Physicians, Pathologists

29-1222.00

Median wage $312,400/yr11,110 employed (US)Rank #662 of 923 scored · top 72% by substitution

Diagnose diseases and conduct lab tests using organs, body tissues, and fluids. Includes medical examiners.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution20

Exposure23

Augmentation63

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

19 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

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%22

panel mean rating 1.9/5 → substitution pressure 22/100

Technical feasibility todayw 20%25

panel mean rating 2.0/5 → substitution pressure 25/100

Cost vs. human wagew 15%21

panel mean rating 1.8/5 → substitution pressure 21/100

Adoption barriersw 20%inverted — strong barriers lower the score11

panel mean rating 4.6/5 (barrier strength) → substitution pressure 11/100

Sector adoption velocityw 10%24

panel mean rating 2.0/5 → substitution pressure 24/100

Task breakdown (19 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.

Diagnose infections, such as Hepatitis B and Acquired Immune Deficiency Syndrome (AIDS), by conducting tests to detect the antibodies that patients' immune systems make to fight such infections.

CI 20–49 · exposure 38 · augmentation 63 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Large clinical laboratories and hospital systems have adopted automated immunoassay platforms and begun piloting AI-assisted interpretation, but most diagnostic decisions still require pathologist review. Adoption is uneven—faster in high-volume settings, slower in smaller labs.
Sector adoption velocity	claude-sonnet-5	2/5	Healthcare/pathology adoption of AI diagnostic tools is proceeding cautiously due to regulatory approval processes, liability concerns, and validation requirements, resulting in slow production deployment despite active piloting.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI significantly augments pathologist productivity by rapidly processing immunoassay data, flagging patterns, and prioritizing results, allowing more tests to be reviewed in less time while the pathologist remains the final decision-maker. This is a strong example of assistive augmentation in routine diagnosis.
Augmentation potential	claude-sonnet-5	3/5	AI can help triage test results, flag likely positives for follow-up, and support differential diagnosis, meaningfully aiding physicians while they retain full diagnostic responsibility.
Task automatability	claude-haiku-4-5-20251001	3/5	AI can automate parts of the diagnostic workflow—test ordering, data interpretation of lab results, and flagging abnormal antibody patterns—but clinical judgment about patient context, symptom correlation, and confirmatory testing decisions requires human pathologist oversight. End-to-end automation without human review falls short of the 50% time-saving threshold at equal quality.
Task automatability	claude-sonnet-5	2/5	AI can assist in interpreting lab assay results and flagging abnormal antibody titers, but ordering, contextualizing with clinical history, and finalizing diagnosis require physician judgment not yet fully automatable end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Diagnosis and test interpretation must be signed by a licensed pathologist or physician in most jurisdictions; regulatory bodies (CLIA, FDA) require human accountability for reportable infectious disease diagnoses. Liability for missed infections creates strong legal and professional barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Diagnosing infectious diseases like HIV/Hepatitis B legally requires a licensed physician's sign-off, with significant liability and regulatory requirements around communicable disease diagnosis and reporting.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	Automated immunoassay platforms and AI-assisted result interpretation are significantly cheaper than dedicated pathologist time per test case, particularly at scale. However, some human oversight and confirmatory testing costs remain, preventing a full order-of-magnitude advantage.
Cost vs. human wage	claude-sonnet-5	2/5	Lab testing itself is already automated and cheap, but the diagnostic interpretation and physician oversight component still requires costly clinical expertise, keeping overall cost comparable to human-led workflows.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Laboratory automation and AI-assisted immunoassay interpretation products exist in pathology workflows, but they operate as decision-support tools with human verification. No mature AI system independently diagnoses infections from antibody tests without pathologist sign-off in production settings.
Technical feasibility today	claude-sonnet-5	2/5	Deployed lab automation systems and immunoassay analyzers handle testing, but AI-driven diagnostic interpretation of infection status is largely research-stage or narrow decision-support, not autonomous diagnosis in production.

Write pathology reports summarizing analyses, results, and conclusions.

CI 29–39 · exposure 45 · augmentation 75 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption remains limited; most pathology labs use AI as an assistive tool rather than autonomous report generation. Pilot programs are common but production deployment of fully autonomous report writing is rare due to liability, validation burden, and physician skepticism.
Sector adoption velocity	claude-sonnet-5	2/5	Healthcare and pathology are historically slow to adopt AI tools for documentation compared to information/finance sectors, with pilots more common than production deployment.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI augments pathologist productivity substantially: image analysis recommendations, structured data extraction, and draft report generation reduce clerical and routine cognitive load, enabling pathologists to focus on complex interpretation and sign-off.
Augmentation potential	claude-sonnet-5	4/5	AI-assisted structured reporting, template population, and language generation from lab data can meaningfully speed up report drafting while the pathologist remains the diagnostic authority.
Task automatability	claude-haiku-4-5-20251001	3/5	AI can generate substantial portions of pathology reports, including structured findings and standard conclusions from image analysis and lab data, but requires pathologist review and sign-off for accuracy, interpretation nuance, and legal liability. This falls short of full end-to-end automation with 50% time savings at equal quality.
Task automatability	claude-sonnet-5	3/5	AI can draft structured report summaries from templated findings, but pathologists must verify diagnostic accuracy and finalize interpretation, so only partial time savings are achievable end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Strong barriers exist: pathology reports are legally mandated outputs signed by a licensed physician, with significant malpractice liability if inaccurate. Regulatory bodies (CAP, CLIA) require physician oversight and sign-off, and clinical use requires documented validation.
Adoption barriers	claude-sonnet-5	5/5	Pathology reports are legal medical documents requiring a licensed physician's diagnostic interpretation and signature, creating a hard regulatory and liability barrier to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI inference and oversight integration cost is approaching but not yet substantially below the loaded wage of a pathologist (typically $200k–$250k+/year); the ratio remains near parity when full validation and correction workflows are factored in.
Cost vs. human wage	claude-sonnet-5	2/5	AI drafting assistance is cheap per query, but integration with LIS/EHR systems, validation, and mandatory physician review keep overall cost closer to human-comparable rather than order-of-magnitude cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Products exist (AI-assisted pathology platforms, LLM-based report generation) and demonstrate utility in production settings, but material gaps remain in complex cases, rare diagnoses, and integration with existing lab information systems. Error rates in critical interpretations exceed clinical acceptance thresholds.
Technical feasibility today	claude-sonnet-5	2/5	Some clinical NLP and structured reporting tools assist with report generation, but no widely deployed product autonomously writes final pathology reports without physician drafting/review at scale.

Analyze and interpret results from tests, such as microbial or parasite tests, urine analyses, hormonal assays, fine needle aspirations (FNAs), and polymerase chain reactions (PCRs).

