the physician performing the task directly, especially given liability and malpractice considerations.
Cost vs. human wageclaude-sonnet-51/5Physical exam, hands-on treatment, and injections cannot be delegated to AI, so the physician's cost remains necessary regardless of AI tools used for ancillary tasks.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI products exist for narrowly scoped components (e.g., MRI segmentation, injury classification from images), but no deployed system independently diagnoses or treats performance-related conditions at clinical reliability; human physicians remain essential for examination, treatment planning, and patient interaction.
Technical feasibility todayclaude-sonnet-51/5No deployed product independently diagnoses or treats musculoskeletal performance injuries in production; AI is used only as a decision-support adjunct for imaging or documentation.

Examine patients to assess mobility, strength, communication, or cognition.

8

CI 0–16 · exposure 13 · augmentation 38 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Despite digital health expansion, actual AI-driven replacement or displacement of physical examination by physicians is minimal. Adoption remains limited to pilot projects and research settings; real clinical practice still centers on direct physician-patient examination as standard of care.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially hands-on clinical exam components, remains a laggard sector for AI deployment due to regulatory, safety, and physical-presence constraints, despite faster adoption in documentation and diagnostics support.
Augmentation potentialclaude-haiku-4-5-202510012/5AI tools can provide some supplementary analysis (e.g., motion tracking, automated cognitive screening scoring) to assist documentation or flag abnormalities, but the core interpretive and diagnostic work remains physician-driven, limiting transformative augmentation potential.
Augmentation potentialclaude-sonnet-53/5AI can assist with cognitive assessment tools, speech/communication analysis, documentation of findings, and pattern recognition from wearable sensor data, improving efficiency and consistency without replacing the physician's direct exam.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with analyzing some aspects (e.g., range-of-motion video analysis, cognitive screening questionnaires), the task fundamentally requires hands-on physical examination, real-time clinical observation, and adaptive interaction that current AI systems cannot reliably replicate end-to-end. Patient examination demands tactile feedback, dynamic reassessment, and clinical judgment that AI cannot match at quality parity.
Task automatabilityclaude-sonnet-51/5Physical examination requires hands-on assessment of strength, range of motion, and direct patient interaction that current AI cannot perform end-to-end; AI cannot physically manipulate limbs or observe subtle physical signs.The core act of examining a patient is fundamentally physical and interpersonal.
Adoption barriersclaude-haiku-4-5-202510015/5Examination of patients for clinical assessment is a licensed clinical function that requires a qualified physician or clinician to perform; liability, clinical responsibility, and regulatory requirements (state medical boards, Medicare/insurance standards) mandate human professional judgment and direct patient contact for valid assessment.
Adoption barriersclaude-sonnet-55/5Physical examination for diagnosis and treatment planning legally requires a licensed physician; liability, medical licensing boards, and standard of care mandate direct physician assessment of mobility, strength, and cognition.
Cost vs. human wageclaude-haiku-4-5-202510011/5The infrastructure, oversight, and integration costs to deploy AI-assisted examination systems remain high relative to physician time, and the necessary human validation/oversight negates meaningful cost advantage. A physician's direct examination is still more cost-effective than hybrid AI systems for this task.
Cost vs. human wageclaude-sonnet-51/5Since AI cannot perform the physical examination itself, there is no viable cost comparison for full task replacement; any AI-assisted components add cost without eliminating the need for the physician's presence.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some narrow components (e.g., gait analysis from video, cognitive screening tests) have prototype or limited-deployment tools, but no mature product reliably performs comprehensive patient examination (mobility, strength, communication, cognition assessment) across the full scope physicians require. Error rates and missing contextual judgments keep feasibility low in production settings.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs physical examinations of mobility or strength; some AI tools assist with cognitive screening or speech analysis, but these are adjuncts, not substitutes for the physician's hands-on exam.

Consult or coordinate with other rehabilitative professionals, including physical and occupational therapists, rehabilitation nurses, speech pathologists, neuropsychologists, behavioral psychologists, social workers, or medical technicians.

