Emergency Medicine Physicians

29-1214.00
Median wage $335,550/yr32,880 employed (US)Rank #861 of 923 scored · top 93% by substitution

Make immediate medical decisions and act to prevent death or further disability. Provide immediate recognition, evaluation, care, stabilization, and disposition of patients. May direct emergency medical staff in an emergency department.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution12
Exposure12
Augmentation56

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

17 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%11

panel mean rating 1.4/5 → substitution pressure 11/100

Technical feasibility todayw 20%15

panel mean rating 1.6/5 → substitution pressure 15/100

Cost vs. human wagew 15%13

panel mean rating 1.5/5 → substitution pressure 13/100

Adoption barriersw 20%inverted — strong barriers lower the score5

panel mean rating 4.8/5 (barrier strength) → substitution pressure 5/100

Sector adoption velocityw 10%20

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

Task breakdown (17 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.

Collect and record patient information, such as medical history or examination results, in electronic or handwritten medical records.

41

CI 32–50 · exposure 42 · augmentation 88 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Hospitals and ED systems have deployed voice transcription and scribe-assist tools, but adoption remains incomplete and use cases narrow; many EDs still rely on direct physician typing or medical scribes, indicating moderate rather than rapid or deep automation.
Sector adoption velocityclaude-sonnet-53/5Ambient AI documentation is spreading quickly in some large hospital systems but adoption in emergency departments specifically is still uneven and pilot-stage in many institutions.
Augmentation potentialclaude-haiku-4-5-202510014/5Voice-to-text transcription, template population, auto-suggestion of common findings, and real-time EHR integration significantly reduce physician documentation time and cognitive load; these tools substantially augment emergency physician productivity while preserving medical decision-making.
Augmentation potentialclaude-sonnet-55/5AI scribes and voice-to-text tools substantially reduce documentation burden and let physicians focus more on patient care while still reviewing and finalizing records.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can capture and transcribe some clinical information from structured inputs (vital signs, dictation), collecting comprehensive patient information requires real-time patient interaction, clarification of symptoms, and clinical judgment about what information is relevant—tasks that remain heavily dependent on human clinicians today.
Task automatabilityclaude-sonnet-53/5AI dictation/ambient scribe tools can transcribe and structure history and exam findings, but physicians must still verify accuracy, especially in fast-paced ED settings with incomplete or noisy information.
Adoption barriersclaude-haiku-4-5-202510014/5Medical record documentation is legally mandated and directly tied to billing, liability, and informed consent; physicians bear legal responsibility for record accuracy, creating strong liability and regulatory barriers to full automation without physician sign-off.
Adoption barriersclaude-sonnet-53/5The physician remains legally responsible for the accuracy and completeness of the medical record, so oversight is mandatory even though the documentation act itself isn't heavily regulated.
Cost vs. human wageclaude-haiku-4-5-202510012/5Voice transcription and EHR integration tools cost on the order of $0.50–$2.00 per encounter after infrastructure, while emergency physician time is $150–$300/hour; full replacement would be cost-effective only if quality and medico-legal liability were equivalent, which they are not.
Cost vs. human wageclaude-sonnet-53/5AI scribe subscriptions are cheaper than dedicated human scribes but still require licensing fees, integration, and physician oversight time, making savings moderate rather than order-of-magnitude.
Technical feasibility todayclaude-haiku-4-5-202510013/5Voice transcription and structured data entry tools exist in production EHR systems, but they still require substantial physician oversight and correction; no system reliably captures complete, accurate patient history without human validation and follow-up questioning.
Technical feasibility todayclaude-sonnet-53/5Ambient clinical documentation products (e.g., Nuance DAX, Abridge) are deployed in EDs today but still require physician review and correction, with variable accuracy in high-noise, high-urgency environments.

Identify factors that may affect patient management, such as age, gender, barriers to communication, and underlying disease.

29

CI 20–37 · exposure 30 · augmentation 75 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Emergency departments have begun adopting AI-assisted triage and record review, but the pace remains slow because the task is tightly integrated with real-time clinical decision-making and carries high liability; adoption remains primarily in pilots rather than mainstream production use.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially emergency medicine, has been slower than other professional sectors to deploy AI directly into real-time clinical judgment tasks, despite growing EHR-integrated decision support pilots.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can effectively assist emergency physicians by surfacing relevant comorbidities, communication barriers, and demographic flags from records, allowing physicians to focus on synthesis and judgment rather than data gathering; this is a high-value augmentation scenario already emerging in well-designed EHR assistants.
Augmentation potentialclaude-sonnet-54/5AI-powered EHR summarization, risk-stratification tools, and clinical decision support can meaningfully help physicians surface relevant patient factors faster, improving efficiency while the physician remains the decision-maker.
Task automatabilityclaude-haiku-4-5-202510012/5AI can extract and flag demographic and clinical data from records, but identifying which factors *affect management* requires complex clinical judgment integrating context, urgency, and treatment options—tasks where AI demonstrates high error rates and cannot reliably replace physician reasoning at equal quality.
Task automatabilityclaude-sonnet-52/5This requires synthesizing real-time clinical context, direct patient assessment, and nuanced judgment about how personal factors interact with acute presentation, which current AI cannot autonomously perform end-to-end in an ED setting.dicated only partial support is feasible.
Adoption barriersclaude-haiku-4-5-202510014/5Emergency medicine requires rapid, autonomous physician judgment for patient safety and liability; there is no regulatory permission for an AI system to independently *identify* management-affecting factors without explicit physician review and sign-off, creating a hard requirement for human oversight.
Adoption barriersclaude-sonnet-55/5This is a core physician responsibility tied to licensure, liability, and legal standard of care in emergency medicine; a licensed physician must perform and be accountable for this judgment.
Cost vs. human wageclaude-haiku-4-5-202510014/5EHR integration and clinical NLP tools have relatively low inference costs compared to physician time, but require substantial oversight and validation; the cost of integration and error-checking is still lower than the loaded wage of an attending physician performing this task from scratch.
Cost vs. human wageclaude-sonnet-52/5A physician must still perform the actual assessment and integration into care; AI tools add cost as decision-support layers rather than replacing the clinician's time-consuming judgment task.
Technical feasibility todayclaude-haiku-4-5-202510013/5Products exist to parse patient records and extract demographics/comorbidities, and some EHR systems flag age/gender-relevant protocols, but deployed systems rarely achieve the nuanced causal reasoning needed to identify management-relevant factors without physician oversight and correction.
Technical feasibility todayclaude-sonnet-52/5Clinical decision support tools can flag relevant factors from structured EHR data, but no deployed product independently identifies and integrates these factors into emergency management decisions reliably at scale.