CI 20–43 · exposure 38 · augmentation 75 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption of AI for autonomous test interpretation in pathology is still largely in pilot phase at major medical centers and reference labs. Most deployments remain assistive rather than replacement-level, and smaller clinical labs lag significantly in digitization and AI integration compared to information/finance sectors.
Sector adoption velocity	claude-sonnet-5	2/5	Healthcare and diagnostic pathology are historically slow adopters of AI due to regulatory hurdles, liability concerns, and validation requirements, with adoption largely limited to pilot programs and narrow FDA-cleared tools.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI demonstrably assists pathologists by highlighting abnormalities, auto-flagging critical values, drafting preliminary reports, and accelerating routine case review. These tools substantially raise review speed and reduce cognitive burden while the pathologist retains final judgment, making augmentation high even where full automation is not yet reliable.
Augmentation potential	claude-sonnet-5	4/5	AI substantially assists pathologists via image analysis, quantification, triage, and flagging abnormalities in digital pathology and some lab tests, improving speed and consistency while the physician retains final interpretive authority.
Task automatability	claude-haiku-4-5-20251001	3/5	AI can automate significant portions of routine test interpretation (e.g., flagging abnormal values, pattern recognition in microbial cultures, standard urine dipstick analysis) and achieve time savings on templated reports. However, complex cases requiring integration of clinical context, rare organism identification, or subtle morphological judgment in FNAs still require human expertise, limiting full automation to roughly half the task.
Task automatability	claude-sonnet-5	2/5	Some AI/ML tools can assist with pattern recognition in specific test types (e.g., digital pathology image analysis), but comprehensive interpretation across diverse test modalities requiring clinical correlation remains beyond current AI's end-to-end capability.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Pathologists are licensed professionals whose interpretations carry legal and clinical liability; most laboratories require a licensed pathologist to sign off on critical results. Regulatory bodies (CAP, CLIA) impose quality and oversight requirements that slow substitution of AI-only interpretation, and error costs in diagnostic medicine are high.
Adoption barriers	claude-sonnet-5	5/5	Pathology diagnosis is a licensed medical act with strict legal requirements for physician sign-off, malpractice liability, and regulatory oversight (CLIA, FDA), making autonomous AI interpretation currently prohibited without human authorization.
Cost vs. human wage	claude-haiku-4-5-20251001	3/5	AI inference and integration costs for lab automation are comparable to the labor cost of a pathologist reviewing routine tests, but human oversight, quality assurance, and system maintenance partially offset savings. Large-scale deployment in high-volume labs may achieve modest savings, but cost per-test is roughly equivalent to human review.
Cost vs. human wage	claude-sonnet-5	2/5	Specialized diagnostic AI systems require significant capital investment, integration, and mandatory pathologist oversight, making all-in costs still comparable to or higher than physician time for many use cases.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed products exist for specific narrower subtasks (e.g., digital pathology image analysis, automated urinalysis flagging, PCR result interpretation), but end-to-end reliable automation across the full range of test types remains limited. Error rates and liability concerns keep most systems in a supportive role rather than autonomous decision-making in production pathology labs.
Technical feasibility today	claude-sonnet-5	2/5	FDA-cleared AI products exist for specific narrow applications (e.g., certain cytology screening, some digital pathology aids) but are adjunctive tools, not autonomous interpreters, and lack broad deployment across the full range of tests listed.

Read current literature, talk with colleagues, or participate in professional organizations or conferences to keep abreast of developments in pathology.

CI 19–41 · exposure 17 · augmentation 75 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Many medical institutions have adopted literature-monitoring tools and AI-powered abstracting systems as pilot or supplemental aids, but actual displacement of the pathologist's personal knowledge maintenance remains limited; adoption is primarily augmentative rather than replacement-driven.
Sector adoption velocity	claude-sonnet-5	3/5	Physicians increasingly use AI-powered literature tools (e.g., summarization, alerts) but adoption for staying current is still supplementary rather than transformative across the field.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI excels at surfacing relevant new papers, generating summaries, and flagging key findings—substantially reducing the time pathologists spend screening literature and allowing them to focus on deeper synthesis and clinical application of new knowledge.
Augmentation potential	claude-sonnet-5	4/5	AI literature summarization, alerting services, and conference content curation meaningfully speed up how pathologists stay current, even though human engagement remains central.
Task automatability	claude-haiku-4-5-20251001	1/5	Reading literature and staying current in a field requires genuine comprehension, synthesis, and judgment about relevance—tasks where AI provides modest support but cannot autonomously maintain professional knowledge currency at the standard required of a pathologist.
Task automatability	claude-sonnet-5	2/5	AI can help surface and summarize literature, but the full task includes networking, colleague discussion, and conference participation that require human presence and judgment, limiting end-to-end automation.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Professional competence and licensure requirements in medicine implicitly mandate that physicians personally maintain current knowledge; professional standards, continuing education requirements, and liability frameworks expect the human to be knowledgeable, creating strong organizational and regulatory barriers to full substitution.
Adoption barriers	claude-sonnet-5	2/5	No licensing requirement mandates this specific task be done a certain way, but professional norms, board certification maintenance requirements, and value of peer interaction create some friction against full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	3/5	AI literature summarization and alerting services are relatively low-cost, but pathologists' time spent discussing findings with colleagues or reflecting on implications is still more economical to perform directly than to automate with meaningful oversight.
Cost vs. human wage	claude-sonnet-5	2/5	AI tools for literature scanning are cheap, but the task as a whole includes professional networking and conference attendance that AI cannot substitute for, so overall cost savings are limited.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can summarize papers and flag recent publications, no deployed product reliably curates a personalized, clinically relevant knowledge update stream that meets a pathologist's professional development standard; systems exist but with significant scope and relevance limitations.
Technical feasibility today	claude-sonnet-5	2/5	AI literature summarization and alerting tools exist and are used by clinicians, but no product autonomously performs the full continuing-education/networking task reliably in practice.

Examine microscopic samples to identify diseases or other abnormalities.

