Ophthalmologists, Except Pediatric
29-1241.00Median wage $300,080/yr8,950 employed (US)Rank #756 of 923 scored · top 82% by substitution
Diagnose and perform surgery to treat and help prevent disorders and diseases of the eye. May also provide vision services for treatment including glasses and contacts.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution16
Exposure18
Augmentation60
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
18 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
0%
Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.
Why this score
The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.
Task automatabilityw 35%16
panel mean rating 1.6/5 → substitution pressure 16/100
Technical feasibility todayw 20%21
panel mean rating 1.8/5 → substitution pressure 21/100
Cost vs. human wagew 15%18
panel mean rating 1.7/5 → substitution pressure 18/100
Adoption barriersw 20%inverted — strong barriers lower the score7
panel mean rating 4.7/5 (barrier strength) → substitution pressure 7/100
Sector adoption velocityw 10%25
panel mean rating 2.0/5 → substitution pressure 25/100
Task breakdown (18 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.
Perform, order, or interpret the results of diagnostic or clinical tests.
54CI 36–71 · exposure 55 · augmentation 75 · importance 4.6/5 · click for rater detail
Perform, order, or interpret the results of diagnostic or clinical tests.
54| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Adoption is growing in specialized screening programs, hospitals, and large practices, but remains patchy; many small ophthalmology practices have not integrated AI, and regulatory/reimbursement clarity is still evolving, placing adoption at middling speed with pilots more common than standardized production rollout. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Ophthalmology is an early adopter of AI diagnostics among medical specialties, with real-world deployment of autonomous screening tools, but broader adoption across all diagnostic test types remains in pilot or narrow-use stages. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially assists ophthalmologists by rapidly flagging abnormalities, prioritizing cases, and providing quantitative measurements, allowing physicians to focus cognitive effort on clinical judgment and patient communication. This augmentation effect is strong across most diagnostic test types. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially augments ophthalmologists by flagging abnormalities in imaging, prioritizing cases, and providing decision support, meaningfully speeding interpretation while the physician retains final authority. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI systems can now interpret many diagnostic tests (retinal imaging, OCT scans, visual fields) with accuracy matching or exceeding human ophthalmologists on specific conditions, and can flag abnormalities for review, achieving substantial time savings on image analysis and initial triage. However, complex multi-test integration and clinical decision-making still require human oversight, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can interpret some diagnostic images (e.g., retinal scans for diabetic retinopathy) with high accuracy, but 'perform' includes hands-on exams and ordering requires clinical judgment integrating patient history, so full end-to-end automation is not yet achievable at equal quality across the breadth of tests. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While ophthalmologists must license and sign off on final clinical decisions, diagnostic test interpretation itself has modest barriers—AI outputs are reviewed/approved by licensed clinicians but do not require the physician to perform the initial interpretation. Liability concerns and the expectation of physician oversight provide moderate friction. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Diagnostic interpretation for treatment decisions generally requires a licensed physician's sign-off, and misdiagnosis carries significant liability, creating strong regulatory and legal barriers to full substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-based diagnostic support (inference + integration) costs pennies to dollars per image or test, while ophthalmologist time for equivalent interpretation costs hundreds of dollars; the all-in ratio heavily favors automation once systems are deployed. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-based screening tools are cheap per scan once deployed, but integration, hardware, liability insurance, and required physician oversight keep overall cost roughly comparable to human interpretation for most test types. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature FDA-cleared and deployed AI products exist for diabetic retinopathy detection, glaucoma screening, and retinal disease classification in production settings at scale; performance is reliable on the image interpretation component. Some tests (e.g., complex visual field interpretation in atypical cases) have lower deployment maturity, justifying a 4 rather than 5. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | FDA-cleared autonomous AI systems (e.g., IDx-DR) reliably screen for diabetic retinopathy in deployed clinical settings, but this covers only a narrow slice of ophthalmic diagnostic testing; most OCT, visual field, and slit-lamp interpretation still relies on physician judgment. |
Educate patients about maintenance and promotion of healthy vision.
41CI 30–51 · exposure 42 · augmentation 75 · importance 4.6/5 · click for rater detail
Educate patients about maintenance and promotion of healthy vision.
41| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare, especially ophthalmology, is a regulated and risk-averse sector where patient education remains a direct physician responsibility; adoption of standalone AI for this task in production practice is currently minimal, with most usage confined to supplementary materials rather than replacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially specialized medical practice, has historically been slower to adopt AI patient-facing tools compared to other professional services, though patient portals and chat-based triage are increasing. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can substantially assist ophthalmologists by drafting patient-friendly explanations, generating visual aids, and preparing pre-visit educational materials, allowing the physician to focus on personalized discussion and addressing patient questions more efficiently while maintaining clinical authority. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can efficiently generate handouts, answer common patient questions, and support the physician's educational efforts, meaningfully improving throughput and consistency while the physician remains the trusted source. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can generate educational materials and scripts about vision health, but patient education requires adapting explanations to individual comprehension levels, addressing specific concerns, and building trust—tasks where current AI lacks the nuanced human judgment to replace the ophthalmologist end-to-end with 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 3/5 | AI chatbots and generative systems can produce accurate, personalized patient education content on eye health, but delivering it within a clinical relationship with tailored context still requires physician time for at least part of the interaction. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Patient education by a licensed physician is a clinical and ethical expectation in most healthcare settings; liability, regulatory oversight of medical information accuracy, and the professional standard that the treating physician educates their own patients create substantial adoption barriers. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement mandates a physician deliver general health education, but liability concerns and patient trust/preference for physician-delivered advice create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | While AI-generated educational content is cheap to produce, the ophthalmologist's time spent in patient education is already integrated into billable clinical encounters; standalone AI tutoring would require additional implementation cost and oversight, making the true marginal cost-per-task comparison less favorable than it appears. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating standardized educational materials or chatbot-delivered advice is far cheaper than physician time, though integration and oversight for accuracy reduce the savings somewhat. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Chatbots and educational AI tools exist to provide standardized vision health information, but they cannot reliably adapt to individual patient contexts, handle complex medical histories, or convey the personalized, authoritative guidance that physicians deliver in clinical settings. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Patient-facing AI education tools and chatbots exist and are used in some clinics/portals, but they are not universally reliable or tailored enough to fully replace physician counseling in production ophthalmology practice. |
Document or evaluate patients' medical histories.
34CI 32–36 · exposure 34 · augmentation 75 · importance 4.6/5 · click for rater detail
Document or evaluate patients' medical histories.
34| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Healthcare organizations are actively piloting and deploying AI-assisted clinical documentation in EHRs, but adoption remains concentrated in large hospital systems and primary care. Specialist practices like ophthalmology show slower, more cautious adoption; widespread production deployment of unsupervised AI evaluation is not yet standard. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Healthcare has been adopting AI scribes and documentation assistants steadily, but overall adoption in clinical care lags behind pure information/finance sectors due to regulatory and workflow constraints. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | Ambient AI transcription and structured data entry assistance meaningfully reduce ophthalmologists' time on documentation and data organization, allowing them to focus on clinical evaluation and patient interaction. Voice-to-note and auto-population of standardized fields demonstrably enhance workflow productivity while keeping the clinician in full control and oversight of medical interpretation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI ambient documentation and summarization tools meaningfully speed up history-taking and note generation, letting physicians focus more on patient interaction while AI handles drafting and organization. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Current AI can extract and organize structured medical history data from documents or patient interviews with reasonable accuracy, but evaluating—interpreting clinical significance, detecting contradictions, and synthesizing history into diagnostic context—requires nuanced clinical judgment that AI struggles with at reliable quality parity. Time savings exist for transcription and initial data entry, but do not yet reach the 50% threshold for end-to-end task completion. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help summarize and structure documentation from dictation or records, but eliciting nuanced history and integrating it with clinical judgment for an ophthalmology exam still requires physician involvement, so full end-to-end automation with equal quality is not yet achievable. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Medical evaluation and documentation carry significant regulatory and liability requirements; a licensed physician must take responsibility for medical history accuracy and clinical interpretation. Malpractice law and medical licensing rules effectively require human clinician signature and judgment on the evaluation component, creating hard barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Medical history documentation feeds directly into diagnosis and billing, and licensed physicians bear legal and clinical responsibility for accuracy, creating strong liability and regulatory barriers to fully autonomous automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI transcription and documentation aids reduce clerical time and data entry cost, but ophthalmologists' fully-loaded wages remain high, and AI integration, oversight, and error correction add overhead. Cost savings are modest and do not yet outweigh human labor for the full task cycle, especially given liability sensitivity in medical evaluation. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI scribing tools are moderately priced per encounter compared to physician time saved, offering some savings, but physician oversight and correction still consume significant time, keeping the ratio only moderately favorable. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Clinical documentation tools with AI-assisted transcription and note generation exist in production EHR systems (e.g., ambient voice capture), but deployed products typically require substantial human review, correction, and clinical sign-off. The evaluation of medical history—clinical interpretation and synthesis—remains a human function; AI products handle only the data-capture portion reliably. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Ambient scribe and EHR summarization products are deployed in clinical practice and reduce documentation burden, but they require physician review/correction and are not fully autonomous at evaluating history for clinical decision-making. |
Diagnose or treat injuries, disorders, or diseases of the eye and eye structures including the cornea, sclera, conjunctiva, or eyelids.
21CI 16–25 · exposure 30 · augmentation 75 · importance 4.8/5 · click for rater detail
Diagnose or treat injuries, disorders, or diseases of the eye and eye structures including the cornea, sclera, conjunctiva, or eyelids.
21| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption has been slow and limited to specific screening workflows (diabetic retinopathy) in integrated health systems and specialized centers. Most independent and hospital-based ophthalmology practices remain traditional; adoption data show pilots rather than deep, production-scale displacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare overall adopts AI diagnostic aids slowly due to regulatory approval processes, liability concerns, and integration challenges, despite promising research in ophthalmic imaging. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI image analysis tools meaningfully assist ophthalmologists in detecting and quantifying specific pathologies (lesion segmentation, glaucoma risk scoring, retinal disease grading), allowing faster screening and documentation. Ophthalmologists remain in the diagnostic loop, with AI raising throughput and diagnostic confidence on imaging-heavy tasks. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-assisted imaging analysis (e.g., OCT interpretation, diabetic retinopathy screening) meaningfully augments ophthalmologists' diagnostic speed and accuracy while they retain clinical decision-making. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI diagnostic tools can assist in detecting certain eye pathologies from imaging (retinopathy, glaucoma screening), the task requires integrating complex clinical judgment, patient history, physical examination findings, and determining treatment plans—functions that remain partially manual. Current AI cannot reliably replace the full diagnostic-to-treatment workflow at quality parity with 50% time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can assist with image-based diagnostic support (e.g., retinal imaging analysis) but the full task involves physical exams, patient history, surgical treatment, and clinical judgment that current systems cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Ophthalmology is a licensed medical specialty; diagnosis and treatment decisions must be rendered or legally signed off by a licensed ophthalmologist. Liability, standard-of-care expectations, and regulatory requirements (state medical boards, malpractice) create hard legal and professional barriers to autonomous AI practice. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Diagnosing and treating eye diseases requires a licensed physician; medical licensing, malpractice liability, and regulatory requirements make this a strongly protected task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Ophthalmologist labor is expensive (~$200–300/hour loaded), but current AI systems require significant infrastructure (validated imaging hardware, ophthalmologist oversight, integration), making the all-in cost per diagnosis comparable to or higher than a full human visit, especially for complex cases. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Narrow AI diagnostic tools can be cheap per screening, but full diagnosis/treatment requiring physician oversight, equipment, and liability coverage keeps overall costs comparable to or above human-only care. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | FDA-cleared AI algorithms exist for specific conditions (diabetic retinopathy detection, age-related macular degeneration screening) and are deployed in some clinics, but they operate as narrow, assistive tools rather than end-to-end diagnostic systems. Real-world adoption shows these tools have material limitations in edge cases and non-English populations. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some FDA-approved diagnostic tools exist (e.g., diabetic retinopathy screening), but comprehensive diagnosis and treatment of diverse eye disorders is not reliably automated in production settings. |
Perform comprehensive examinations of the visual system to determine the nature or extent of ocular disorders.
20CI 20–20 · exposure 25 · augmentation 63 · importance 4.8/5 · click for rater detail
Perform comprehensive examinations of the visual system to determine the nature or extent of ocular disorders.
20| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Ophthalmology is a regulated medical specialty with strong professional standards and relatively lower digitization of core clinical workflows compared to radiology or pathology. While some practices pilot AI screening tools, production adoption of AI for comprehensive diagnosis remains limited; the sector lags information and finance sectors. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially specialized clinical diagnostics, adopts AI tools cautiously due to regulatory approval processes, liability concerns, and the need for extensive validation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-assisted image analysis and automated screening can meaningfully support ophthalmologists by flagging abnormalities in fundus or OCT images, reducing manual review time and improving consistency. However, augmentation is confined to specific imaging modalities rather than transforming the full clinical examination, which remains highly interactive and dependent on real-time patient assessment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-based imaging analysis, decision support, and documentation tools meaningfully speed up parts of the diagnostic workflow while the ophthalmologist retains full clinical control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist with limited parts of comprehensive eye exams—image analysis of retinal fundus photos or optical coherence tomography scans—but cannot yet perform the full examination end-to-end. Clinical judgment, patient history integration, tonometry, visual acuity assessment, and dynamic examination require human presence and expertise; automation falls well short of 50% time savings at equal quality across the complete diagnostic workflow. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can assist with imaging interpretation and diagnostic support, but the full comprehensive exam requires hands-on procedures, patient interaction, and integrated clinical judgment not achievable end-to-end by current systems.rsatile |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Comprehensive eye examinations involve clinical diagnosis and prescription of treatment; regulatory frameworks (FDA, state medical boards) require a licensed physician to perform or legally sign off on these determinations. Liability for missed diagnoses, malpractice considerations, and patient safety law create hard legal and regulatory barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Comprehensive eye exams require a licensed physician to perform, interpret, and take legal/clinical responsibility for diagnosis, representing a hard regulatory and licensure barrier. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI imaging analysis tools have moderate deployment costs (software, integration, image acquisition) but still require an ophthalmologist to interpret results, perform clinical maneuvers, and make diagnoses. The loaded cost of the full examination (physician time + overhead) remains lower than the combined cost of AI infrastructure plus required physician review. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI imaging analysis is cheap per scan, but the exam requires physical equipment operation, patient contact, and physician oversight, keeping overall costs comparable to human-delivered care. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Research-grade AI systems perform well on narrow subtasks (diabetic retinopathy detection, glaucoma screening from images), and some tools exist in clinical settings for triage, but no deployed product reliably performs a comprehensive visual system examination without substantial physician oversight and re-examination. Material error rates and narrow scope persist in production. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI diagnostic tools (e.g., retinal screening for diabetic retinopathy) are deployed for narrow subtasks, but no product performs a full comprehensive ocular exam autonomously in production. |
Prescribe corrective lenses such as glasses or contact lenses.