Select, request, perform, or interpret diagnostic procedures, such as laboratory tests, electrocardiograms, emergency ultrasounds, and radiographs.

26

CI 20–32 · exposure 30 · augmentation 75 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Hospitals and emergency departments are actively piloting AI diagnostic interpretation tools, particularly for radiology and ECG, but adoption remains in the pilot and early-production phase. Deep, system-wide displacement has not yet occurred despite interest, and adoption varies significantly by institution size and resources.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially acute/emergency care, is a slower-adopting sector for AI diagnostics due to regulatory hurdles, liability concerns, and integration complexity with hospital systems.
Augmentation potentialclaude-haiku-4-5-202510014/5AI-powered diagnostic interpretation and clinical decision support significantly assist emergency physicians by accelerating image review, flagging abnormalities, and suggesting differential diagnoses, thereby improving speed and reducing cognitive load. The human physician remains responsible for selection and final judgment, but AI transforms productivity on interpretation tasks.
Augmentation potentialclaude-sonnet-54/5AI tools meaningfully assist by flagging abnormal ECGs, pre-screening radiographs, and speeding lab result interpretation, helping physicians work faster while retaining decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with interpretation of some diagnostic outputs (e.g., radiology image analysis), the full task requires selecting which procedures are appropriate given clinical context, requesting them within hospital workflows, physically performing procedures (ultrasound, EKG), and integrating results into complex clinical reasoning. Current systems cannot autonomously perform the full cycle or match the 50% time-saving threshold for the integrated decision-making required.
Task automatabilityclaude-sonnet-52/5AI can assist with interpretation of certain images (e.g., radiograph triage, ECG analysis) but the full task of selecting appropriate tests, integrating clinical context, performing bedside ultrasound, and synthesizing results end-to-end is far from fully automatable today.
Adoption barriersclaude-haiku-4-5-202510014/5Ordering, performing, and interpreting diagnostic procedures are legally restricted to licensed physicians in most jurisdictions, and malpractice liability for missed or misinterpreted results creates strong regulatory and liability barriers. Patient contact and legal responsibility requirements mean a human physician must ultimately perform or approve these tasks.
Adoption barriersclaude-sonnet-55/5Ordering and interpreting diagnostic tests in emergency care requires a licensed physician's judgment and legal responsibility, with strict liability, credentialing, and regulatory requirements for clinical decision-making.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI diagnostic interpretation tools reduce interpretation time and cost for specific modalities, but the integrated task of selecting, requesting, performing, and interpreting procedures requires a physician's full labor cost for legal, safety, and liability reasons. Even with AI assistance, the human cost per task remains substantial and comparable to or higher than the AI augmentation value.
Cost vs. human wageclaude-sonnet-52/5AI diagnostic aids add cost on top of physician oversight rather than replacing it, and liability requires physician review, so total cost is not meaningfully lower than physician-only workflows today.
Technical feasibility todayclaude-haiku-4-5-202510013/5Deployed AI products exist for image interpretation (radiology and ultrasound analysis), and some EHR-integrated diagnostic support systems are in use, but they operate with error rates requiring human oversight and typically cover only narrow subsets of the full diagnostic selection and interpretation workflow. Production deployment is real but materially incomplete.
Technical feasibility todayclaude-sonnet-52/5Some FDA-cleared products exist for narrow interpretation tasks (ECG arrhythmia detection, chest X-ray triage flags), but no deployed product performs the full selection-order-perform-interpret workflow reliably in emergency settings.

Refer patients to specialists or other practitioners.

23

CI 20–25 · exposure 25 · augmentation 63 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Emergency medicine has adopted AI for triage and diagnosis support, but referral automation remains low-adoption. Most referral workflows are still manually initiated by physicians; uptake of autonomous referral agents in production is minimal.
Sector adoption velocityclaude-sonnet-52/5Emergency medicine has been slower to adopt autonomous AI decision-making tools compared to administrative or diagnostic-support software, given liability and real-time constraints.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can significantly assist by surfacing relevant specialists, checking patient history and insurance constraints, and highlighting time-sensitive conditions, allowing the physician to make faster, better-informed referral decisions while remaining in control.
Augmentation potentialclaude-sonnet-53/5AI can help identify relevant specialists, generate referral documentation, and flag risk factors, offering moderate assistance while the physician retains final decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist in flagging appropriate specialties based on symptoms and clinical data, but the referral decision requires human judgment about timing, urgency, patient context, and specialist availability. Current systems lack the contextual reasoning and accountability required for autonomous referrals.
Task automatabilityclaude-sonnet-52/5Referral decisions require synthesizing clinical judgment, patient context, and specialist availability that current AI cannot reliably perform end-to-end; AI can assist with drafting referral letters but not the decision itself.
Adoption barriersclaude-haiku-4-5-202510014/5Referrals carry legal and liability implications; physicians are responsible for appropriateness and timeliness. Regulatory and professional standards require a licensed physician to make the clinical judgment and sign the referral. Patient safety liability asymmetry strongly protects this task.
Adoption barriersclaude-sonnet-55/5Referral decisions are a licensed medical judgment requiring physician authority and legal accountability; this cannot be delegated to AI without a licensed practitioner's sign-off.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI infrastructure costs (model inference, integration, clinical oversight) are substantial relative to the time saved, since the physician must still verify appropriateness, check insurance, and make final referral decisions. Full automation would be needed for cost advantage, which is not achieved here.
Cost vs. human wageclaude-sonnet-52/5A physician's referral decision requires minimal marginal time, and the liability/complexity of automating this makes AI oversight costs comparable to or exceeding the human cost for this narrow task.
Technical feasibility todayclaude-haiku-4-5-202510012/5Clinical decision support tools exist that suggest specialist consultations, but no deployed product reliably performs autonomous referrals at quality parity with physician judgment. Systems require human validation and lack integration with the full referral workflow including authorization and scheduling.
Technical feasibility todayclaude-sonnet-52/5Some clinical decision-support tools suggest specialist referrals or flag risk, but no deployed product autonomously manages the referral decision and process reliably in emergency settings.