CI 20–39 · exposure 38 · augmentation 75 · importance 4.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Digital pathology adoption remains uneven across U.S. labs, particularly outside large academic and commercial centers; while pilots are common, production deployment with AI triage is still nascent compared to adoption in lower-regulation digital sectors.
Sector adoption velocity	claude-sonnet-5	2/5	Digital pathology adoption is growing but still nascent overall, with most labs still using traditional microscopy and AI tools deployed only in leading academic or specialty centers.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI pathology assistants demonstrably enhance pathologist productivity by pre-screening slides, highlighting regions of interest, and reducing time to diagnosis, allowing pathologists to handle higher volumes while maintaining oversight and judgment on complex cases.
Augmentation potential	claude-sonnet-5	4/5	AI-assisted image analysis meaningfully speeds up screening, quantification, and detection of subtle features (e.g., cancer cells, biomarkers), helping pathologists work faster and more consistently while retaining final diagnostic authority.
Task automatability	claude-haiku-4-5-20251001	3/5	AI-powered digital pathology systems can identify many diseases in histopathology images with high accuracy, potentially automating large portions of sample screening and triage. However, complex cases, rare conditions, and integration with clinical context still require pathologist expertise, preventing full end-to-end automation at the ≥50% time-saving threshold for all work.
Task automatability	claude-sonnet-5	2/5	AI image analysis can flag abnormalities in some tissue types, but comprehensive diagnostic interpretation across the full range of pathology specimens still requires expert human judgment and integration with clinical context, so the ≥50% time-saving-at-equal-quality bar is not met end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Pathology diagnoses carry significant medicolegal weight and liability; regulators (FDA, CAP, CLIA) require credentialed pathologists to validate findings and sign reports, creating a hard requirement for human sign-off that prevents full substitution regardless of AI capability.
Adoption barriers	claude-sonnet-5	5/5	Diagnosis of disease from tissue samples is a licensed medical act requiring a board-certified pathologist's sign-off, with high liability exposure and strict regulatory oversight (CLIA, FDA) preventing full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	While AI inference costs are low, the full system cost including validation infrastructure, integration with lab information systems, ongoing oversight, and liability management approaches human pathologist wages, especially when accounting for the need to retain expert review capacity.
Cost vs. human wage	claude-sonnet-5	2/5	Digital pathology AI requires expensive slide scanners, software licensing, and validation infrastructure, plus mandatory pathologist review, so total cost per case is not dramatically lower than physician time alone yet.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed AI products (e.g., Paige, Proscia, PathAI) demonstrate reliable detection of common cancers and some other pathologies in production settings, but they typically flag or score regions rather than fully autonomous diagnosis, and error rates remain material enough that pathologist review remains standard practice.
Technical feasibility today	claude-sonnet-5	2/5	FDA-cleared digital pathology AI tools exist for narrow applications (e.g., prostate cancer detection, mitosis counting) but are used as adjuncts in limited settings, not as reliable standalone diagnostic systems at scale.

Conduct genetic analyses of deoxyribonucleic acid (DNA) or chromosomes to diagnose small biopsies and cell samples.

CI 20–34 · exposure 38 · augmentation 75 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Pathology labs are adopting AI-assisted image and data analysis incrementally, but adoption remains cautious due to regulatory hurdles, high stakes of diagnostic error, and reliance on credentialed expertise. Penetration in production workflows is still limited compared to lower-barrier information sectors.
Sector adoption velocity	claude-sonnet-5	2/5	Clinical genomics and pathology labs are adopting AI-assisted tools gradually, but adoption is slower than in pure information sectors due to regulatory approval cycles and validation requirements for diagnostic use.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI significantly augments pathologist productivity by automating image preprocessing, highlighting suspicious regions, and pre-screening variants, allowing the expert to focus on complex cases and clinical integration. Tools like digital pathology platforms with AI assistance demonstrably improve throughput and consistency while keeping the pathologist in the loop for final diagnosis.
Augmentation potential	claude-sonnet-5	4/5	AI substantially augments pathologists by accelerating variant annotation, flagging chromosomal abnormalities, and prioritizing regions of interest, meaningfully speeding up analysis while the pathologist retains diagnostic authority.
Task automatability	claude-haiku-4-5-20251001	3/5	Image analysis of genetic samples can be partially automated using AI for feature detection and preliminary classification, but definitive diagnosis requires integration of complex patient history, morphologic interpretation, and clinical correlation. Current AI can handle routine screening and flagging, achieving roughly 50% time savings on labor-intensive parts, but full end-to-end diagnosis at equal quality remains dependent on expert review.
Task automatability	claude-sonnet-5	2/5	AI can assist in variant calling, sequence alignment, and pattern detection within genomic data, but final diagnostic interpretation integrating clinical context, sample quality assessment, and confirmatory judgment still requires a pathologist and cannot yet be fully automated end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Regulatory and legal barriers are substantial: in most jurisdictions, a licensed pathologist must perform or directly supervise genetic diagnosis and sign off on reports. Clinical Laboratory Improvement Amendments (CLIA) and FDA oversight of genetic tests create hard requirements for professional licensure and liability accountability that prevent full automation.
Adoption barriers	claude-sonnet-5	5/5	Genetic diagnosis for clinical biopsies requires licensed pathologist sign-off, is tightly regulated (CLIA, FDA), and carries high liability for misdiagnosis, making full automation legally and professionally barred.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI inference and integration costs for genetic analysis platforms are falling but remain significant when accounting for validation, oversight by board-certified pathologists, and regulatory compliance infrastructure. The loaded wage of a pathologist is high, and AI has not yet achieved an order of magnitude cost advantage when full workflow costs are included.
Cost vs. human wage	claude-sonnet-5	2/5	Genetic sequencing and bioinformatics pipelines have real infrastructure, reagent, and specialist oversight costs; while software components are cheap to run, the overall diagnostic workflow still requires costly wet-lab and expert validation steps.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed products exist for automated microscopy image analysis and some genetic variant calling in clinical labs, but these operate within narrow scopes (e.g., specific stain types, known variant panels) and still require pathologist review. Material error rates and the need for human sign-off mean production systems are assistive rather than fully autonomous.
Technical feasibility today	claude-sonnet-5	2/5	Some FDA-cleared bioinformatics pipelines and AI-assisted variant classification tools are deployed in genomics labs, but comprehensive automated diagnosis from raw biopsy/cytogenetic samples to final report is not yet standard production practice.

Conduct research and present scientific findings.

