Preventive Medicine Physicians
29-1229.05Median wage $265,930/yr342,720 employed (US)Rank #657 of 923 scored · top 71% by substitution
Apply knowledge of general preventive medicine and public health issues to promote health care to groups or individuals, and aid in the prevention or reduction of risk of disease, injury, disability, or death. May practice population-based medicine or diagnose and treat patients in the context of clinical health promotion and disease prevention.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution21
Exposure19
Augmentation61
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%19
panel mean rating 1.8/5 → substitution pressure 19/100
Technical feasibility todayw 20%19
panel mean rating 1.8/5 → substitution pressure 19/100
Cost vs. human wagew 15%21
panel mean rating 1.8/5 → substitution pressure 21/100
Adoption barriersw 20%inverted — strong barriers lower the score22
panel mean rating 4.1/5 (barrier strength) → substitution pressure 22/100
Sector adoption velocityw 10%25
panel mean rating 2.0/5 → substitution pressure 25/100
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.
Design or use surveillance tools, such as screening, lab reports, and vital records, to identify health risks.
33CI 25–41 · exposure 38 · augmentation 75 · importance 4.2/5 · click for rater detail
Design or use surveillance tools, such as screening, lab reports, and vital records, to identify health risks.
33| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Healthcare has moderate digitization and adoption of analytics tools, but the preventive medicine and surveillance design phase is often slower to adopt autonomous AI than acute clinical settings. Pilots are common; deep production deployment remains limited. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Public health and healthcare surveillance systems are adopting AI/analytics tools gradually, but institutional, regulatory, and interoperability barriers slow deployment compared to fast-adopting sectors like finance or general IT. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially improves a physician's ability to review and synthesize screening and lab data at scale, reducing time spent on routine review and surfacing patterns a human might miss. The physician remains the decision-maker and designer, but productivity gains are substantial. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly enhance surveillance by rapidly processing lab reports, vital records, and screening data to flag anomalies and trends, greatly aiding physicians in identifying health risks faster and more comprehensively. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can assist significantly with data review and flagging anomalies in screening results, lab reports, and vital records, but designing comprehensive surveillance systems and interpreting complex clinical contexts still require physician judgment and integration with patient history. Automating the full task end-to-end meets the threshold for some narrow workflows but not consistently across the diversity of health risks. |
| Task automatability | claude-sonnet-5 | 2/5 | Designing surveillance systems and interpreting complex population-level health data requires clinical judgment, epidemiological expertise, and contextual decision-making that current AI cannot fully replicate end-to-end, though data aggregation and pattern flagging can be partially automated. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Medical liability, regulatory oversight (FDA classification of clinical decision support), and the legal expectation that a licensed physician designs and validates surveillance protocols create strong barriers. Patient trust and informed consent also favor human endorsement of risk identification decisions. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Public health surveillance often involves legally mandated reporting, licensed physician oversight, and liability for missed or misinterpreted health risks, creating strong regulatory and professional barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI infrastructure and continuous validation add meaningful cost; oversight by a physician remains essential, so labor cost is not eliminated. The all-in cost per risk identification remains comparable to or slightly lower than unassisted physician review, not a decisive cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools can cheaply process large datasets, but the physician's role in designing surveillance strategy, interpreting risk, and ensuring regulatory compliance still requires significant human cost, keeping overall cost comparable rather than dramatically cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Products exist that analyze lab results, vital signs, and screening data (EHR analytics, clinical decision support systems), but they operate with material limitations in sensitivity, specificity, and false-positive rates. Production use is common but typically as physician-supervised tools rather than autonomous systems. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some deployed products (e.g., syndromic surveillance dashboards, lab result flagging systems) exist, but they operate as decision-support tools requiring physician oversight rather than autonomously designing or running full surveillance programs. |
Deliver presentations to lay or professional audiences.