Evaluate patients' vital signs or laboratory data to determine emergency intervention needs and priority of treatment.

20

CI 20–20 · exposure 25 · augmentation 75 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Despite digitization, emergency medicine adoption of autonomous AI triage systems remains limited. Most adoption is in assistive alerting or documentation, not substitution; organizational and regulatory conservatism in high-acuity settings slows deployment.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially emergency medicine, adopts AI decision-support slowly due to regulatory scrutiny, liability concerns, and integration complexity with EHR/lab systems, despite being a well-resourced sector.
Augmentation potentialclaude-haiku-4-5-202510014/5AI augments this task substantially through real-time vital sign monitoring, automated alert systems, clinical decision support summaries, and lab-to-patient correlation, all of which can accelerate physician assessment and reduce missed criticals while keeping the physician in control.
Augmentation potentialclaude-sonnet-54/5AI-based early warning scores, sepsis alerts, and lab interpretation tools meaningfully help physicians prioritize patients faster and catch abnormalities, while the physician retains final decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5Current AI can extract and flag abnormal vital signs and lab values, but cannot independently determine the full clinical context, competing priorities, or nuanced intervention decisions that emergency medicine demands. Human judgment on symptom interpretation, risk stratification, and triage priority remains essential and not meaningfully automatable.
Task automatabilityclaude-sonnet-52/5AI can flag abnormal vitals/labs and suggest triage priorities, but the actual clinical judgment integrating context, physical exam, and rapidly evolving patient status still requires a physician; full end-to-end automation is not viable today.
Adoption barriersclaude-haiku-4-5-202510015/5Strong regulatory and legal barriers exist: physicians are statutorily required to evaluate and sign off on emergency interventions; malpractice liability for AI errors in life-or-death triage falls on the organization and physician, not the tool vendor; and medical licensing law mandates physician judgment in acute care.
Adoption barriersclaude-sonnet-55/5Triage and emergency treatment decisions legally and ethically require a licensed physician's judgment and sign-off; liability for missed emergencies is severe, making this a hard-barrier task.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of developing, validating, deploying, and maintaining AI systems for emergency decision-making, plus required human oversight and liability management, rivals or exceeds the cost of physician time given liability and error consequences.
Cost vs. human wageclaude-sonnet-52/5Deploying and maintaining clinical decision support systems with integration, validation, and physician oversight is costly relative to the marginal cost of a physician's assessment for this specific sub-task.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI tools can assist with vital sign analysis and flag critical thresholds, but no deployed system reliably performs independent emergency triage and intervention prioritization at production scale. Clinical decision support exists but requires human physician oversight and final judgment.
Technical feasibility todayclaude-sonnet-52/5Clinical decision support tools and sepsis/early-warning algorithms are deployed in some EDs, but they are advisory only with notable false positive/negative rates, not autonomous decision-makers.

Analyze records, examination information, or test results to diagnose medical conditions.

20

CI 20–20 · exposure 25 · augmentation 75 · importance 4.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Emergency medicine has adopted AI for specific tasks (imaging triage, lab alerts) but diagnostic automation remains experimental and fragmented across institutions. Physician skepticism, regulatory caution, and liability concerns slow production deployment; adoption is far below the pace seen in information or finance sectors.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially emergency medicine, is a comparatively slow-adopting sector for autonomous decision-making tools due to regulatory, liability, and safety concerns, though pilot decision-support tools are spreading.
Augmentation potentialclaude-haiku-4-5-202510014/5AI demonstrably assists emergency physicians by flagging abnormal findings, suggesting differential diagnoses, and prioritizing urgent cases, measurably improving diagnostic speed and coverage while physicians retain final decision-making authority and clinical judgment.
Augmentation potentialclaude-sonnet-54/5AI tools (imaging analysis, risk stratification, symptom-checkers, record summarization) meaningfully speed up data synthesis and flag risks, augmenting physician diagnostic reasoning without replacing final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with pattern recognition in imaging and lab data interpretation, diagnosis requires synthesis of clinical history, physical exam findings, and often judgment calls under uncertainty that current AI cannot reliably perform end-to-end with 50% time savings at equal quality. Diagnostic errors carry high stakes and require physician oversight.
Task automatabilityclaude-sonnet-52/5AI can support differential diagnosis by synthesizing records and test results, but emergency diagnosis requires integrating rapidly evolving, ambiguous, high-stakes clinical data and physical exam nuance that current systems cannot reliably do end-to-end.'
Adoption barriersclaude-haiku-4-5-202510015/5Diagnosis by unlicensed entities is legally prohibited in most jurisdictions; a licensed physician must legally perform or formally validate diagnostic conclusions. Liability for misdiagnosis falls on the responsible physician, creating strong regulatory and liability barriers to full automation.
Adoption barriersclaude-sonnet-55/5Diagnosing patients is a licensed medical act with direct liability and legal requirements that a physician must render diagnoses, creating a hard regulatory and professional barrier to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI diagnostic tools require significant infrastructure, training data, validation, and physician oversight; the all-in cost per diagnosis supported is still comparable to or exceeds the marginal cost of physician time for similar diagnostic output when accounting for liability and integration costs.
Cost vs. human wageclaude-sonnet-52/5AI diagnostic aids are relatively cheap to run, but given the need for physician oversight, liability review, and integration with EHRs, the net cost savings versus physician time are modest, not order-of-magnitude.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI products exist for narrow diagnostic support (e.g., imaging interpretation, lab anomaly flagging) but no deployed system reliably performs full diagnostic reasoning from mixed record types independently. Real-world diagnostic AI remains research-heavy or narrowly scoped; production systems function as decision-support, not autonomous diagnosticians.
Technical feasibility todayclaude-sonnet-52/5Diagnostic decision-support tools exist and are used in some EDs (e.g., sepsis alerts, triage algorithms), but no deployed product autonomously performs full diagnostic synthesis reliably across the breadth of ED presentations.