CI 25–28 · exposure 25 · augmentation 75 · importance 3.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Academic and medical research sectors have adopted AI incrementally for supporting tasks (data analysis, writing assists), but core research conception and execution remain human-led. Adoption is limited by epistemic requirements and professional norms favoring human originality.
Sector adoption velocity	claude-sonnet-5	3/5	Academic medicine and research institutions have begun adopting AI for literature review, data analysis, and manuscript drafting, but adoption for core research design and presentation remains at the pilot stage.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI substantially augments pathology research through literature mining, statistical analysis, image analysis in pathology datasets, manuscript drafting, and hypothesis generation. Researchers increasingly use AI assistants to accelerate workflows while maintaining critical judgment and experimental oversight.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids literature synthesis, statistical analysis, figure generation, and drafting of manuscripts/presentations, meaningfully boosting researcher productivity while humans retain intellectual control.
Task automatability	claude-haiku-4-5-20251001	2/5	Pathologists conducting original research involves hypothesis formulation, experimental design, literature synthesis, and interpretation of novel findings—creative, domain-specific judgment tasks AI cannot yet perform end-to-end. AI can assist with literature review and data analysis, but cannot independently conceive research questions or validate novel scientific insights at the 50% time-savings threshold.
Task automatability	claude-sonnet-5	2/5	AI can assist with literature review, data analysis, and drafting text, but designing pathology research, generating novel scientific insight, and validating findings requires human expertise that current AI cannot autonomously replace end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Research integrity standards, authorship attribution, institutional review boards, and funding agency requirements create strong organizational and regulatory barriers to AI autonomy. Pathologists must personally take responsibility for research accuracy and intellectual contribution, limiting substitution.
Adoption barriers	claude-sonnet-5	4/5	Scientific research involving patient data and publication requires physician authorship, ethical oversight, IRB approval, and accountability for findings, creating strong professional and regulatory barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	LLM inference is cheap, but the overhead of fact-checking AI outputs, correcting scientific errors, and human oversight of research direction makes the all-in cost competitive with or exceeding the cost of a researcher's time on this task.
Cost vs. human wage	claude-sonnet-5	2/5	AI tools reduce some costs of literature search and drafting, but the core research (experimental design, specimen analysis, interpretation) still requires expensive physician-scientist time, keeping overall cost comparable to human-driven work.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI tools (GPT, ChatGPT) can draft text and summarize literature, no deployed product reliably conducts original scientific research or validates experimental results. Current systems lack the ability to design rigorous experiments, troubleshoot failures, or make novel scientific judgments that would constitute reliable end-to-end performance.
Technical feasibility today	claude-sonnet-5	2/5	Products like AI writing assistants and literature-summarization tools exist and are used by researchers, but no deployed system independently conducts pathology research or presents findings without heavy human oversight.

Educate physicians, students, and other personnel in medical laboratory professions, such as medical technology, cytotechnology, or histotechnology.

CI 25–25 · exposure 25 · augmentation 63 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Medical education is a conservative, credentialed domain where adoption of AI as a replacement for instruction is minimal; most AI use is supplementary (study aids, content generation). Institutions move slowly on curricular changes and are wary of delegating teaching to unproven systems.
Sector adoption velocity	claude-sonnet-5	2/5	Medical education is adopting AI tools like case simulations and quiz generators, but adoption in accredited teaching roles remains slow and pilot-stage.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist pathologists and educators by drafting lecture notes, generating practice questions, providing explanatory diagrams, and summarizing literature—materially reducing preparation time. However, the core teaching act (interaction, feedback, assessment) remains human-driven, limiting the productivity multiplier.
Augmentation potential	claude-sonnet-5	4/5	AI significantly aids in creating teaching slides, case studies, quizzes, and explanations, meaningfully boosting educator productivity while the physician retains oversight of accuracy and pedagogy.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can generate educational content and draft explanations of pathology concepts, medical education requires interactive feedback, real-time question answering, and adaptive teaching tailored to learner level—capabilities current systems struggle with at scale. Automating the full teaching loop with equal pedagogical quality would require supervision that negates time savings.
Task automatability	claude-sonnet-5	2/5	AI can generate teaching materials, quizzes, and explanations, but live interactive teaching, mentorship, and hands-on lab supervision require human presence and expertise that current AI cannot fully replace.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Medical education is heavily regulated; accreditation bodies, medical boards, and institutional curricula typically require credentialed physician educators. Liability for incorrect instruction, professional standards for teaching pathology, and the need for human certification create substantial legal and organizational barriers to substituting AI alone.
Adoption barriers	claude-sonnet-5	4/5	Medical education and certification are governed by accreditation bodies requiring qualified physician educators; liability and credentialing requirements limit full AI substitution.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI-generated educational content can be cheap to produce, but effective medical education requires human instructors for mentoring, assessment, and credentialing—costs that remain high. Blended models reduce instructor burden incrementally rather than achieving dramatic cost reduction.
Cost vs. human wage	claude-sonnet-5	2/5	While AI-generated content is cheap, effective medical education still requires substantial physician oversight, curriculum design, and supervised practice, keeping overall costs closer to human-led instruction.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably performs end-to-end medical education delivery comparable to a qualified pathologist instructor. Chatbots and LLMs can supplement learning materials but lack the clinical judgment, ability to assess student comprehension dynamically, and credibility required for professional medical training in real institutions.
Technical feasibility today	claude-sonnet-5	2/5	AI tutoring and content-generation tools exist and are used to supplement medical education, but no deployed product independently delivers pathology instruction to trainees at production scale.

Identify the etiology, pathogenesis, morphological change, and clinical significance of diseases.

CI 20–30 · exposure 30 · augmentation 75 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Despite pilot programs and research enthusiasm, actual production deployment of AI for autonomous pathology diagnosis remains limited; most healthcare organizations use AI as a triage or second-opinion tool under pathologist control rather than as a replacement.
Sector adoption velocity	claude-sonnet-5	2/5	Pathology and clinical diagnostics are adopting AI tools cautiously through pilots and FDA-cleared narrow applications, but broad production deployment for full disease characterization remains rare.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI augmentation is strong and measurable: image analysis algorithms highlight suspicious regions, differential diagnosis tools suggest etiologies, and literature search agents retrieve mechanistic context, all of which measurably accelerate and improve pathologist performance when integrated into workflow.
Augmentation potential	claude-sonnet-5	4/5	AI tools assist pathologists significantly by flagging abnormalities, suggesting differential diagnoses, and summarizing literature, meaningfully boosting productivity while the physician retains final judgment.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI excels at pattern recognition in imaging and histology, the full task of identifying etiology, pathogenesis, morphological change, and clinical significance requires integrating complex causal reasoning, multi-modal data synthesis, and contextual judgment that current systems cannot reliably perform end-to-end without substantial human oversight and interpretation.
Task automatability	claude-sonnet-5	2/5	This requires deep clinical reasoning integrating morphology, patient history, and disease mechanism; AI can support with pattern recognition but cannot reliably perform the full synthesis at equal quality end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Pathology diagnosis carries high liability and error-cost asymmetry; regulatory bodies (CAP, CLIA) and clinical practice require a licensed pathologist to sign out and take responsibility for diagnostic statements, creating a hard barrier to full automation.
Adoption barriers	claude-sonnet-5	5/5	Diagnosis and disease characterization is a licensed medical act requiring physician sign-off, with major liability exposure, making full automation legally and professionally barred.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Current AI pathology tools require significant infrastructure, training data curation, and human validation by pathologists; the integration cost and per-case overhead remain comparable to or exceed the labor cost of a specialist pathologist reviewing a case.
Cost vs. human wage	claude-sonnet-5	2/5	Given the need for extensive physician oversight and validation of any AI output in this high-stakes domain, cost savings are limited despite cheaper inference for narrow subtasks.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	AI-based diagnostic tools (pathology image analysis, differential diagnosis assistants) exist in production for narrow domains (e.g., cancer subtyping), but they function as high-error-rate classifiers rather than systems that reliably establish full causal and mechanistic understanding across the breadth of pathology.
Technical feasibility today	claude-sonnet-5	2/5	AI diagnostic aids exist for narrow tasks like tumor detection in images, but no deployed product independently determines etiology and clinical significance across diseases reliably in production.