31CI 25–37 · exposure 25 · augmentation 75 · importance 3.6/5 · click for rater detail
Deliver presentations to lay or professional audiences.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare organizations are early in adopting AI for presentation tasks; most still rely on physician-authored and physician-delivered presentations. Adoption is limited to narrow use cases (slide templates, draft bullet points) rather than end-to-end automation or replacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Medical and public health sectors adopt AI slower than typical professional services due to cautious institutional culture and low digitization of live presentation delivery. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI tools demonstrably help physicians prepare presentations faster by drafting content, organizing evidence, generating speaker notes, and refining slides, significantly reducing preparation time while the physician retains full control and credibility for delivery. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools substantially help physicians create slides, structure talks, summarize research, and prepare Q&A materials, improving efficiency while the human still delivers the presentation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate presentation slides and draft talking points automatically, delivering a live presentation requires real-time audience engagement, adaptive pacing, and credibility that current systems cannot reliably replicate. The task involves nuanced communication and professional judgment that falls short of the 50% time-saving threshold for full task automation. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft content and slides but the live delivery, audience interaction, and credibility of a physician presenting cannot be fully automated end-to-end at equal quality today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Regulatory and professional norms strongly prefer that physicians deliver medical education and health information directly, particularly on preventive health matters where credibility and accountability matter. Institutional policies and audience expectations create meaningful friction against AI-only delivery. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement to give a talk, but professional credibility, audience trust, and organizational norms favor a human physician delivering such presentations. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-assisted presentation creation (slides, notes, outlines) costs significantly less than the professional time required to write and rehearse presentations from scratch, though a physician's time to deliver remains irreplaceable today. The ratio favors AI for preparation but not delivery. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI-assisted slide creation is cheap, an AI-only substitute for a credible presenter still requires significant human oversight and production costs, making all-in cost savings limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI can assist with presentation preparation (generating content, organizing slides), but no deployed product reliably delivers professional-quality live presentations end-to-end. Systems lack the ability to read audience dynamics, adjust delivery in real time, and maintain the ethos required for medical professionals addressing stakeholders. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI presentation tools (avatars, generative slide decks) exist but are rarely deployed to actually replace physicians delivering talks to professional or lay audiences in production settings. |
Prepare preventive health reports, including problem descriptions, analyses, alternative solutions, and recommendations.
31CI 25–37 · exposure 33 · augmentation 63 · importance 3.9/5 · click for rater detail
Prepare preventive health reports, including problem descriptions, analyses, alternative solutions, and recommendations.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Preventive medicine operates in regulated healthcare settings where digitization is moderate, liability concerns are high, and adoption of AI for clinical decision-making and reporting remains limited to pilots. Actual displacement of this task in production is minimal. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially clinical and public health domains, adopts AI more slowly than information-sector fields due to regulatory caution, liability concerns, and data privacy constraints. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by organizing patient/population data, summarizing evidence, flagging risk factors, and suggesting report structure, enabling physicians to focus on clinical synthesis and recommendations rather than data compilation. This support meaningfully raises physician productivity without full automation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools substantially assist in drafting, summarizing data, and suggesting alternative solutions, meaningfully speeding up report preparation while the physician remains responsible for final content and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist with data aggregation and initial report structure, but preventive medicine reports require synthesis of complex patient/population health data, clinical judgment about risk prioritization, and physician-level interpretation of evidence. Current AI lacks the context integration and medical decision-making depth to produce end-to-end reports meeting the 50% time-saving threshold without substantial physician revision. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft report sections, synthesize epidemiological data, and generate alternative recommendations, but requires physician verification of clinical accuracy and contextual judgment, so full end-to-end automation with equal quality is not yet reliable. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Legal and ethical barriers are substantial: physicians bear liability for report accuracy and clinical appropriateness, regulatory frameworks (state medical boards) require physician judgment and sign-off, and professional standards demand human accountability for clinical recommendations in preventive medicine. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Preventive health reports often carry regulatory, legal, and clinical liability implications requiring physician sign-off, creating strong professional and legal barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI inference plus the required physician review, validation, and rewriting to ensure clinical accuracy approaches or exceeds the cost of a physician drafting the report directly, especially for complex preventive assessments. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI drafting can cut report preparation time significantly, but the need for expert oversight, data verification, and liability review keeps blended costs only moderately below fully human-authored reports. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI tools exist for clinical documentation and some report generation, no deployed product reliably generates complete preventive health reports with problem descriptions, analyses, and clinically sound recommendations at production scale. Existing systems require significant physician oversight and correction. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | LLM-based drafting tools exist and are used for medical writing assistance, but no deployed product autonomously produces validated preventive health reports without substantial physician review and correction. |
Identify groups at risk for specific preventable diseases or injuries.