Monitor patients' conditions, and reevaluate treatments, as necessary.

14

CI 3–25 · exposure 13 · augmentation 63 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI for autonomous monitoring and treatment reevaluation in emergency departments remains pilot-stage; most deployments are advisory dashboards requiring physician validation rather than replacing the physician's decision-making role in real time.
Sector adoption velocityclaude-sonnet-52/5Hospitals and emergency departments are historically slow, highly regulated adopters of autonomous clinical AI, with deployment limited mostly to alerting and documentation support rather than treatment decision-making.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist emergency physicians by surfacing vital sign trends, flagging deterioration patterns, and suggesting evidence-based treatment protocols, reducing cognitive load and supporting faster decision-making while the physician remains fully in control.
Augmentation potentialclaude-sonnet-54/5AI-based monitoring alerts, sepsis prediction models, and decision-support tools can meaningfully help physicians track deteriorating patients and flag needed reevaluation, improving situational awareness while the physician remains in control.
Task automatabilityclaude-haiku-4-5-202510012/5Continuous patient monitoring involves integrating vital signs, clinical history, and contextual factors to assess condition changes—tasks where AI shows promise in routine cases. However, reevaluation of treatments in emergency settings requires clinical judgment about complex, evolving presentations and trade-offs that current AI cannot reliably do end-to-end without substantial physician oversight, preventing the 50% time-saving threshold.
Task automatabilityclaude-sonnet-51/5Continuous clinical monitoring and real-time treatment reevaluation require physical examination, judgment under uncertainty, and legal accountability that no current AI system can perform end-to-end in the ED setting.
Adoption barriersclaude-haiku-4-5-202510014/5Legal and regulatory barriers are substantial: emergency medicine is practiced under licensure and malpractice liability that requires a licensed physician to make treatment decisions and remain accountable for patient outcomes. Clinical oversight and sign-off by a physician are essentially non-negotiable in emergency care.
Adoption barriersclaude-sonnet-55/5Reevaluating and modifying treatment in emergency care is a licensed physician responsibility with direct liability, mandated clinical oversight, and legal scope-of-practice restrictions that bar AI from independently performing this task.
Cost vs. human wageclaude-haiku-4-5-202510012/5Monitoring infrastructure (hardware + AI models + integration + physician oversight) remains costly relative to a physician's marginal time spent on continuous assessment in an emergency department; the loaded cost is not yet favorable for full substitution.
Cost vs. human wageclaude-sonnet-51/5There is no AI system that performs this task alone, so no valid cost comparison to a human physician exists; any AI tools used are additive costs on top of physician labor.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI systems can assist with vital sign trend detection and flagging anomalies in monitored data, but no deployed product autonomously reevaluates emergency treatments with reliability sufficient for independent use. Most deployed tools are narrow (e.g., sepsis alerts) and require physician confirmation, not replacement of the monitoring and reevaluation cycle.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously monitors ED patients and adjusts treatment plans; existing clinical decision support and monitoring alert systems are narrow adjuncts, not substitutes for physician reevaluation.

Select and prescribe medications to address patient needs.

14

CI 7–20 · exposure 17 · augmentation 75 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While EHR-integrated clinical decision support is common, it remains assistive and advisory. Active substitution of physician prescribing authority by AI is not occurring in production; adoption velocity of autonomous or semi-autonomous AI prescribing remains minimal due to regulatory and liability constraints.
Sector adoption velocityclaude-sonnet-52/5Hospitals and EDs adopt AI slowly for high-stakes clinical decisions due to safety, liability, and regulatory constraints, despite faster adoption of administrative AI tools.
Augmentation potentialclaude-haiku-4-5-202510014/5AI-powered drug interaction checkers, allergy alerts, formulary guidance, and medication suggestion systems meaningfully augment physician prescribing workflows in modern EHRs, reducing cognitive load and error rates. These tools assist the physician while the physician retains full decision authority and responsibility.
Augmentation potentialclaude-sonnet-54/5AI-powered clinical decision support, drug interaction alerts, and dosing calculators meaningfully assist physicians in choosing appropriate medications faster and more safely.
Task automatabilityclaude-haiku-4-5-202510011/5Medication selection and prescription requires real-time assessment of patient history, contraindications, drug interactions, clinical judgment under uncertainty, and legal responsibility for adverse outcomes. Current AI lacks the integrated clinical reasoning and accountability framework necessary to perform this end-to-end; it can only assist with lookups and suggestion, not bear the decision-making burden.
Task automatabilityclaude-sonnet-52/5Selecting and prescribing medications requires integrating physical exam findings, patient history, allergies, and real-time clinical judgment in acute settings; AI can suggest options but cannot independently perform this end-to-end reliably yet.'
Adoption barriersclaude-haiku-4-5-202510015/5Prescribing is legally and professionally restricted to licensed physicians in all jurisdictions; liability for adverse drug events falls on the prescriber, and regulatory bodies (FDA, DEA, medical boards) require human accountability. No automation can remove these hard barriers.
Adoption barriersclaude-sonnet-55/5Prescribing is a licensed medical act requiring physician authority, with strict legal, regulatory, and liability requirements mandating human sign-off.
Cost vs. human wageclaude-haiku-4-5-202510011/5The integration, validation, oversight, and liability costs of AI-assisted medication selection, combined with mandatory physician review and signature, do not reduce total cost per prescription below that of direct physician decision-making. The human remains the rate-limiting step.
Cost vs. human wageclaude-sonnet-52/5AI decision-support software has licensing and integration costs plus mandatory physician review, so total cost is not dramatically cheaper than the physician's marginal time for this specific subtask.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some decision-support tools and drug-interaction checkers exist in EHRs, but no deployed AI system independently selects and prescribes medications in production without physician oversight and final sign-off. Benchmark studies show AI can suggest medications, but real-world deployment requires human validation on every prescription.
Technical feasibility todayclaude-sonnet-52/5Clinical decision support tools exist and offer drug interaction checks and dosing suggestions, but no deployed product autonomously selects and prescribes ED medications without physician oversight.