Consult with physicians about ordering and interpreting tests or providing treatments.

CI 20–28 · exposure 30 · augmentation 75 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Healthcare is adopting AI assistants at moderate pace (pilots and some clinical deployments), but regulatory caution, liability concerns, and entrenched physician workflows limit velocity; adoption is faster in imaging and lab analytics than in active consultation roles.
Sector adoption velocity	claude-sonnet-5	2/5	Healthcare overall adopts AI diagnostic aids cautiously and unevenly, with heavy regulatory oversight, so physician-to-physician consultation workflows have seen little real displacement by AI to date.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI significantly augments pathologists and consulting physicians by rapidly summarizing prior results, flagging relevant findings, and suggesting differential diagnoses, boosting their consultation speed and confidence while the physician retains full decision authority and communication responsibility.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist by summarizing test results, flagging anomalies, and providing rapid literature or guideline lookups, enhancing the pathologist's consultation without replacing the interactive judgment involved.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with test interpretation via image analysis or lab result summaries, the full task requires complex clinical judgment, patient context integration, and real-time communication with other physicians—elements that current systems cannot reliably handle end-to-end without substantial human oversight, falling short of 50% time savings at equal quality.
Task automatability	claude-sonnet-5	2/5	This involves nuanced clinical dialogue, integrating patient-specific context and judgment that current AI cannot reliably replicate end-to-end; only narrow sub-components like retrieving reference data could be offloaded.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Strong legal and regulatory barriers exist: only licensed physicians can order tests, interpret critical results, and consult on treatment; malpractice liability and standard-of-care requirements mandate physician sign-off, creating hard substitution barriers.
Adoption barriers	claude-sonnet-5	5/5	Interpreting tests and recommending treatment is a licensed medical act with direct liability; only a credentialed physician can legally provide this consultation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The task involves high-stakes clinical communication and decision-making where physician time commands premium rates; AI tools may reduce some overhead but do not eliminate the need for expert physician review and consultation, making all-in costs still substantially higher than outright AI automation.
Cost vs. human wage	claude-sonnet-5	2/5	AI tools can cheaply generate suggested interpretations, but the liability, oversight, and integration costs of using them in place of a consulting pathologist keep the effective all-in cost comparable to or higher than human time for this interactive task.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	AI products exist for specific subtasks (radiology interpretation, lab value flagging), but deployed systems have material limitations in nuance, contextual reasoning, and liability; consultation itself requires human physicians, though AI can support parts of the decision pipeline in production environments.
Technical feasibility today	claude-sonnet-5	2/5	Clinical decision support and AI diagnostic tools exist but are used as adjuncts, not as substitutes for pathologist-to-physician consultation, which requires interactive, accountable dialogue not yet handled by deployed products.

Diagnose diseases or study medical conditions, using techniques such as gross pathology, histology, cytology, cytopathology, clinical chemistry, immunology, flow cytometry, or molecular biology.

CI 20–25 · exposure 30 · augmentation 75 · importance 4.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Despite clinical interest, actual deployment of autonomous diagnostic AI in pathology labs remains limited. Most adoption is assistive (flagging regions of interest) rather than autonomous; organizational friction around liability, regulatory approval, and trust in AI results slows deep adoption.
Sector adoption velocity	claude-sonnet-5	2/5	Digital pathology adoption is growing but remains at pilot/limited-deployment stage in most health systems, constrained by regulatory approval, workflow integration, and conservative clinical culture.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI demonstrates strong assistive value in pathology: automated slide scanning, attention-to-detail on large datasets, highlighting suspicious areas, and speeding up routine screening significantly augment pathologist productivity while maintaining human control over final diagnosis and clinical judgment.
Augmentation potential	claude-sonnet-5	4/5	AI-assisted image analysis, quantification (e.g., IHC scoring, mitotic counts), and triage meaningfully speed up pathologists' workflow and improve consistency while the physician remains the final diagnostician.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with image analysis and pattern recognition in histology and cytology, diagnosis requires integration of multiple modalities, clinical context, and judgment that current systems cannot reliably perform end-to-end. Pathology diagnosis involves complex reasoning and specimen-specific factors that typically require >50% human oversight.
Task automatability	claude-sonnet-5	2/5	AI tools (e.g., digital pathology image analysis) can assist in specific sub-steps like slide screening, but the full diagnostic synthesis across multiple modalities and clinical context still requires physician judgment and legal sign-off, so end-to-end time savings well below 50% at equal quality.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Diagnosis by a pathologist is often a legally required step in clinical workflows; results must be signed off by a licensed physician. Regulatory bodies (CAP, CLIA) and malpractice liability create hard barriers—AI cannot legally replace the pathologist's professional judgment and certification.
Adoption barriers	claude-sonnet-5	5/5	Diagnosing disease is a licensed medical act requiring physician certification and legal accountability; regulatory and liability frameworks mandate human sign-off on pathology diagnoses.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI inference costs for image analysis are low, but integration, validation, oversight by licensed pathologists, and liability management add significant overhead. The total cost per diagnosis remains comparable to or exceeds pathologist labor, particularly when factoring in required human review.
Cost vs. human wage	claude-sonnet-5	2/5	AI image analysis tools have real licensing, integration, and validation costs, and still require pathologist oversight, making all-in cost per diagnosis comparable to or only modestly cheaper than human-only workflows.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed AI products exist for specific subtasks (e.g., histopathology image classification, cancer detection in slides), but they operate within narrow scopes and typically require pathologist review. Production systems exist but with material error rates and limited generalization across disease types and specimen variations.
Technical feasibility today	claude-sonnet-5	2/5	Some FDA-cleared AI products exist for narrow tasks (e.g., detecting cancerous cells in Pap smears or prostate biopsies), but no deployed product performs comprehensive multi-technique diagnosis reliably at scale in production.