29CI 25–32 · exposure 30 · augmentation 75 · importance 4.2/5 · click for rater detail
Identify groups at risk for specific preventable diseases or injuries.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Public health agencies and health systems increasingly pilot AI for disease outbreak detection and risk scoring, but adoption remains mixed. Some healthcare organizations deploy predictive analytics for population health; others rely on traditional epidemiological methods. Adoption is faster in large hospital systems and slower in smaller or under-resourced settings. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially public health and preventive medicine, has historically been a slower adopter of AI systems in production compared to information/finance sectors, despite growing pilots in predictive analytics. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially augments preventive medicine work by rapidly processing large datasets, identifying patterns, and suggesting risk groups that physicians can then validate and act upon. Physicians remain essential for refining risk models, considering equity and local context, and translating findings into public health policy. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully enhance this task by analyzing large datasets (EHRs, epidemiological trends, genetic/social risk factors) to flag at-risk populations for physician review, substantially improving efficiency and coverage. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist with identifying risk factors and stratifying populations using epidemiological data, but the task requires clinical judgment, contextual knowledge of community factors, and integration with patient history that current systems cannot fully perform end-to-end. A human physician must ultimately synthesize complex, multifactorial risk profiles. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can assist with data analysis and risk stratification but identifying at-risk groups requires clinical judgment, epidemiological context, and synthesis of diverse data sources that current systems cannot fully replace end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: clinical decision-making and public health authority rests with licensed physicians; liability and negligence law require physician accountability for population health decisions; regulatory bodies (FDA, public health agencies) oversee risk classification tools; and health systems emphasize human review before actionable recommendations. |
| Adoption barriers | claude-sonnet-5 | 4/5 | This task is embedded in physician licensure and clinical judgment requirements; liability for missed at-risk populations and regulatory expectations around clinical decision-making create strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI models for risk stratification incur licensing, data integration, and validation costs, while preventive medicine physicians' time is already accounted for in health systems. The marginal cost of AI is often lower than the physician's loaded wage, but the setup and maintenance overhead is substantial relative to the task's scope. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI-driven analytics tools have upfront and integration costs comparable to or exceeding the marginal cost of physician review for this task, since physician oversight remains essential. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Risk stratification tools and predictive models exist in production (e.g., clinical risk calculators, EHR-integrated tools), but they typically focus on narrow conditions or populations and require physician oversight to validate findings and account for local conditions. Deployed systems work well for specific, well-studied populations but lack generalizability. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some population health analytics and risk-stratification tools are deployed (e.g., in insurance or public health systems), but they are narrow in scope and require significant human interpretation and validation. |
Teach or train medical staff regarding preventive medicine issues.
28CI 25–30 · exposure 25 · augmentation 63 · importance 4.0/5 · click for rater detail
Teach or train medical staff regarding preventive medicine issues.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for medical education is still in pilot phases; most healthcare organizations rely on traditional in-person or instructor-led webinars. Medical training culture prioritizes human expertise and personalized feedback, slowing substitution relative to information-sector adoption patterns. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare training and CME sectors are adopting AI slowly for content support, with pilots more common than production-scale automated training delivery. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by drafting lecture notes, generating evidence summaries on preventive topics, and creating assessment materials, allowing physicians to focus on interactive and adaptive teaching. However, the assistance is partial—design, delivery, and credibility remain human-anchored. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly help physicians prepare training materials, summarize new guidelines, and create quizzes or case studies, boosting efficiency while the physician still delivers instruction. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate educational content and draft training materials on preventive medicine topics, the task fundamentally requires live instruction, real-time adaptation to learner questions, and credibility-building interaction that current AI systems cannot reliably deliver end-to-end. Limited automation of lecture preparation, but core teaching cannot meet the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | Teaching involves synthesizing content, adapting to trainee questions, and modeling clinical judgment; AI can support material creation but cannot independently run interactive training with the credibility of a physician instructor. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Teaching of medical staff is often governed by accreditation standards and continuing medical education (CME) requirements that mandate qualified physician or credentialed educator participation. Liability for incorrect training content and regulatory expectations for human oversight create substantial legal and organizational barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing barrier prevents AI-assisted content, but organizational expectation of an authoritative clinician trainer and liability for clinical accuracy create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-generated training materials reduce some preparation costs, but the human physician's loaded wage for actual instruction time is not substantially undercut by AI inference costs when overhead and quality assurance are included. Training design, customization, and live delivery remain human-intensive. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply draft materials, but the actual training delivery, credibility, and interactive Q&A still require paid physician time, keeping overall cost comparable to human-led training. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs full medical staff training on preventive medicine independently. AI can assist with content generation and practice quizzes, but educational institutions and medical systems rely on humans to deliver and validate training; systems are not mature enough for unsupervised deployment. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products exist for generating educational content, quizzes, and slide decks, but no deployed system autonomously delivers physician-led staff training on preventive medicine at scale. |
Document or review comprehensive patients' histories with an emphasis on occupation or environmental risks.