4

CI 0–7 · exposure 5 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Rehabilitation and multidisciplinary coordination remain human-contact-intensive and relational; adoption is primarily limited to narrow administrative support tools, not autonomous task replacement in production settings.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially inpatient rehabilitation medicine, has lagged in AI adoption for clinical coordination tasks despite growth in administrative AI tools.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by summarizing prior therapy notes, flagging communication tasks, or organizing input from different specialists, but the core act of consultation and clinical integration remains physician-led and human-intensive.
Augmentation potentialclaude-sonnet-53/5AI can assist by summarizing patient records, drafting communication notes, or flagging care gaps, but the interpersonal coordination itself remains human-led.
Task automatabilityclaude-haiku-4-5-202510011/5This task fundamentally requires real-time professional judgment, interpersonal negotiation, and coordinated care decisions across disciplines. AI systems cannot meaningfully replace the clinical reasoning, accountability, and trust-building inherent in physician-led multidisciplinary consultations.
Task automatabilityclaude-sonnet-51/5Interdisciplinary care coordination requires real-time clinical judgment, negotiation, and relationship-building among human professionals that current AI cannot perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Regulatory and legal frameworks require a licensed physician to be accountable for care coordination decisions; liability, scope-of-practice laws, and the necessity of professional judgment create hard barriers to full substitution.
Adoption barriersclaude-sonnet-55/5Physician licensure, liability for treatment decisions, and legal requirements for physician oversight of rehabilitation plans make this a hard barrier task.
Cost vs. human wageclaude-haiku-4-5-202510011/5Current AI systems would require significant human oversight to validate clinical recommendations and maintain care continuity, making the all-in cost comparable to or exceeding the physician's time savings from automation.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this coordination task, so cost comparison favors the human physician entirely.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can draft summary documents or flag communication gaps, no deployed product reliably performs the active consultation and coordination role itself. Tools exist to assist documentation but not to substitute for the physician's decision-making authority in orchestrating care.
Technical feasibility todayclaude-sonnet-51/5No deployed product coordinates multidisciplinary rehabilitation care autonomously; existing tools only support scheduling or documentation, not the substantive clinical consultation.

Instruct interns and residents in the diagnosis and treatment of temporary or permanent physically disabling conditions.

4

CI 0–7 · exposure 5 · augmentation 50 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Medical training institutions remain dependent on direct physician-to-trainee instruction as a regulatory and accreditation requirement; AI adoption in this specific pedagogical role is not observed in practice and regulatory frameworks actively resist substitution.
Sector adoption velocityclaude-sonnet-52/5Academic medicine adopts AI tools slowly for administrative and educational support, but the core teaching/supervisory role remains largely untouched by AI adoption trends.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist instructors by generating case summaries, diagnostic decision-support references, or multimedia educational materials, moderately enhancing preparation and supplementary learning, but does not transform the core teaching interaction itself.
Augmentation potentialclaude-sonnet-53/5AI can assist by generating case studies, quizzes, differential diagnosis exercises, and summarizing literature to support teaching, improving efficiency without replacing the mentor role.
Task automatabilityclaude-haiku-4-5-202510011/5Instruction of trainees requires real-time interactive teaching, demonstration of clinical reasoning, personalized feedback, and mentorship—elements that demand human judgment, adaptive communication, and relationship-building that current AI cannot perform end-to-end.
Task automatabilityclaude-sonnet-51/5Clinical teaching requires live supervision of trainees performing exams, procedures, and patient interactions, plus adaptive mentorship that current AI cannot replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Medical education and resident training are heavily regulated (ACGME accreditation standards require licensed physician oversight and direct instruction); liability, credential requirements, and legal mandates for human physician supervision create hard barriers to automation.
Adoption barriersclaude-sonnet-55/5Medical education and residency supervision are governed by accreditation bodies (ACGME) requiring licensed attending physicians to directly teach and evaluate trainees, a hard regulatory barrier.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI cannot yet replace the physician-educator role; the cost of deploying and integrating AI systems to assist would add expense rather than reduce the need for attending physicians to instruct residents in hands-on clinical settings.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing the full supervisory teaching function, so cost comparison favors the human physician who also carries licensure and liability responsibilities.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can generate educational content or provide diagnostic references to support teaching, no deployed system can reliably conduct live clinical instruction, observe trainee performance, or deliver mentored diagnosis and treatment planning at production scale in medical education.
Technical feasibility todayclaude-sonnet-51/5No deployed product replaces physician-led clinical instruction of interns/residents; existing AI is limited to supplementary educational content, not the supervisory teaching role itself.

Related occupations — Healthcare Practitioners & Technical

How to read this

A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.

What would change this score

New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.