Conduct primary patient assessments that include information from prior medical care.

7

CI 3–11 · exposure 5 · augmentation 50 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Healthcare is moderately digitized and adopting AI, but adoption of autonomous or semi-autonomous assessment systems in emergency departments remains limited and highly cautious due to liability, regulation, and patient safety stakes. Most deployments are in documentation support, not primary assessment.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially emergency medicine, has been slow to adopt AI for direct clinical assessment tasks due to safety, regulatory, and workflow constraints, though administrative AI tools are spreading.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist with record retrieval, summarization of prior care, and differential diagnosis suggestions, raising a physician's efficiency in reviewing history. However, the augmentation is partial—AI does not substantially transform the core task of conducting the physical and verbal assessment itself.
Augmentation potentialclaude-sonnet-53/5AI can assist by summarizing prior medical records, flagging relevant history, or suggesting differential diagnoses, improving efficiency while the physician retains full responsibility for the assessment.
Task automatabilityclaude-haiku-4-5-202510011/5While AI can summarize prior medical records and suggest differential diagnoses, the task requires real-time clinical examination, patient interaction, and judgment under urgency that current AI cannot perform end-to-end. The integration of physical exam findings, patient communication, and dynamic decision-making remains beyond autonomous AI capability.
Task automatabilityclaude-sonnet-51/5Primary patient assessment in an emergency setting requires hands-on physical examination, real-time clinical judgment, and integration of dynamic patient presentation with history—no current AI can perform this end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Emergency medicine assessments require a licensed physician to perform or legally assume responsibility for the evaluation. Strong regulatory and liability barriers exist: no unlicensed entity can autonomously conduct primary patient assessments, and the standard of care mandates physician judgment.
Adoption barriersclaude-sonnet-55/5Licensed physicians are legally required to perform patient assessments in emergency medicine, with strong liability, regulatory, and human-contact requirements.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI can assist with record review and documentation, reducing clerical overhead, but the core assessment task requires a physician. Even accounting for time savings on documentation, the loaded cost of human physician time far exceeds the marginal cost savings from partial AI assistance.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the physician performing this task, so there is no viable cost comparison—human physicians remain mandatory.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI systems can extract and summarize historical medical data from records, but no deployed product reliably performs the full primary assessment—which demands direct patient examination, verbal history-taking, and immediate clinical judgment. Existing tools support components only, not the integrated clinical assessment.
Technical feasibility todayclaude-sonnet-51/5No deployed product independently conducts primary emergency patient assessments; AI is at most used for documentation or triage support, not the assessment itself.

Stabilize patients in critical condition.

1

CI 0–3 · exposure 0 · augmentation 50 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5This task is inherently human-centric and non-automatable; adoption velocity for AI replacement is zero, though AI tools for diagnostic support are slowly entering some emergency departments as assistants.
Sector adoption velocityclaude-sonnet-52/5Emergency medicine is adopting AI for triage support, imaging, and documentation, but direct clinical stabilization tasks remain almost entirely human-performed with slow, cautious integration of any AI tools.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist emergency physicians through rapid diagnostic image analysis, clinical decision support, and data synthesis to inform triage and treatment choices, but the physician remains the decision-maker and performer of critical interventions.
Augmentation potentialclaude-sonnet-53/5AI can assist via predictive analytics (sepsis alerts, deterioration scores), rapid documentation, and diagnostic support that help physicians act faster, though it doesn't perform the stabilization itself.
Task automatabilityclaude-haiku-4-5-202510011/5Stabilizing critical patients requires real-time clinical judgment, physical intervention (intubation, chest compressions, medication administration), and rapid adaptation to changing physiology—capabilities far beyond current AI systems, which cannot perform the hands-on medical procedures and dynamic decision-making this task demands.
Task automatabilityclaude-sonnet-51/5Stabilizing critically ill patients requires hands-on physical intervention, rapid multi-sensory clinical judgment, and physical procedures (intubation, chest tubes, defibrillation) that no current AI system can perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Hard legal and regulatory barriers exist: only licensed physicians can make clinical decisions and perform emergency interventions; medical liability, malpractice law, and medical practice acts require human physician accountability and cannot be substituted by AI.
Adoption barriersclaude-sonnet-55/5Licensed physician judgment and hands-on procedures are legally required, with extreme liability exposure and mandatory human accountability for life-critical interventions.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI has no role in performing this task, so cost comparison is moot; the full labor cost remains with the emergency physician and cannot be displaced by automation.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical, high-stakes task, so cost comparison favors the human physician entirely; any AI role is supplementary, not substitutive.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product performs patient stabilization end-to-end; this remains a purely human clinical function requiring physical presence, licensure, and accountability for life-or-death decisions that no current system is authorized or capable of executing independently.
Technical feasibility todayclaude-sonnet-51/5No deployed product independently stabilizes critical patients; AI exists only as decision-support or monitoring adjuncts, not as an autonomous actor in resuscitation.

Consult with hospitalists and other professionals, such as social workers, regarding patients' hospital admission, continued observation, transition of care, or discharge.