26CI 25–28 · exposure 25 · augmentation 63 · importance 4.3/5 · click for rater detail
Document or review comprehensive patients' histories with an emphasis on occupation or environmental risks.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI-assisted occupational history documentation remains limited; most preventive medicine practices rely on template-driven EHRs rather than autonomous AI systems. Uptake is slower than in high-volume, lower-stakes documentation tasks due to specialization and liability concerns. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Healthcare has moderate AI adoption via ambient documentation tools and EHR-integrated scribes, though full production-scale replacement of history-taking is still emerging. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by suggesting relevant occupational exposure questions, flagging known high-risk industries, or auto-populating routine history sections, thereby reducing physician time spent on data entry and recall. However, the core task of clinical assessment remains physician-driven. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI scribes and structured intake tools meaningfully speed up documentation and can prompt for occupational/environmental risk factors, aiding physicians while they remain responsible for accuracy and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Documenting patient history requires synthesis of complex occupational and environmental information with clinical judgment about relevance, which AI systems struggle with reliably. Current AI can draft outlines or suggest relevant questions, but cannot independently assess risk significance or validate accuracy against patient context without substantial human oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft or summarize histories from structured data, but eliciting nuanced occupational/environmental exposure histories requires interactive clinical judgment and probing that current systems cannot fully replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Legal and liability barriers are high: preventive medicine physicians are responsible for the medical-legal adequacy of documented histories, and errors in occupational risk assessment could expose patients to missed diagnoses. Regulatory and professional standards require physician attestation and judgment, not automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Physician licensure, liability for missed occupational exposures, and legal requirements for clinician-authored/verified medical records create strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted documentation tools reduce typing time modestly but require physician validation and specialized occupational health knowledge integration, making the total cost (tool + oversight + training) still substantial relative to the time saved compared to physician wage. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI scribing tools reduce documentation time but still require physician oversight, review, and correction, so all-in costs remain closer to human cost than an order-of-magnitude cheaper alternative. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While EHR systems exist with templated history fields and some vendors offer AI-assisted documentation, no deployed product reliably captures nuanced occupational/environmental risk assessment autonomously. Products require heavy physician review and correction, particularly for novel exposure scenarios or complex employment histories. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Ambient scribes and EHR summarization tools exist in production, but reliable extraction and review of occupational/environmental risk-specific history remains narrow and error-prone in deployed products. |
Perform epidemiological investigations of acute and chronic diseases.
25CI 25–25 · exposure 25 · augmentation 63 · importance 4.2/5 · click for rater detail
Perform epidemiological investigations of acute and chronic diseases.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Public health organizations and epidemiologists have been slow to adopt end-to-end AI automation, relying instead on traditional investigative methods and selective tool-use for data management. Adoption remains pilot-stage despite increasing data availability; structural and expertise factors limit velocity. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Public health and government agencies are historically slow adopters of AI tools relative to finance or tech, with pilots more common than production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist epidemiologists by automating literature synthesis, analyzing case data, generating hypotheses, and modeling transmission patterns, raising their analytical productivity. However, augmentation is concentrated in the analytical phase; core field investigation and judgment remain largely human-driven. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids in analyzing large datasets, detecting outbreak clusters, and drafting reports, meaningfully increasing investigator productivity while the physician remains the decision-maker. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Epidemiological investigations require complex field work, qualitative judgment about disease patterns, and human interviews—tasks where AI struggles with real-world variability and contextual reasoning. While AI can assist with data analysis and literature review, the core investigative activities (site visits, exposure assessment, interviewing patients/contacts) remain human-dependent. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can assist with data analysis, literature review, and pattern detection within an investigation, but designing and executing the full epidemiological investigation (fieldwork, hypothesis generation, causal reasoning, stakeholder coordination) requires human judgment that current systems cannot replace end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong professional and legal barriers exist: epidemiological investigations of disease outbreaks often fall under public health authority mandates requiring qualified physicians or epidemiologists; findings inform regulatory and public health action that demand credentialed sign-off. Liability and institutional requirements protect human oversight. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Epidemiological investigations often require licensed physicians/public health officials for legal reporting, interpretation, and decision-making, plus regulatory and institutional review requirements that limit full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | An epidemiologist's labor-intensive field work, expert judgment, and required travel make this expensive to perform. AI can reduce some analytical overhead, but cannot replace the on-site investigation, interviewing, and sampling expertise, keeping the full-task cost ratio unfavorable compared to human execution. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply process surveillance data, but the overall investigation still requires expensive expert time for design, verification, and field validation, making all-in cost savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed system reliably performs end-to-end epidemiological investigations independently. AI tools exist for data analysis and case identification but lack the field investigation, sampling strategy design, and adaptive hypothesis-testing that define the discipline; published studies remain research-stage for automation. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools (e.g., outbreak detection algorithms, statistical software with AI-assisted modeling) are deployed for surveillance and signal detection, but no product performs full epidemiological investigations reliably in production without heavy human oversight. |
Evaluate the effectiveness of prescribed risk reduction measures or other interventions.