Review cases by analyzing autopsies, laboratory findings, or case investigation reports.

CI 20–25 · exposure 25 · augmentation 63 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption of AI in pathology is in pilot and early-stage phases. While radiology has moved faster, autopsy and full case review automation lags due to lower case volume, strict regulatory requirements, and the specialized expertise involved. Production deployment remains rare.
Sector adoption velocity	claude-sonnet-5	2/5	Pathology and forensic medicine adopt digital tools slowly due to regulatory, legal, and specialized workflow constraints; AI adoption is mostly pilot-stage in image analysis, not widespread production use for full case review.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist pathologists by highlighting abnormal findings in images, cross-referencing lab data, and flagging unusual patterns, thereby raising review efficiency and diagnostic completeness. However, the human remains essential for final interpretation and medicolegal responsibility.
Augmentation potential	claude-sonnet-5	4/5	AI-assisted image analysis, natural language summarization of lab reports, and pattern recognition in histopathology can meaningfully speed up a pathologist's review process while they retain final judgment.
Task automatability	claude-haiku-4-5-20251001	2/5	Pathologists review complex case materials requiring deep domain knowledge, clinical judgment, and interpretation of subtle findings. While AI can assist with image analysis and flagging abnormalities, the synthesis of autopsy results, lab data, and case reports into a clinically sound conclusion requires human expertise and accountability.
Task automatability	claude-sonnet-5	2/5	AI can assist in reviewing lab findings and flagging anomalies in images or reports, but the integrative diagnostic judgment across autopsy findings, history, and case context remains beyond full automation today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Strong barriers exist: pathologists are licensed professionals whose expertise is legally required for death certification and medicolegal case conclusions. Liability and regulatory oversight of autopsy findings are substantial, and organizational and legal frameworks mandate qualified human review.
Adoption barriers	claude-sonnet-5	5/5	Pathology case review and death certification legally require a licensed physician's sign-off, with strong liability and regulatory requirements preventing full substitution.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Deployed AI systems for pathology image analysis carry significant infrastructure, validation, and oversight costs. The loaded cost of these systems plus required human review and sign-off remains comparable to or exceeds the cost of a pathologist performing the task directly.
Cost vs. human wage	claude-sonnet-5	2/5	AI tools for image analysis are cheap per query, but the overall task requires physician-level synthesis and liability, so the all-in cost including oversight remains comparable to or higher than human cost for full case review.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI tools exist for specific subtasks like histopathology image analysis and lab data interpretation, but no deployed product reliably performs the holistic case review that a pathologist conducts. Systems lack the contextual reasoning and medicolegal accountability required in clinical practice.
Technical feasibility today	claude-sonnet-5	2/5	Digital pathology AI tools exist for specific tasks (e.g., tumor detection, slide triage) but no deployed product performs comprehensive case review integrating autopsy, lab, and investigative findings in production.

Develop or adopt new tests or instruments to improve diagnosis of diseases.

CI 6–25 · exposure 13 · augmentation 63 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	While digital pathology and AI-assisted analysis are emerging, the pace of actual adoption of new diagnostic instruments remains slow due to regulatory requirements, validation timelines, and conservative institutional adoption patterns in healthcare.
Sector adoption velocity	claude-sonnet-5	2/5	While AI adoption in diagnostics research is growing, actual development and clinical adoption of new tests remains slow due to regulatory and validation requirements.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist pathologists by accelerating literature synthesis, identifying patterns in validation datasets, and automating preliminary statistical analysis, thereby raising productivity in parts of the development and evaluation workflow.
Augmentation potential	claude-sonnet-5	4/5	AI can significantly assist pathologists in identifying patterns in data, generating hypotheses, and analyzing large datasets to inform new diagnostic approaches, even though humans must design and validate the final tests.
Task automatability	claude-haiku-4-5-20251001	2/5	This task requires conceptual innovation, experimental design, and clinical judgment to identify diagnostic gaps and evaluate new tools. While AI can assist in literature review and data analysis, the core creative and evaluative work of developing or adopting new diagnostic instruments remains dependent on human expertise and cannot achieve 50% time savings end-to-end.
Task automatability	claude-sonnet-5	1/5	Developing or adopting novel diagnostic tests/instruments requires original scientific research, validation, and creative problem-solving that current AI cannot execute end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory barriers are substantial: new diagnostic tests and instruments require FDA approval, clinical validation, and often institutional review board oversight. Liability and accuracy requirements are high, and adoption requires buy-in from laboratory directors and clinicians, all of which enforce human oversight.
Adoption barriers	claude-sonnet-5	5/5	New diagnostic tests require rigorous clinical validation, regulatory approval (e.g., FDA), and physician sign-off, making this one of the most heavily regulated and liability-sensitive tasks in medicine.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The overhead of AI-assisted development (requiring human pathologists to oversee, validate, and integrate findings) is comparable to or potentially higher than direct human research and development effort, given the specialized expertise required.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cheaply assist with literature review or data analysis, but the overall cost of developing/validating a new test still requires expensive expert labor, trials, and regulatory processes.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI systems autonomously develop or adopt diagnostic tests at scale. AI tools exist for supporting components (literature mining, statistical analysis) but no production system performs the full cycle of needs assessment, prototype evaluation, validation, and institutional adoption without substantial human direction.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously develops or validates new diagnostic tests; this remains a human-led R&D and clinical adoption process with AI only as a tool.

Manage medical laboratories.