1

CI 0–3 · exposure 0 · augmentation 50 · importance 4.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Adoption of AI for actual consultation decisions is minimal. Healthcare institutions remain highly conservative on replacing clinician consultation and disposition decisions due to liability, regulatory requirements, and patient safety concerns.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially acute/emergency care coordination, adopts AI slowly due to regulatory, liability, and interoperability constraints, though documentation tools are spreading.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by summarizing patient records, flagging relevant lab or imaging results, or drafting documentation templates, which could modestly improve consultation efficiency. However, augmentation is limited because the core task—interpersonal clinical negotiation and judgment—remains fundamentally human-driven.
Augmentation potentialclaude-sonnet-53/5AI can help summarize patient records, flag risk factors, or draft handoff notes to support these discussions, but the core interpersonal consultation remains human-driven.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time clinical judgment, understanding of complex patient circumstances, and negotiation with multiple professionals. Current AI cannot reliably conduct consultations that meet the standard of care, assess nuanced clinical and social factors, or make binding decisions on patient disposition.
Task automatabilityclaude-sonnet-51/5This requires real-time clinical judgment, negotiation, and coordination among multiple professionals about complex patient trajectories; no current AI system can conduct these consultations autonomously.'
Adoption barriersclaude-haiku-4-5-202510015/5Strong legal and regulatory barriers exist: physicians must exercise independent clinical judgment on admission and discharge decisions, carry malpractice liability for those decisions, and hospital credentialing requires a licensed physician to sign off. Patient safety regulations and standard of care mandate human physician involvement.
Adoption barriersclaude-sonnet-55/5Admission, discharge, and care transition decisions legally require licensed physician judgment and sign-off, with high liability exposure for errors.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of AI consultation systems (accounting for integration, EHR connectivity, liability oversight, and physician review of recommendations) would exceed the cost of direct physician-to-physician consultation, which takes minutes of physician time.
Cost vs. human wageclaude-sonnet-51/5There is no AI product performing this function, so no cost comparison favors AI; human clinicians remain the only option.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs this multi-stakeholder consultation and clinical decision-making task. While AI can assist with information retrieval or documentation, independent performance of consultation and disposition decisions remains research-stage and not deployable at patient-care scale.
Technical feasibility todayclaude-sonnet-51/5No deployed product substitutes for physician-to-physician or physician-to-social-worker case consultation regarding disposition decisions.

Communicate likely outcomes of medical diseases or traumatic conditions to patients or their representatives.

1

CI 0–3 · exposure 0 · augmentation 38 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Healthcare adoption of AI remains slow and cautious in high-stakes communication tasks; no measurable displacement of physician-patient outcome discussions has occurred, and regulatory and professional norms strongly discourage substitution.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially acute/emergency care, has been slow to adopt AI for direct patient communication due to liability, trust, and regulatory constraints, despite faster AI adoption in administrative healthcare functions.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could assist by drafting talking points or pulling relevant outcome statistics, but the task itself—the actual conversation—depends critically on the physician's judgment and presence, limiting practical augmentation value in the clinical context.
Augmentation potentialclaude-sonnet-53/5AI can help physicians by summarizing clinical data, drafting explanatory materials, or suggesting communication frameworks, but the actual sensitive conversation still relies heavily on human delivery and judgment.
Task automatabilityclaude-haiku-4-5-202510011/5Communicating medical outcomes requires genuine empathy, nuanced judgment about patient readiness, and real-time adaptation to emotional cues that current AI systems cannot reliably replicate. The task inherently demands a licensed human to deliver news and answer follow-up questions in a medically and ethically appropriate manner.
Task automatabilityclaude-sonnet-51/5Communicating prognosis in emergency settings requires real-time clinical judgment, empathy, and adaptive dialogue with distressed patients/families that current AI cannot perform end-to-end reliably or ethically.
Adoption barriersclaude-haiku-4-5-202510015/5Medical practice law and standard of care require a licensed physician to communicate critical clinical information and prognosis directly to the patient or surrogate. Liability, informed consent, and regulatory requirements create absolute barriers to automation or independent AI deployment.
Adoption barriersclaude-sonnet-55/5Licensure, medical liability, informed consent law, and the ethical/legal requirement for a qualified physician to deliver diagnoses and prognoses create hard barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5The physician's presence and legal liability for this conversation cannot be substituted by AI inference; oversight costs and human involvement remain necessary, making the all-in cost of any AI system additive rather than cost-saving.
Cost vs. human wageclaude-sonnet-51/5Since AI cannot perform this task independently, there is no viable cost comparison—any attempt would require full physician involvement plus additional oversight, making AI more expensive in practice.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs this task end-to-end in clinical settings. While AI can generate factual medical summaries or outcome statistics, the communicative and relational aspects—explaining uncertainty, addressing patient fears, adjusting tone and pacing—remain outside the scope of production systems.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously delivers prognosis communication to patients in emergency care; this remains firmly a physician-performed task with no substitution in production.

Direct and coordinate activities of nurses, assistants, specialists, residents, and other medical staff.