25CI 25–25 · exposure 25 · augmentation 75 · importance 4.2/5 · click for rater detail
Evaluate the effectiveness of prescribed risk reduction measures or other interventions.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Preventive medicine relies on longitudinal patient relationships and complex risk modeling; adoption of AI for effectiveness evaluation remains slow, mostly limited to academic medical centers and large integrated systems, with widespread clinical deployment still nascent. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare overall adopts AI more cautiously than digital-native sectors; preventive medicine's evaluative and regulatory-heavy tasks see slower deployment of autonomous systems. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can substantially augment physician productivity by automating data aggregation, flagging outcome trends, and generating evidence-based comparisons, allowing physicians to focus on interpretation, patient communication, and individualized decision-making while staying fully in control. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered analytics, literature synthesis, and predictive modeling substantially help physicians assess intervention outcomes and identify patterns, meaningfully boosting their evaluative efficiency. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist in analyzing clinical outcome data and comparing intervention metrics against population baselines, but evaluating effectiveness requires synthesizing patient context, comorbidities, behavioral factors, and longitudinal follow-up that demands physician judgment and patient interaction. Current systems cannot reliably perform this holistic assessment end-to-end at the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help analyze outcome data and literature but judging the true effectiveness of an intervention for a specific patient population requires clinical judgment, contextual knowledge, and accountability that current AI cannot fully replicate end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Medical licensing laws and professional liability frameworks require a licensed physician to evaluate clinical effectiveness and take responsibility for intervention recommendations. Regulatory oversight (FDA, clinical trial standards) also constrains automated decision-making in this domain. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Clinical evaluation of interventions typically requires a licensed physician's judgment and sign-off, especially given liability and patient safety concerns, creating strong regulatory and professional barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI analytics and data infrastructure have meaningful upfront costs and require continuous oversight by physicians to validate results and clinical relevance. The all-in cost per evaluation remains comparable to or exceeds the cost of physician time for this task. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Analytics tools reduce some data-crunching cost, but the overall evaluation still requires expensive physician time for interpretation and liability, keeping cost savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While EHR-integrated analytics and outcome dashboards exist in some healthcare systems, deployed AI products do not reliably evaluate intervention effectiveness across diverse clinical scenarios without substantial physician oversight and validation. Most systems are in pilot or prototype phase rather than mature production deployment. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Clinical decision support and analytics tools exist that surface trends and flag outcomes, but no deployed product autonomously evaluates intervention effectiveness reliably in production without physician oversight. |
Provide information about potential health hazards and possible interventions to the media, the public, other health care professionals, or local, state, and federal health authorities.
25CI 25–25 · exposure 25 · augmentation 75 · importance 4.1/5 · click for rater detail
Provide information about potential health hazards and possible interventions to the media, the public, other health care professionals, or local, state, and federal health authorities.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for autonomous health hazard communication to authorities is minimal; public health agencies and physicians remain conservative, and regulatory environments actively discourage uncredentialed entity claims about health risks. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Public health and clinical medicine sectors have been slower and more cautious in adopting AI for external-facing authoritative communications compared to fast-moving information/finance sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can substantially assist by drafting hazard summaries, synthesizing epidemiological data, generating talking points, and organizing evidence—transforming physician productivity in preparing communications while the physician retains judgment and accountability. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by drafting explanatory materials, summarizing research, and tailoring messages for different audiences, significantly speeding up the physician's communication tasks while they retain final oversight. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can draft health communications and summarize hazard information, this task fundamentally requires human judgment about messaging context, audience needs, and responsibility for health claims. The liability and accuracy sensitivity make autonomous end-to-end performance infeasible; human oversight remains essential. |
| Task automatability | claude-sonnet-5 | 2/5 | Drafting communications about health hazards can be partially automated, but synthesizing current epidemiological data, exercising judgment about public messaging, and engaging with press or authorities requires human expertise and accountability that AI cannot fully replace end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Legal and regulatory barriers are substantial: health communications to authorities and media carry liability, require professional credibility and licensure, and often necessitate physician sign-off. Organizational culture in public health also favors credentialed professionals for official guidance. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Official health communications typically require credentialed physicians or public health officials to vet and authorize messaging, given liability, regulatory reporting duties, and public trust considerations. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The human cost of a preventive medicine physician is high (~$200k+ loaded), and AI tools for communication drafting are relatively inexpensive, but the task requires physician review and accountability, so the all-in cost ratio remains unfavorable for pure AI execution. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can cheaply generate draft text, the physician's expert review, liability, and authoritative sign-off remain necessary, so overall cost savings are modest rather than order-of-magnitude. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products perform this task end-to-end reliably. AI tools exist for drafting communications and information synthesis, but they lack the medical authority, contextual judgment, and accountability required to directly inform authorities or the public about health hazards without physician review. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI writing tools can help draft public health messaging, but no deployed product autonomously communicates health hazard information to media/authorities in production without physician oversight and validation. |
Direct public health education programs dealing with topics such as preventable diseases, injuries, nutrition, food service sanitation, water supply safety, sewage and waste disposal, insect control, and immunizations.