CI 3–25 · exposure 13 · augmentation 50 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	While healthcare is digitizing, laboratory management automation is limited. Most adoption centers on ancillary tools (LIS optimization, data analytics) rather than autonomous management. Regulatory conservatism, physician oversight norms, and liability concerns slow displacement of human laboratory directors in production settings.
Sector adoption velocity	claude-sonnet-5	2/5	Healthcare administration adopts AI slowly for operational tasks, with pilots for scheduling or inventory but not for core managerial decision-making in labs.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist pathologists in laboratory management through workflow analytics, quality control monitoring, scheduling optimization, and decision support. These tools improve human productivity and error detection, though the pathologist remains responsible for strategic and compliance decisions.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with data analytics, inventory forecasting, quality control monitoring, and compliance documentation, improving efficiency while the pathologist retains managerial responsibility.
Task automatability	claude-haiku-4-5-20251001	2/5	Laboratory management involves complex coordination of personnel, equipment, quality control, budgeting, and regulatory compliance. While AI can assist with scheduling, inventory tracking, and some quality metrics, the task fundamentally requires human judgment on staffing decisions, error investigation, and regulatory interpretation that current systems cannot fully automate. Most of the task remains human-driven.
Task automatability	claude-sonnet-5	1/5	Managing a laboratory involves personnel supervision, budgeting, quality assurance oversight, regulatory compliance, and strategic decision-making that requires human judgment and accountability far beyond current AI capabilities.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Pathologists must be licensed physicians, and regulatory bodies (CAP, CLIA) hold the laboratory's director legally accountable for operations, quality, and test validity. Liability and compliance requirements create strong barriers to full AI substitution without human sign-off on critical decisions.
Adoption barriers	claude-sonnet-5	5/5	Lab directorship requires licensed physician oversight under regulations like CLIA, with legal accountability for quality, safety, and compliance that cannot be delegated to software.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Laboratory management requires domain expertise and accountability; the total cost of AI systems (including setup, integration, ongoing compliance monitoring, and required human supervision) approaches or exceeds the cost of experienced human managers given the low error tolerance and regulatory requirements.
Cost vs. human wage	claude-sonnet-5	1/5	There is no AI substitute performing this managerial role, so cost comparison favors the human entirely; any AI tools used are supplementary, not replacements.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably performs end-to-end laboratory management autonomously. AI tools exist for specific subtasks (workflow optimization, data analysis) but production systems still require significant human oversight, manual exception handling, and judgment calls that AI cannot consistently perform across the full scope.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product manages medical laboratories end-to-end; existing lab software assists with data/workflow tracking but does not perform managerial functions.

Communicate pathologic findings to surgeons or other physicians.

CI 3–16 · exposure 13 · augmentation 50 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Pathology departments use AI-assisted diagnosis and report generation in limited pockets, but the communication of findings to surgeons remains a core physician function with slow adoption of AI-driven substitution. Most organizations retain human pathologists in the communication loop for both legal and clinical reasons.
Sector adoption velocity	claude-sonnet-5	2/5	Healthcare, especially clinical communication and diagnostic sign-off, remains a slow-adopting sector for full automation due to regulatory and safety constraints, though AI drafting tools are piloted.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist by pre-drafting findings summaries, flagging critical results, and suggesting relevant comparisons or follow-up tests, helping the pathologist communicate more clearly and efficiently. However, the pathologist's judgment, authority, and direct interaction with colleagues remain central to the task.
Augmentation potential	claude-sonnet-5	3/5	AI can help draft structured reports, summarize findings, and flag key results to support the pathologist's communication, offering moderate productivity benefits while human judgment and delivery remain essential.
Task automatability	claude-haiku-4-5-20251001	1/5	Communicating pathologic findings requires clinical judgment, peer-to-peer professional dialogue, real-time responsiveness to follow-up questions, and the ability to adjust explanation based on the recipient's specialty and context. Current AI cannot reliably participate in this nuanced, bidirectional professional communication.
Task automatability	claude-sonnet-5	2/5	This requires synthesizing complex diagnostic findings into clinically actionable communication with real-time judgment and dialogue, which current AI cannot reliably do end-to-end despite drafting assistance possibilities.'
Adoption barriers	claude-haiku-4-5-20251001	5/5	Legal and regulatory requirements mandate that a licensed pathologist must review findings, sign off on reports, and communicate directly with treating physicians. Liability for misinterpretation, patient safety, and professional accountability create hard barriers to full automation or substitution.
Adoption barriers	claude-sonnet-5	5/5	Pathology reporting and inter-physician communication is legally and professionally mandated to be performed by licensed physicians, with strong liability and regulatory requirements.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	This task fundamentally requires a licensed pathologist's time and expertise to ensure accuracy and accountability. AI assistance in drafting findings does not eliminate the need for professional review and real-time communication, so the full-task cost remains driven by pathologist labor.
Cost vs. human wage	claude-sonnet-5	1/5	Since AI cannot substitute for the licensed pathologist's judgment and liability-bearing communication, there is no viable AI cost comparison—human physicians must perform this.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI can draft summary statements of pathologic findings from structured data or prior reports, but deployed systems do not reliably handle the interactive, context-dependent communication required when a surgeon calls with urgent questions or needs clarification on implications for treatment. Current practice requires a licensed physician.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously communicates pathology findings to treating physicians; this remains a physician-to-physician clinical communication task, not automated in practice.

Perform autopsies to determine causes of deaths.

CI 0–5 · exposure 5 · augmentation 38 · importance 3.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Autopsy rates have been declining for decades, and the sector remains low-digitization with strong professional and legal gatekeeping. Digital pathology adoption is slow, and there is minimal economic pressure to automate what is already a constrained, specialized procedure.
Sector adoption velocity	claude-sonnet-5	1/5	Forensic and hospital pathology is a highly manual, physically-grounded medical specialty with minimal AI deployment for the core procedural task.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI can assist with image analysis and flagging of potential diagnostic findings in pathology slides, but the autopsy process itself—tissue sampling, gross examination, decision-making about which samples to analyze—relies heavily on pathologist judgment and cannot be meaningfully augmented by current AI tools.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with digital pathology image analysis, report generation, and literature/differential diagnosis support, but not the dissection or determination process itself.
Task automatability	claude-haiku-4-5-20251001	1/5	Autopsies require physical tissue examination, dissection, and interpretation of anatomical findings in context of individual case history—activities that demand hands-on laboratory work and real-time decision-making that current AI cannot perform end-to-end. While AI can assist with image analysis of pathology slides, the full autopsy procedure involves irreducible manual and investigative components.
Task automatability	claude-sonnet-5	1/5	Autopsy performance requires physical dissection, tissue handling, and expert judgment integrating gross and microscopic findings; no AI system can perform the physical procedure end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Autopsies are legally required to be performed or directly overseen by a licensed medical examiner or pathologist in most jurisdictions, and next-of-kin authorization is mandatory. Regulatory and liability requirements create hard barriers to AI substitution.
Adoption barriers	claude-sonnet-5	5/5	Autopsies are legally required to be performed by licensed pathologists/medical examiners, with strict chain-of-custody, forensic, and legal accountability requirements.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The cost of autopsy equipment, facilities, and the technical infrastructure required to support even partial AI automation remains high relative to the procedural time saved, and a pathologist's involvement is legally and professionally mandatory, making full cost displacement impossible.
Cost vs. human wage	claude-sonnet-5	1/5	There is no AI substitute for the physical task, so cost comparison favors the human entirely; any AI use is supplementary at added cost.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI systems can support autopsy interpretation through digital pathology image analysis and some diagnostic assistance, but no deployed product performs the full autopsy autonomously or reliably. Pathologists use AI as a supplementary tool, not as a substitute for the procedure itself.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs autopsies; AI is at most used in ancillary image analysis or report drafting, not the procedure itself.