1

CI 0–3 · exposure 0 · augmentation 38 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Emergency medicine is a high-regulation, liability-sensitive domain with entrenched hierarchical protocols; there is minimal industry movement toward AI-directed or autonomous staff coordination in production EDs.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially emergency medicine, adopts AI slowly for hands-on and supervisory clinical roles despite faster uptake in documentation or diagnostics support.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could provide marginal assistance via real-time scheduling suggestions or staff allocation alerts, but the core coordination task—making decisions, communicating priorities, and accountably directing humans—remains deeply human-centric with little genuine productivity lift today.
Augmentation potentialclaude-sonnet-53/5AI can assist with information relay, decision support, staffing/resource optimization, and documentation that supports coordination, but the core directive and leadership function remains human-driven.
Task automatabilityclaude-haiku-4-5-202510011/5Directing and coordinating medical staff requires real-time human judgment, emotional intelligence, conflict resolution, and accountability for clinical decisions—capabilities that current AI cannot replicate end-to-end in a high-stakes emergency setting where authority and legal responsibility vest in the physician.
Task automatabilityclaude-sonnet-51/5Real-time leadership and coordination of a multidisciplinary clinical team during dynamic, high-stakes emergency situations requires physical presence, authority, and split-second judgment that current AI cannot replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Legal and regulatory frameworks place clinical authority and accountability on the licensed physician; hospitals require a human MD to direct care teams, and liability for delegation and coordination decisions cannot be transferred to an AI system.
Adoption barriersclaude-sonnet-55/5Medical licensure, legal accountability for patient outcomes, and hierarchical clinical authority structures require a licensed physician to direct care teams, making this a hard regulatory and liability barrier.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI systems capable of meaningful coordination support would require extensive custom integration, continuous oversight by the physician, and would not reduce the physician's workload or salary requirements, making the all-in cost higher than the AI benefit.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute for this supervisory role, so cost comparison favors the human entirely; any AI attempt would require extensive human oversight negating savings.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI system today can autonomously manage clinical staff coordination, prioritization decisions, or real-time triage leadership in an emergency department; research chatbots and workflow tools exist but do not perform this orchestration function in production.
Technical feasibility todayclaude-sonnet-51/5No deployed product directs or coordinates emergency department staff; AI is not used as a team leader or supervisor in clinical practice today.

Discuss patients' treatment plans with physicians and other medical professionals.

1

CI 0–3 · exposure 0 · augmentation 38 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Healthcare adoption of AI remains cautious and heavily supervised; clinical discussion and consensus-building between physicians has shown minimal AI displacement. Physician workflows still depend on human-to-human communication for treatment-plan deliberation.
Sector adoption velocityclaude-sonnet-52/5Healthcare overall adopts AI unevenly, and physician-to-physician clinical consultation is one of the slowest areas due to liability, regulation, and the primacy of human judgment.
Augmentation potentialclaude-haiku-4-5-202510012/5AI can provide background summaries, evidence synthesis, or decision-support notes that inform a discussion, but cannot augment the discussion itself meaningfully. The interaction remains fundamentally human-driven, and AI's role is limited to pre-discussion preparation rather than real-time augmentation of the collaborative process.
Augmentation potentialclaude-sonnet-53/5AI can summarize charts, suggest differential diagnoses, or draft consult notes to prepare physicians for these discussions, offering moderate but not transformative support.
Task automatabilityclaude-haiku-4-5-202510011/5This task fundamentally requires real-time clinical judgment, negotiation, and accountability in high-stakes settings. Current AI cannot independently conduct meaningful discussions with physicians about complex treatment decisions, validate clinical reasoning across specialists, or own responsibility for care recommendations.
Task automatabilityclaude-sonnet-51/5This is a synchronous, high-stakes clinical communication task requiring real-time judgment, accountability, and trust between licensed professionals; no AI system can substitute for the physician in this interaction today.jaw
Adoption barriersclaude-haiku-4-5-202510015/5Medical decision-making and collaborative care discussions are heavily regulated and involve malpractice liability. Licensing laws, standard-of-care requirements, and accountability frameworks legally require licensed physicians to lead treatment discussions; AI participation as a discussion peer is not permitted or recognized in practice.
Adoption barriersclaude-sonnet-55/5Only licensed physicians can legally discuss and authorize treatment plans, with strong liability, licensure, and clinical-standard-of-care requirements barring any AI substitution.
Cost vs. human wageclaude-haiku-4-5-202510011/5An emergency medicine physician's time discussing treatment plans with colleagues is high-value clinical work ($50–150+/hour loaded). Current AI produces mainly supportive documents rather than conduct discussions, offering no cost advantage over human collaboration.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this task, so no meaningful cost comparison exists; the human physician remains the only viable performer.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably conducts genuine treatment-plan discussions with medical professionals today. AI can draft summaries or suggest talking points, but cannot participate as a peer in interactive clinical deliberation with the judgment and accountability physicians expect.
Technical feasibility todayclaude-sonnet-51/5No deployed product conducts or replaces inter-physician treatment plan discussions; AI is at best a passive documentation or decision-support aid, not a participant in the discussion itself.

Assess patients' pain levels or sedation requirements.

1

CI 0–3 · exposure 0 · augmentation 38 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Emergency medicine remains highly conservative on autonomous clinical decision-making due to liability and regulatory constraints. Adoption of AI for pain/sedation assessment has been minimal because the task fundamentally requires physician judgment and responsibility.
Sector adoption velocityclaude-sonnet-52/5Healthcare, especially emergency medicine, is a slower-adopting sector for clinical decision-making AI due to regulatory, liability, and safety constraints, though documentation tools are spreading.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could modestly assist by aggregating vital signs, suggesting pain scales, or flagging physiological changes, but the core clinical assessment and decision-making remain physician-centric. Augmentation potential is limited because most of the value already lies in human observation and expertise.
Augmentation potentialclaude-sonnet-53/5AI can assist with pain scale documentation, flagging vital sign trends, or suggesting sedation protocols based on guidelines, but the physician must still perform the core assessment.
Task automatabilityclaude-haiku-4-5-202510011/5Assessing pain and sedation requires subjective clinical judgment, direct patient observation, behavioral interpretation, and real-time interactive communication—none of which current AI systems can reliably perform autonomously in a clinical setting. This task demands human clinical expertise and cannot meet the 50% time-saving threshold without human oversight.
Task automatabilityclaude-sonnet-51/5This requires direct physical examination, real-time patient interaction, observation of vital signs, and clinical judgment in a dynamic emergency setting that current AI cannot perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Legal and regulatory barriers are absolute: only licensed physicians can assess and manage pain/sedation in emergency settings. Liability, patient safety, and medical licensure laws require human physician sign-off on all analgesia and sedative decisions.
Adoption barriersclaude-sonnet-55/5Determining sedation requirements involves prescribing controlled substances and clinical judgment that legally must be performed by a licensed physician, with high liability for error.
Cost vs. human wageclaude-haiku-4-5-202510011/5Current AI solutions that could assist with pain assessment (e.g., physiological monitoring systems) still require full physician review and decision-making, making the combined cost of AI plus human expert time higher than direct physician assessment alone.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this task alone, so cost comparison favors the human physician who must be present regardless.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product reliably assesses patient pain or sedation requirements independently in emergency medicine. While AI can aid in analyzing physiological signals, the clinical decision itself requires a licensed physician's real-time evaluation.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously assesses patient pain or sedation needs in emergency settings; this remains a hands-on clinical judgment task requiring physical presence.