23CI 20–25 · exposure 20 · augmentation 63 · importance 4.2/5 · click for rater detail
Direct public health education programs dealing with topics such as preventable diseases, injuries, nutrition, food service sanitation, water supply safety, sewage and waste disposal, insect control, and immunizations.
23| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Public health agencies adopt AI for content support and data analysis, but program direction remains human-led; the sector is slow to fully automate physician-directed public health functions due to regulatory requirements and the need for human credibility in community health messaging. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Public health agencies, especially government and municipal bodies, are typically slow adopters of AI compared to private-sector information industries. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist physicians by generating evidence-based content drafts, identifying at-risk populations through data analysis, and suggesting program frameworks, thereby raising productivity in research and material preparation while the physician directs strategy and community engagement. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly help physicians draft educational materials, analyze epidemiological data, and generate communication content, meaningfully boosting productivity while the physician retains directive control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can help generate educational content and materials on public health topics, this task requires designing comprehensive programs tailored to specific populations, determining messaging strategy, selecting delivery channels, and building community trust—activities requiring human judgment and oversight that prevent end-to-end automation with 50% time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | Directing a public health education program requires strategic planning, stakeholder coordination, community engagement, and adaptive decision-making that current AI cannot perform end-to-end, though AI can assist with content drafting and data analysis components. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Directing public health programs typically requires licensed medical credentials and legal/regulatory authority to certify program safety and validity; liability for health misguidance creates high error costs, and public health authorities require human physician accountability. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Public health leadership roles often require licensed physician oversight, regulatory accountability, and legal responsibility for public safety recommendations, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI tools reduce content creation costs but do not substantially lower the labor costs of program design, community engagement, and oversight by qualified physicians or public health professionals, making the cost ratio roughly comparable or slightly favorable but not decisively cheaper. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can cheaply generate educational materials, the actual directing role (program management, partnerships, community outreach) still requires substantial human labor, keeping overall cost savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI systems can draft educational materials and suggest program structures, but no deployed product reliably performs the full task of directing a public health education program independently; production systems lack the strategic decision-making and stakeholder engagement coordination required. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously directs public health education programs; this remains a human leadership and coordination function with AI only used as a supporting tool. |
Design, implement, or evaluate health service delivery systems to improve the health of targeted populations.
16CI 7–25 · exposure 13 · augmentation 63 · importance 3.8/5 · click for rater detail
Design, implement, or evaluate health service delivery systems to improve the health of targeted populations.
16| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare systems are slow to adopt AI for strategic planning and system redesign, with most adoption concentrated in narrow diagnostics or administrative tasks. Population health system innovation remains largely human-driven and resistant to algorithmic substitution due to accountability and regulatory requirements. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Public health and health administration sectors are slower adopters of AI for systemic design tasks compared to clinical documentation or diagnostics, with pilots more common than production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist with data aggregation, population health analytics, identifying evidence-based practices, and simulation of system designs, allowing physicians to make more informed strategic decisions. However, the core task of negotiating implementation, managing stakeholders, and ensuring accountability remains fundamentally human. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist with epidemiological data analysis, predictive modeling, literature synthesis, and drafting evaluation reports, enhancing physician productivity in designing and assessing health systems. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Designing, implementing, and evaluating health service delivery systems requires extensive human judgment, stakeholder engagement, regulatory navigation, and contextual understanding of population health needs. Current AI cannot end-to-end execute this complex strategic task with 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | This task requires systems-level design, stakeholder negotiation, policy judgment, and contextual evaluation that current AI cannot perform end-to-end; AI can assist with data analysis but not the full design/implementation/evaluation cycle. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Health service system design and implementation are heavily regulated, require physician oversight and licensure, involve liability for population health outcomes, and demand direct engagement with institutional leadership, clinical staff, and communities. Legal and professional responsibility strongly protect this role. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Health system design decisions often require licensed physician oversight, regulatory compliance, and accountability for population health outcomes, creating strong professional and liability barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The AI systems that could contribute to this task (healthcare analytics platforms, consultative AI) are expensive relative to the labor they replace, and substantial physician time remains required for decision-making, stakeholder engagement, and accountability. Overall cost remains higher than pure automation would suggest. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools can cheaply generate data summaries or models, but the overall task still requires expensive expert oversight, stakeholder engagement, and implementation work, so cost savings are limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with data analysis and literature review for system design, no deployed product reliably performs the full system design, implementation, and evaluation workflow. Existing tools support components (predictive analytics, workflow optimization) but not the integrated strategic planning and stakeholder coordination this task demands. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously designs or evaluates health service delivery systems; this remains a human strategic and administrative function with AI only used for supporting analytics. |
Develop or implement interventions to address behavioral causes of diseases.