Obtain specimens by performing procedures, such as biopsies or fine needle aspirations (FNAs) of superficial nodules.

CI 0–0 · exposure 0 · augmentation 38 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	No meaningful AI adoption has occurred for autonomous biopsy/FNA performance because the task remains legally and clinically non-delegable to machines in all healthcare sectors.
Sector adoption velocity	claude-sonnet-5	1/5	Physical, invasive medical procedures remain among the least digitized/automated tasks, with no meaningful trend toward AI or robotic replacement in this specific act.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist pathologists by analyzing ultrasound/imaging to guide needle placement or by pre-screening imaging to flag suspicious nodules for biopsy, but the actual specimen extraction remains physician-performed.
Augmentation potential	claude-sonnet-5	2/5	AI can assist with pre-procedure imaging guidance or specimen adequacy assessment, but offers minimal direct enhancement to the physical act of specimen collection itself.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires precise physical manipulation of needles and tissue extraction in a live patient, combined with real-time anatomical judgment and procedural adaptation—capabilities that current AI systems cannot perform autonomously today.
Task automatability	claude-sonnet-5	1/5	This is a hands-on invasive physical procedure requiring manual dexterity, patient contact, and real-time tactile feedback that current AI systems cannot perform.
Adoption barriers	claude-haiku-4-5-20251001	5/5	This task is legally and ethically restricted to licensed physicians; regulatory frameworks (FDA, state medical boards) explicitly prohibit unlicensed or automated performance of needle-based tissue sampling in living patients.
Adoption barriers	claude-sonnet-5	5/5	Performing invasive diagnostic procedures on patients requires a licensed physician, with strict legal, liability, and safety requirements that bar non-human execution.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The procedure requires a licensed physician's direct hands-on labor; AI has no economic displacement advantage since it cannot legally or safely perform the invasive procedure itself.
Cost vs. human wage	claude-sonnet-5	1/5	There is no AI substitute performing this physical task, so AI cost is not comparable—human physician labor is the only viable option.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI system can independently perform biopsies or FNAs on patients; these remain exclusively human-executed procedures in all clinical settings.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI or robotic product autonomously performs biopsies or FNAs on patients; this remains purely a human clinical procedure.

Plan and supervise the work of the pathology staff, residents, or visiting pathologists.

CI 0–0 · exposure 0 · augmentation 38 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Healthcare organizations have shown virtually no adoption of AI systems for staff management and supervision; these are core human leadership functions unlikely to be delegated to AI agents in the foreseeable future.
Sector adoption velocity	claude-sonnet-5	1/5	Healthcare administrative/supervisory functions show minimal AI adoption for actual personnel management, lagging far behind clinical decision-support tools.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI might assist with scheduling, workload analysis, or compliance tracking, but the core supervisory, developmental, and evaluative aspects of leading pathology staff remain inherently human work where AI offers only marginal support tools.
Augmentation potential	claude-sonnet-5	3/5	AI can help with scheduling, workflow tracking, performance metrics, or case-load management tools that assist a supervising pathologist, but does not replace the supervisory judgment itself.
Task automatability	claude-haiku-4-5-20251001	1/5	Planning and supervising human staff requires nuanced judgment about individual capabilities, priorities, institutional dynamics, and complex problem-solving that current AI systems cannot perform end-to-end. This task fundamentally depends on real-time human interaction and accountability that AI cannot replicate.
Task automatability	claude-sonnet-5	1/5	Supervising staff and trainees requires interpersonal judgment, mentorship, accountability, and organizational leadership that current AI cannot perform end-to-end.'
Adoption barriers	claude-haiku-4-5-20251001	5/5	Medical staff supervision is fundamentally a human leadership and accountability function; institutional policy, medical licensing, and professional standards require human pathologists to direct and evaluate their teams' work.
Adoption barriers	claude-sonnet-5	5/5	Supervision of residents and staff requires a licensed, credentialed physician with legal and institutional accountability, making this a hard-barrier task.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	An AI system that could supervise pathology staff would require continuous oversight, integration with institutional systems, and human review of decisions—making it more expensive than the supervising physician's existing work, not cheaper.
Cost vs. human wage	claude-sonnet-5	1/5	There is no AI substitute performing this managerial task, so no cost comparison favors AI; a human supervisor remains the only option.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI product reliably supervises or manages human staff in medical settings today. This requires human leadership judgment, conflict resolution, performance evaluation, and accountability that remains firmly within human domain.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product manages or supervises human clinical staff or residents; this remains entirely outside current AI product scope.

Testify in depositions or trials as an expert witness.

CI 0–0 · exposure 0 · augmentation 50 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Legal and regulatory frameworks globally require human expert witnesses; there is no measurable adoption of AI as an independent testifying expert because such substitution is prohibited by law and professional ethics.
Sector adoption velocity	claude-sonnet-5	1/5	Legal and courtroom testimony processes are highly resistant to automation, with essentially no adoption of AI substitutes for sworn expert testimony.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist pathologists in preparing testimony by synthesizing literature, organizing case evidence, or drafting talking points, but the expert themselves must deliver and defend all testimony in court.
Augmentation potential	claude-sonnet-5	3/5	AI can help pathologists prepare testimony by organizing case data, summarizing records, or drafting reports, but it does not participate in the actual testimony itself.
Task automatability	claude-haiku-4-5-20251001	1/5	Expert witness testimony requires real-time adversarial cross-examination, credibility judgment by opposing counsel and judges, and dynamic response to unanticipated legal and factual challenges—capabilities that current AI cannot reliably perform without human authority and accountability.
Task automatability	claude-sonnet-5	1/5	Expert witness testimony requires a licensed pathologist's personal medical judgment, credibility, and live sworn testimony subject to cross-examination; no AI system can perform this in place of the human today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Legal testimony requires a human expert to be sworn under oath, subject to cross-examination, and personally liable for perjury; statutes and court rules explicitly mandate human testimony and do not permit AI substitution.
Adoption barriers	claude-sonnet-5	5/5	Legal rules of evidence and expert witness qualification require a licensed, credentialed human to testify under oath, with strict liability, perjury, and professional accountability implications.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	Expert witness services command substantial hourly rates ($300–$1000+) including deposition and trial prep, and AI cannot legally substitute for the human expert, making direct cost comparison inapplicable and AI purely supplementary if used at all.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI substitute cost to compare; the human expert's fee is the only real option, making AI effectively unusable and thus not cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI system can independently serve as an expert witness in legal proceedings; testimony requires a licensed human expert who can be sworn, deposed, and held legally liable for false statements.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product testifies as an expert witness; this remains entirely outside current AI product capability and courts require a human expert.

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.