Perform emergency resuscitations on patients.

0

CI 0–0 · exposure 0 · augmentation 38 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Adoption of AI for autonomous resuscitation is essentially zero—the task remains firmly in the domain of human clinicians by necessity. No healthcare sector is displacing physicians from resuscitation duties with AI agents.
Sector adoption velocityclaude-sonnet-51/5Emergency medicine's hands-on acute care remains almost entirely human-delivered with minimal AI displacement of the physical resuscitation act itself.
Augmentation potentialclaude-haiku-4-5-202510012/5AI offers limited augmentation: rhythm recognition and drug-dosing reminders can assist decision-making, but the task's dynamic, time-critical, and physically demanding nature means AI assistance has modest productivity impact compared to the physician's own expertise and judgment.
Augmentation potentialclaude-sonnet-53/5AI-enabled monitoring, defibrillator guidance, and decision-support tools (e.g., real-time vital sign analysis, protocol prompts) can assist clinicians during resuscitation without replacing the physical task.
Task automatabilityclaude-haiku-4-5-202510011/5Emergency resuscitation requires real-time physical intervention, manual dexterity, and immediate adaptation to dynamic physiological changes—tasks that no current AI system can perform autonomously. While AI can assist with decision support, the core manual and sensorimotor components (chest compressions, airway management, drug administration) remain entirely dependent on human clinicians.
Task automatabilityclaude-sonnet-51/5Emergency resuscitation requires real-time physical manipulation (chest compressions, airway management, defibrillation, IV access) that no current AI system can physically execute.
Adoption barriersclaude-haiku-4-5-202510015/5Emergency resuscitation is legally and clinically required to be performed by licensed physicians or paramedics under strict protocols. Patient contact, immediate life-or-death accountability, and regulatory requirements (state medical boards, hospital credentialing) create hard barriers to any form of autonomous substitution.
Adoption barriersclaude-sonnet-55/5Resuscitation is a licensed medical procedure requiring physician/clinician authority, legal liability, and immediate physical human presence—hard regulatory and physical barriers.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI has no cost advantage here because it cannot perform the task end-to-end; human emergency physicians remain indispensable and command high wages, while AI tools provide only marginal support.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical task, so no meaningful cost comparison exists; a trained physician/team is mandatory.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product can autonomously perform emergency resuscitation. Current systems offer only decision-support (rhythm interpretation, drug dosing prompts), which requires a physician to execute all interventions.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs hands-on resuscitation; AI exists only in adjacent decision-support or documentation roles, not the physical act itself.

Perform such medical procedures as emergent cricothyrotomy, endotracheal intubation, and emergency thoracotomy.

0

CI 0–0 · exposure 0 · augmentation 25 · importance 5.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Emergency medicine remains highly physician-dependent with no meaningful adoption of autonomous AI for these critical procedures. Adoption velocity is negligible because legal, safety, and immediate-care requirements preclude it.
Sector adoption velocityclaude-sonnet-51/5Emergency medicine's hands-on procedural care remains a highly physical, low-digitization domain with essentially no movement toward automating invasive procedures.
Augmentation potentialclaude-haiku-4-5-202510012/5AI may assist with pre-procedure imaging analysis or decision support, but provides minimal augmentation during the invasive procedure itself. Real-time guidance systems exist in research, but meaningful in-situ augmentation during emergency cricothyrotomy or thoracotomy remains limited.
Augmentation potentialclaude-sonnet-52/5AI may assist with pre-procedure decision support (e.g., ultrasound guidance interpretation, checklists) but offers minimal augmentation for the actual manual execution of these emergency interventions.
Task automatabilityclaude-haiku-4-5-202510011/5These are complex surgical/invasive procedures requiring real-time manual dexterity, spatial reasoning, and immediate adaptive judgment in life-threatening conditions. Current AI systems cannot perform physical interventions and lack the tactile feedback and in-situ decision-making necessary for emergency airway or chest procedures.
Task automatabilityclaude-sonnet-51/5These are hands-on invasive emergency procedures requiring physical dexterity, real-time tactile feedback, and rapid decision-making in life-threatening situations; no AI system can physically perform these interventions today.
Adoption barriersclaude-haiku-4-5-202510015/5These procedures are legally restricted to licensed physicians and often require specific credentialing. Liability, malpractice, and regulatory oversight (state medical boards, hospital privileging) create hard barriers that prevent non-physician automation.
Adoption barriersclaude-sonnet-55/5These procedures legally and ethically require a licensed physician (or credentialed provider) to perform, with direct liability, hands-on skill, and immediate human judgment demanded in emergencies.
Cost vs. human wageclaude-haiku-4-5-202510011/5Emergency procedures demand immediate physician presence and cannot be delegated to AI at any meaningful cost savings. The human physician's expertise and liability remain irreplaceable; AI cannot substitute at lower cost.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical task, so cost comparison is moot; the human physician remains the only means of delivering this care, making AI infinitely more 'expensive' in the sense of being nonexistent as an alternative.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product can autonomously perform cricothyrotomy, intubation, or thoracotomy. These require licensed human physicians and remain research-stage in robotics; no production surgical robots perform them independently in emergency contexts.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs cricothyrotomy, intubation, or thoracotomy; this remains entirely in the domain of trained physicians and is not addressed by any commercial or research robotic system for emergency use.

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.