9CI 3–16 · exposure 5 · augmentation 63 · importance 3.6/5 · click for rater detail
Develop or implement interventions to address behavioral causes of diseases.
9| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare sectors show slower AI adoption compared to information and finance, with behavioral medicine adoption particularly cautious due to licensure, liability, and patient autonomy concerns. Pilots exist but production-level displacement of physician-led intervention design remains limited. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially public health and preventive medicine, adopts AI slowly for clinical decision-making tasks, with pilots more common than production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by synthesizing behavioral health literature, suggesting evidence-based frameworks, and tracking intervention outcomes, raising physician productivity in research and planning phases. However, the core task of tailoring and implementing interventions remains primarily physician-driven despite these assistive tools. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by synthesizing epidemiological data, generating patient education content, and suggesting evidence-based intervention strategies, boosting physician productivity. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Developing or implementing behavioral health interventions requires nuanced clinical judgment, understanding of patient context, and ethical decision-making tailored to individual circumstances. Current AI cannot autonomously design treatment plans or implement interventions that meet clinical standards for patient safety and efficacy. |
| Task automatability | claude-sonnet-5 | 1/5 | Designing behavioral health interventions requires clinical judgment, patient rapport, contextual understanding of community/lifestyle factors, and iterative program design that current AI cannot execute end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Preventive medicine physicians must be licensed clinicians, and behavioral intervention design and implementation are legally and ethically restricted to credentialed practitioners. Medical liability, patient safety requirements, and regulatory oversight create hard barriers to full automation or unsupervised AI deployment. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Physician licensure, liability for clinical recommendations, and the need for professional judgment in designing health interventions create strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The cost of developing clinically sound, evidence-based behavioral interventions with AI oversight currently exceeds the cost of a physician doing so directly, particularly given validation and liability requirements. Integration costs and necessary human oversight make AI more expensive than direct physician work today. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can generate low-cost drafts of educational materials or program outlines, but the physician's design, implementation, and oversight still dominate the cost, limiting overall savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI tools can assist with evidence synthesis and suggest intervention frameworks, no deployed system reliably performs end-to-end development and implementation of behavioral interventions. Behavioral medicine requires clinical licensure, patient assessment, and accountability that AI systems do not yet provide in production healthcare settings. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently develops or implements clinical behavioral interventions; AI is at most a research or decision-support aid, not an autonomous practitioner. |
Supervise or coordinate the work of physicians, nurses, statisticians, or other professional staff members.
4CI 0–7 · exposure 0 · augmentation 38 · importance 4.2/5 · click for rater detail
Supervise or coordinate the work of physicians, nurses, statisticians, or other professional staff members.
4| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare organizations have not adopted AI to supervise or coordinate professional medical staff; supervision remains a distinctly human leadership function with no evidence of meaningful AI displacement in this domain. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare administration adoption of AI for personnel management/coordination is nascent, with most AI use focused on clinical decision support rather than staff supervision. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI might assist with scheduling, performance metrics, or administrative tracking, but augmentation is limited because the core supervisory function—accountability, mentoring, and personnel judgment—must remain human-led and cannot be substantially enhanced by current AI tools. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI tools can assist with scheduling, performance data aggregation, and communication drafting, moderately supporting the supervisory function without replacing judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Supervising and coordinating professional staff requires human judgment, relationship management, conflict resolution, and accountability for personnel decisions. Current AI systems cannot meaningfully replace this interpersonal and organizational leadership function. |
| Task automatability | claude-sonnet-5 | 1/5 | Supervising and coordinating diverse professional staff requires interpersonal leadership, performance management, and real-time decision-making that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Healthcare leadership roles are legally bound to human responsibility; supervisory authority over licensed professionals (physicians, nurses) requires licensed management oversight and carries full liability for team performance and decisions. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Supervisory and coordination responsibilities often carry legal accountability (e.g., physician oversight requirements, licensure, liability) that require a qualified human to hold authority. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI cannot yet take over managerial responsibilities, so the cost comparison is not applicable; a human supervisor remains mandatory and AI offers no viable cost reduction for this core function. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this supervisory role, so cost comparison favors the human by default. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product reliably performs end-to-end supervision and coordination of professional teams in healthcare settings. This requires legal accountability and real-time adaptive management that exceeds current AI capabilities. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product manages or supervises multidisciplinary clinical/professional teams autonomously; this remains a human leadership function. |
Coordinate or integrate the resources of health care institutions, social service agencies, public safety workers, or other organizations to improve community health.
4CI 0–7 · exposure 0 · augmentation 38 · importance 3.9/5 · click for rater detail
Coordinate or integrate the resources of health care institutions, social service agencies, public safety workers, or other organizations to improve community health.
4| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare institutions remain largely hierarchical and relationship-driven; adoption of AI for cross-organizational strategy and governance is minimal even in digitized settings. Coordination decisions are made by senior physicians and administrators, not delegated to automated systems. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Public health and social service sectors are generally slower AI adopters, with fragmented systems and limited interoperability, apart from isolated data-sharing pilots. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with data synthesis, identifying gaps in service coverage, or scheduling meetings, but these supports address only the information-gathering periphery. The core task—persuading organizations to reallocate resources—remains fundamentally human and offers limited augmentation potential. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help by aggregating data, drafting communications, tracking cross-agency initiatives, and identifying resource gaps, meaningfully assisting the physician who still leads coordination. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires high-level stakeholder negotiation, strategic judgment, and institutional coordination across heterogeneous organizations with conflicting incentives. Current AI systems cannot independently identify priorities, broker agreements, or navigate organizational politics—the human judgment and relationship-building are central to success. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires cross-organizational relationship-building, negotiation, and judgment across institutions with different mandates and cultures, which current AI cannot execute end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Preventive medicine physicians hold professional credentials and legal authority to coordinate clinical and public health resources. Hospitals, public health agencies, and social services typically require licensed medical professionals to sign off on major coordination initiatives, creating a formal licensing and accountability barrier. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Coordination often requires accountable individuals with institutional authority, trust, and sometimes statutory public health roles, creating strong organizational and legal barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The cost of deploying oversight systems and human review to validate AI-driven coordination would likely exceed the cost of a physician directly managing these relationships, given the high stakes and low error tolerance in institutional coordination. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the human coordinator role itself, so any comparison favors the human whose judgment, authority, and relationships are essential; AI tools only add marginal cost as support. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI products perform end-to-end institutional coordination and resource integration. While AI can assist with data aggregation or meeting scheduling, the core task of synthesizing diverse organizational goals and securing cross-sector commitment requires human leadership and decision-making. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs multi-agency coordination and integration of health, social service, and public safety resources; this remains a human relational and administrative function. |
Direct or manage prevention programs in specialty areas such as aerospace, occupational, infectious disease, and environmental medicine.
1CI 0–3 · exposure 0 · augmentation 38 · importance 4.6/5 · click for rater detail
Direct or manage prevention programs in specialty areas such as aerospace, occupational, infectious disease, and environmental medicine.
1| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare and occupational medicine sectors show slow adoption of autonomous AI in management and clinical oversight roles due to regulatory constraints, liability concerns, and institutional reliance on credentialed physicians in decision-making authority. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Public health and occupational medicine settings are slow adopters of AI for managerial/leadership functions, with pilots limited mostly to data analytics rather than program direction. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI can assist with data gathering, literature synthesis, and program analytics, but the core functions of directing strategy, managing personnel, and making program-level decisions remain inherently human and offer limited scope for transformative augmentation. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with epidemiological data analysis, report drafting, and monitoring dashboards, aiding the physician's decision-making without replacing the directive role. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Directing and managing prevention programs requires strategic decision-making, stakeholder engagement, regulatory interpretation, and adaptive program design based on evolving organizational and epidemiological contexts—functions that demand human judgment and accountability. Current AI lacks the authority, domain-specific contextual reasoning, and ability to assume responsibility for programmatic outcomes. |
| Task automatability | claude-sonnet-5 | 1/5 | Directing and managing a specialty prevention program requires strategic planning, stakeholder coordination, regulatory judgment, and accountability that current AI cannot perform end-to-end.rating.rating |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Directing prevention programs in regulated specialty areas (aerospace, occupational, infectious disease) typically requires a licensed physician to authorize, sign off on, and assume liability for program decisions. Legal and professional licensing frameworks mandate human oversight and responsibility. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This role requires a licensed physician with specialty board credentials and legal accountability for program oversight, creating hard regulatory and liability barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI cannot perform the core task of directing prevention programs, so cost comparison is not meaningful; the task requires a physician's salary and benefits. Integration of AI assistance tools would supplement rather than replace this role. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this managerial/directive role, so cost comparison favors the human physician who provides the actual service. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system can independently direct or manage comprehensive prevention programs; these tasks require licensed medical professionals to interpret regulations, design interventions, oversee staff, and make high-stakes decisions in specialty domains. AI may assist with data analysis or documentation, but the core management function remains non-delegable. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product manages or directs a medical prevention program autonomously; existing tools only support narrow analytic or documentation subtasks. |
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.