20CI 20–20 · exposure 25 · augmentation 63 · importance 4.3/5 · click for rater detail
Prescribe corrective lenses such as glasses or contact lenses.
20| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI-assisted refraction and recommendation tools in ophthalmology remains slow and limited to pilot programs in well-resourced settings. The requirement for clinician sign-off and regulatory conservatism in healthcare slows penetration significantly. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially direct clinical decision-making, adopts AI cautiously due to regulation and liability, with pilots more common than full production deployment for prescribing. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by automating refraction measurement, flagging common lens options, and organizing patient data, meaningfully speeding up the prescription workflow. However, final clinical decision-making and patient counseling remain centered on the ophthalmologist. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-enabled autorefractors, OCT analysis, and decision-support tools meaningfully speed up and improve accuracy of the refraction process feeding into the prescription decision. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in refractive measurements and lens selection, prescribing corrective lenses requires clinical judgment about patient-specific factors (comfort, lifestyle, prior experience) and legal authority that only licensed ophthalmologists possess. The task cannot be performed end-to-end by AI without human oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | Refraction measurement can be partially automated with autorefractors, but prescribing requires integrating exam findings, patient history, and clinical judgment that current AI cannot fully replicate end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Prescribing corrective lenses is a licensed clinical function; only licensed ophthalmologists or optometrists can legally issue prescriptions in virtually all jurisdictions. Regulatory and legal barriers are absolute—no non-licensed entity can substitute. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Prescribing corrective lenses is a licensed medical act requiring a physician's legal authorization in virtually all jurisdictions, creating a hard regulatory barrier. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI-based refraction and recommendation systems, plus required ophthalmologist review and sign-off, remains comparable to or more expensive than direct clinician prescription, offering minimal cost savings when human oversight is mandatory. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Devices reduce time on manual refraction but still require licensed clinician oversight and equipment costs, so total cost savings versus a human optometrist/ophthalmologist are modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product independently prescribes corrective lenses at scale. AI tools exist for refraction assistance and lens recommendation, but they operate as decision-support systems requiring a licensed clinician to make and sign the prescription, not as autonomous prescription systems. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Autorefractors and some AI-assisted diagnostic tools are deployed, but no product independently generates a final prescription without physician verification and sign-off. |
Develop treatment plans based on patients' histories and goals, the nature and severity of disorders, and treatment risks and benefits.
18CI 16–20 · exposure 25 · augmentation 63 · importance 4.7/5 · click for rater detail
Develop treatment plans based on patients' histories and goals, the nature and severity of disorders, and treatment risks and benefits.
18| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While ophthalmology uses AI for image analysis (diabetic retinopathy screening), adoption of AI for treatment planning itself remains limited to pilot programs and research settings; production deployment in clinical workflows is rare. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially specialized surgical fields like ophthalmology, adopts AI cautiously due to regulatory, liability, and workflow integration constraints, with pilots more common than broad production use. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by summarizing patient histories, presenting evidence-based treatment guidelines, and highlighting relevant risk factors, but the ophthalmologist remains firmly in the loop for final clinical decision-making. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-based imaging analysis, risk calculators, and literature synthesis tools meaningfully assist ophthalmologists in evaluating disease severity and treatment options, enhancing decision quality while the physician retains final judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in summarizing patient histories and displaying treatment options, developing a comprehensive treatment plan requires integrating complex clinical judgment, patient-specific risk assessment, and nuanced goal prioritization that current AI systems cannot reliably execute end-to-end to meet the 50% time-saving bar at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can support treatment plan development with diagnostic and decision-support tools, but integrating patient history, personal goals, and nuanced risk-benefit judgment for individualized care requires physician synthesis that current systems cannot fully replicate end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | A licensed ophthalmologist must legally develop and sign off on treatment plans; regulatory standards (FDA, state medical boards) and malpractice liability create hard barriers to full automation of this task. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Treatment planning is a core licensed medical function requiring physician judgment, informed consent processes, and legal accountability, making this a hard-barrier task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI-assisted ophthalmologic decision support still requires a licensed ophthalmologist's time for validation, modification, and legal responsibility, making the combined cost comparable to or exceeding the cost of direct human planning. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools reduce time on data synthesis and imaging analysis, but physician oversight, liability, and integration costs keep overall cost comparable to or only modestly below physician-led planning. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Clinical decision-support tools exist and can flag treatment guidelines, but no deployed product reliably performs the full task of developing individualized treatment plans without substantial physician oversight and modification; this remains research-stage or narrow-scope in production. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Clinical decision support and diagnostic AI (e.g., retinal imaging analysis) are deployed in ophthalmology, but no product autonomously generates complete, individualized treatment plans reliably in production. |
Conduct clinical or laboratory-based research in ophthalmology.
18CI 11–25 · exposure 13 · augmentation 75 · importance 2.6/5 · click for rater detail
Conduct clinical or laboratory-based research in ophthalmology.
18| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI in ophthalmic research remains in pilot and early production phases, concentrated in data analysis and image processing within academic centers. Most ophthalmic research is still conducted in traditional ways with limited AI integration in core research operations, reflecting the sector's slower digitization compared to tech/finance. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Academic medicine and biomedical research adopt AI tools (e.g., for data analysis, image processing) at a moderate pace, but full research task automation remains rare and pilot-stage. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments ophthalmology research through automated image analysis of retinal scans, literature synthesis, statistical analysis, and hypothesis generation from large datasets. These tools raise researcher productivity substantially while keeping human researchers in control of study design, interpretation, and scientific judgment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids literature synthesis, statistical analysis, image analysis, and manuscript drafting, meaningfully boosting researcher productivity while humans retain control of study design and clinical judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Clinical and laboratory-based ophthalmic research involves complex experimental design, hypothesis formulation, and interpretation of biological data that require human judgment. AI can assist with data analysis and literature review, but cannot independently conduct experiments, manage research protocols, or make the iterative theoretical decisions that define research work. |
| Task automatability | claude-sonnet-5 | 1/5 | Clinical/laboratory ophthalmology research requires original experimental design, hypothesis generation, hands-on lab or clinical trial execution, and interpretation that AI cannot autonomously perform end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Research involving human subjects or novel therapeutic targets faces regulatory requirements (IRB approval, FDA oversight for certain studies) that mandate human investigator responsibility and accountability. Publishing and funding agencies require human researchers to take ownership of research conduct and conclusions, creating hard barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Research involving human subjects or clinical data requires IRB approval, licensed physician oversight, and regulatory compliance, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Conducting ophthalmic research at the quality and novelty required by the field demands high-skill human expertise; even sophisticated AI tools require significant oversight, validation, and integration costs. The all-in cost of AI-assisted research remains comparable to or exceeds the cost of human researchers who bring deep domain knowledge. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply assist with literature review, data analysis, or drafting, but the core research (experiments, trials, physical lab work) still requires expensive human specialists, so overall cost savings are limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI tools exist for data analysis, image processing, and literature mining in medical research, no deployed product reliably conducts end-to-end ophthalmic research. Production systems handle narrow subtasks (image segmentation, statistical analysis) but not the full research workflow including experiment design, hypothesis testing, and results interpretation. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product conducts independent ophthalmology research; AI is used only as a tool within researcher-led workflows, not as an autonomous investigator. |
Develop or implement plans and procedures for ophthalmologic services.
14CI 3–25 · exposure 13 · augmentation 50 · importance 4.7/5 · click for rater detail
Develop or implement plans and procedures for ophthalmologic services.
14| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While healthcare systems are digitizing operations, adoption of AI-driven service planning specifically remains limited and cautious. Most implementations are pilots or administrative support; genuine AI-led protocol development is rare and typically involves human review at every step. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare administration and clinical governance adopt AI slowly due to regulatory, liability, and safety concerns, with pilots more common than deployed autonomous planning tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by analyzing best practices, comparing benchmarks across institutions, drafting procedure templates, and flagging gaps in existing plans. However, clinical validation and strategic decision-making remain physician-led, limiting the transformative impact to moderate productivity gain. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist by summarizing best practices, benchmarking data, or drafting protocol documents, but the physician must still evaluate, adapt, and implement them. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Developing procedural plans requires understanding clinical protocols, regulatory compliance, resource allocation, and organizational strategy—tasks involving substantial human judgment. While AI could draft templates or suggest standard protocols, the integration of medical liability, institutional context, and professional standards requires human oversight that prevents ≥50% time savings on the full task. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires clinical leadership, organizational judgment, and integration of medical, regulatory, and business considerations that AI cannot autonomously perform end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong regulatory and liability barriers exist: ophthalmologists must personally design and validate clinical service plans; malpractice and regulatory frameworks require physician accountability. Hospital credentialing and accreditation standards require licensed oversight of service procedures. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Developing clinical service procedures typically requires licensed physician oversight, institutional accreditation standards, and legal accountability, creating strong barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI assistance (templating, analysis) could reduce some planning overhead, but the task requires expert ophthalmologist time for final design and implementation sign-off. The loaded cost of a physician's involvement means the human labor component remains dominant; AI is not an order of magnitude cheaper. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this task, so cost comparison favors the human physician who must ultimately develop and implement such plans. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably develops or implements ophthalmologic service plans end-to-end. Generic workflow tools and documentation systems exist, but none demonstrably handle the clinical, regulatory, and operational complexity of ophthalmology services planning in production healthcare settings. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product plans or implements ophthalmology service protocols independently; such work remains research-stage or entirely human-driven. |
Provide ophthalmic consultation to other medical professionals.
11CI 3–20 · exposure 13 · augmentation 63 · importance 4.5/5 · click for rater detail
Provide ophthalmic consultation to other medical professionals.
11| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Ophthalmology has begun adopting AI for image screening and triage, but adoption of AI for autonomous consultation to other physicians remains rare. The specialty is digitizing but remains highly human-centered and resistant to algorithmic delegation of clinical judgment. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially specialist consultation, adopts AI slowly due to regulatory, liability, and workflow integration constraints, with pilots more common than production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist ophthalmologists in consultation by quickly analyzing fundus or anterior segment images, summarizing relevant literature, or flagging critical findings. These tools do raise efficiency on parts of the consultation workflow, though the core advisory role remains human-driven. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can meaningfully assist by analyzing imaging, flagging abnormalities, summarizing patient history, and suggesting differential diagnoses, enhancing the ophthalmologist's consultative efficiency. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Providing ophthalmic consultation requires integrating patient history, examination findings, and complex clinical judgment to advise other physicians. While AI can assist in image analysis and preliminary assessment, the real-time interactive consultation role—involving live questioning, nuanced reasoning about comorbidities, and responsibility for clinical recommendations—cannot be performed end-to-end by current systems at equal quality with 50% time savings. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires synthesizing complex clinical exam findings, imaging, and patient context into judgment-based recommendations that carry legal and clinical responsibility; no current system can perform this end-to-end reliably.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Strong regulatory and liability barriers exist: consultations require a licensed physician to examine, evaluate, and sign off on recommendations to other clinicians. Medical-legal responsibility for diagnostic and treatment advice cannot be delegated to an unlicensed AI system. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Consultation between medical professionals is a licensed medical activity requiring a physician's judgment and legal accountability, making this a hard regulatory and liability barrier. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The infrastructure, integration, oversight, and liability management required for AI to meaningfully substitute for ophthalmic consultation would be costly and unproven, while ophthalmologist consultation is already a standard billable service. AI cost per task is unlikely to undercut loaded physician wages. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI inference is cheap, the human oversight, liability, and specialist judgment required make the effective all-in cost comparable to or only marginally cheaper than physician consultation time given current tool limitations. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products reliably perform full ophthalmic consultations autonomously in production. AI tools exist for image grading and triage suggestions, but these are narrow decision-support aids, not autonomous consultants. A licensed ophthalmologist must still integrate findings and own the clinical recommendation. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently provides ophthalmic consultations to other physicians in production; AI diagnostic tools exist for narrow tasks (e.g., diabetic retinopathy screening) but not full consultative reasoning. |
Instruct interns, residents, or others in ophthalmologic procedures and techniques.
10CI 0–20 · exposure 13 · augmentation 50 · importance 3.4/5 · click for rater detail
Instruct interns, residents, or others in ophthalmologic procedures and techniques.
10| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Despite some pilots of simulation and asynchronous educational content, teaching of ophthalmologic procedures remains anchored to live faculty oversight and residency program accreditation standards that require human instruction, limiting rapid AI substitution. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare training programs adopt AI slowly for hands-on clinical instruction, with adoption limited to simulation aids rather than replacing supervisory teaching. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist instructors by generating procedure animations, pre-op simulations, or automatically flagging technique deviations in recorded video, thereby reducing preparation time and enriching trainee learning—but the instructor remains essential to the teaching loop. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-based simulators, video analysis, and surgical planning tools can supplement teaching materials and feedback, but the instructor role stays central. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Teaching complex surgical procedures requires real-time mentorship, adaptive feedback, and judgment of trainee competence that current AI systems cannot reliably provide end-to-end. While AI could assist with generating instructional content or pre-recorded procedure explanations, the interactive, corrective, and relationship-driven aspects of clinical instruction remain beyond meaningful automation. |
| Task automatability | claude-sonnet-5 | 1/5 | Clinical teaching involves live demonstration, hands-on supervision, mentorship, and real-time judgment during procedures that current AI cannot replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Regulatory and legal requirements (GME oversight, hospital credentialing, malpractice liability) mandate that instruction and sign-off of resident competence be performed by licensed, credentialed physicians; AI cannot assume this fiduciary or legal responsibility. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Medical education and supervision of trainees performing invasive procedures require licensed physician oversight and legal accountability, making this a hard-barrier task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI systems to generate instructional content or simulations does not offset the high salary of an experienced ophthalmologist instructor, nor does it eliminate the need for that instructor to supervise trainees in high-liability clinical settings. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute delivering equivalent supervisory teaching, so cost comparison favors the human attending entirely. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably substitutes for a qualified ophthalmologist instructor in clinical training environments. AI-generated educational videos and simulations exist but do not replace the live supervision, error correction, and credentialing function that an attending physician performs during training. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs supervised clinical instruction of trainees in ophthalmologic procedures; this remains firmly a human faculty responsibility in production medical training. |
Prescribe ophthalmologic treatments or therapies such as chemotherapy, cryotherapy, or low vision therapy.
9CI 3–16 · exposure 13 · augmentation 50 · importance 4.2/5 · click for rater detail
Prescribe ophthalmologic treatments or therapies such as chemotherapy, cryotherapy, or low vision therapy.
9| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Ophthalmology has adopted imaging AI for diagnosis, but therapeutic prescription remains physician-controlled. Adoption of AI-assisted treatment planning is slow because of regulatory constraints, liability concerns, and physician resistance to ceding prescription authority. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially specialized clinical decision-making, has been slow to adopt autonomous AI tools; adoption is limited to diagnostic support and imaging, not treatment prescription. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can augment prescription by suggesting evidence-based treatment protocols, summarizing patient data, or flagging contraindications, but the physician retains decision authority and remains the bottleneck, offering moderate productivity gains rather than transformation. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help synthesize diagnostic imaging, suggest treatment options based on guidelines, and flag drug interactions, providing meaningful support to physicians while they retain final decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Prescribing treatments requires complex clinical judgment integrating patient history, examination findings, and individualized risk–benefit assessment. While AI can assist in pattern recognition, the legal and medical requirement for physician sign-off, and the high consequence of error, means current systems cannot autonomously perform this end-to-end at the quality threshold. |
| Task automatability | claude-sonnet-5 | 1/5 | Prescribing treatment requires integrating clinical exam findings, patient history, risk tolerance, and legal accountability into a personalized decision; no AI system today performs this end-to-end.showable time savings are minor relative to the judgment and liability involved. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Prescribing medications and therapies is legally restricted to licensed physicians; regulatory frameworks (FDA, state medical boards) explicitly require physician oversight of treatment decisions, and liability falls on the prescribing physician, creating hard legal barriers to autonomous automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Prescribing medical treatment is legally restricted to licensed physicians, with direct malpractice liability and regulatory oversight of drug and treatment authorization, making this a hard-barrier task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The cost of AI infrastructure, clinical validation, compliance, and mandatory physician oversight approaches or exceeds the marginal cost of physician time already required for prescription authority, especially for specialized ophthalmologic treatments. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | Since AI cannot independently perform this task, there is no valid cost substitution; any AI assistance still requires full physician time and liability coverage, so no cost savings materialize at the task level. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Decision-support tools exist to suggest treatment options based on clinical data, but no deployed product reliably prescribes these specific therapies without a licensed physician making the final determination and handling individual patient factors. Prescription authority remains a physician function in production practice. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously prescribes ophthalmologic treatments; clinical decision-support tools exist only as advisory aids reviewed by physicians. |
Prescribe or administer topical or systemic medications to treat ophthalmic conditions and to manage pain.
7CI 0–15 · exposure 8 · augmentation 63 · importance 4.7/5 · click for rater detail
Prescribe or administer topical or systemic medications to treat ophthalmic conditions and to manage pain.
7| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Adoption of autonomous AI prescribing in ophthalmology is negligible because legal and regulatory frameworks explicitly require a licensed physician to prescribe. No sector has moved toward delegating prescriptive authority to AI; the profession remains gatekept by licensure requirements. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially direct prescribing/treatment decisions, adopts AI cautiously due to regulation and malpractice risk, with pilots for diagnostics far outpacing prescribing automation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can augment prescribing decisions by suggesting evidence-based treatment options, summarizing contraindications, or flagging drug interactions based on patient history. However, the ophthalmologist remains the primary decision-maker; the assistance is moderate because diagnosis and patient-specific judgment still rest with the human. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by flagging drug interactions, suggesting dosing based on guidelines, and summarizing relevant literature, improving physician efficiency while they retain prescribing authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist in medication selection based on diagnostic findings and clinical guidelines, but the task requires diagnosis of the underlying condition first, patient-specific medical history evaluation, and legal prescription authority. Current systems cannot independently diagnose eye conditions or legally prescribe medications; this requires a licensed physician's judgment and signature. |
| Task automatability | claude-sonnet-5 | 1/5 | Prescribing medication requires clinical judgment, physical exam findings, patient history integration, and legal authority that current AI cannot independently exercise end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Prescribing medications is a legally restricted activity requiring a licensed ophthalmologist's credentials and signature. Medical liability, regulatory frameworks (FDA, state licensing boards), and malpractice risk create hard barriers that prevent substitution by AI systems without human physician oversight and sign-off. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Prescribing controlled and systemic medications is legally restricted to licensed physicians; strict regulatory, licensing, and liability requirements prevent AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI systems cannot legally or reliably substitute for physician prescribing. The cost comparison is moot because this task cannot be automated end-to-end; human ophthalmologists must retain responsibility for all prescription decisions, making the human cost irreducible. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI decision-support has low marginal inference cost, but since a licensed physician must still evaluate, prescribe, and bear liability, overall cost savings versus the human process are minimal today. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product can independently prescribe medications for ophthalmologic conditions. While AI diagnostic aids exist for some eye conditions, they do not perform the full prescribing task, which requires legal authority and integration with patient records that only licensed physicians can execute in practice. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously prescribes or administers ophthalmic medications; clinical decision-support tools exist but require physician sign-off and action. |
Perform ophthalmic surgeries such as cataract, glaucoma, refractive, corneal, vitro-retinal, eye muscle, or oculoplastic surgeries.
5CI 3–7 · exposure 5 · augmentation 50 · importance 4.7/5 · click for rater detail
Perform ophthalmic surgeries such as cataract, glaucoma, refractive, corneal, vitro-retinal, eye muscle, or oculoplastic surgeries.
5| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of surgical robots in ophthalmology is nascent and limited to specialized centers; most procedures remain manually performed by surgeons. Industry adoption is slow relative to other healthcare domains, with pilot deployments but minimal production displacement to date. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Surgical AI/robotics adoption in ophthalmology is slow and limited to adjunct tools (e.g., femtosecond laser cataract systems) rather than autonomous task replacement, reflecting a physically-grounded, highly regulated sector. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-assisted surgical platforms can augment surgeons with real-time imaging overlay, eye-tracking, and motion stabilization, improving precision and reducing fatigue on complex cases. However, the surgeon remains the primary decision-maker and executor, with assistance on specific subtasks rather than transformation of overall productivity. |
| Augmentation potential | claude-sonnet-5 | 3/5 | Image-guided systems, femtosecond lasers, and surgical planning software assist surgeons in precision tasks like corneal incisions or lens calculations, improving outcomes while the surgeon remains fully in control. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Ophthalmic surgeries require real-time 3D spatial reasoning, manual dexterity at microscopic scales, and rapid intraoperative decision-making in response to unpredictable tissue responses. Current AI systems cannot perform surgery end-to-end; robotic surgical systems exist but require continuous human control and oversight. |
| Task automatability | claude-sonnet-5 | 1/5 | Surgical execution requires physical dexterity, real-time tactile feedback, and judgment under variable anatomy that no current AI system can perform end-to-end; robotic assistance exists but does not replace the surgeon.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Medical licensing and malpractice liability require a licensed ophthalmologist to perform or directly oversee surgery; regulatory bodies (FDA, medical boards) mandate human accountability for patient safety. Legal and liability frameworks create hard barriers to autonomous surgical automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Ophthalmic surgery is a licensed medical procedure with strict regulatory, liability, and credentialing requirements mandating a qualified physician perform and be accountable for the operation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Surgical robotic systems and AI-assisted platforms require high capital investment, ongoing maintenance, and specialized training. The all-in cost per procedure remains substantially above the loaded wage of an ophthalmologist for the task output, especially given the need for human supervision. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | Surgical robots and laser platforms are capital-intensive and still require a licensed surgeon present, so AI does not reduce cost relative to the human physician performing the surgery. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While some robotic platforms (e.g., ROBOT, Lensar) assist with specific surgical steps or pre-surgical planning, no deployed AI system independently performs complete ophthalmic surgeries at production scale. Existing tools require surgeon oversight and manual intervention, falling short of reliable autonomous performance. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously performs ophthalmic surgery; femtosecond lasers and robotic-assist tools support but do not replace surgeon-led procedures in production settings. |
Perform laser surgeries to alter, remove, reshape, or replace ocular tissue.
5CI 3–7 · exposure 5 · augmentation 63 · importance 4.5/5 · click for rater detail
Perform laser surgeries to alter, remove, reshape, or replace ocular tissue.
5| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Although some surgical centers use robotic-assisted platforms for certain procedures, adoption remains limited to high-end facilities and concentrated in research/early-adoption phases rather than mainstream clinical practice displacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Surgical specialties adopt AI-assisted tools (image guidance, planning software) gradually, but full autonomous surgical execution has not entered clinical practice at any scale. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-assisted imaging and surgical planning tools can help surgeons visualize anatomy and plan trajectories, improving preparation and execution, but the surgeon remains the active decision-maker and controller throughout the critical phases of laser tissue manipulation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI and computer-guided laser systems already assist surgeons with precision planning, real-time tracking, and pre-surgical diagnostics, meaningfully improving outcomes while the surgeon remains in control. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Laser surgery on the eye requires real-time adaptive control, micro-precision targeting of delicate tissue, and immediate response to unforeseen complications. Current AI cannot perform end-to-end surgical procedures autonomously with the precision and safety margin required for ocular tissue manipulation. |
| Task automatability | claude-sonnet-5 | 1/5 | This is hands-on invasive surgery requiring physical dexterity, real-time judgment, and manipulation of surgical equipment on a living patient; no AI system performs this task end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Surgical procedures on the human eye are legally and ethically bound to licensed physician supervision and direct control; liability, regulatory approval (FDA), and malpractice exposure create hard barriers to autonomous or minimally supervised automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Ocular laser surgery requires a licensed ophthalmologist to perform and bear legal/medical liability; regulatory, licensing, and safety requirements make autonomous AI performance essentially prohibited. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Ophthalmologic laser surgery equipment, maintenance, integration with AI systems, and required human oversight costs substantially exceed the loaded wage of performing the surgery manually, with current technology. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this task independently, so no meaningful cost comparison exists; surgical lasers are tools operated by humans, not autonomous replacements. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While surgical robots exist and AI-assisted vision systems help with planning, no deployed product autonomously performs complete laser eye surgeries in production. Existing systems require continuous human surgeon control and decision-making throughout the procedure. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | While laser platforms (e.g., femtosecond lasers for LASIK/cataract) have automated components and image-guided targeting, the surgeon performs and controls the procedure; no deployed product autonomously performs the surgery. |
Collaborate with multidisciplinary teams of health professionals to provide optimal patient care.
3CI 0–6 · exposure 0 · augmentation 38 · importance 4.3/5 · click for rater detail
Collaborate with multidisciplinary teams of health professionals to provide optimal patient care.
3| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Because collaboration is inherently human-centric and legally required to be performed by licensed professionals, there is minimal adoption of AI automation in this dimension. Organizational practices in healthcare reinforce the necessity of direct physician-to-physician and physician-to-team coordination. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare adoption of AI for clinical collaboration remains slow and pilot-stage, constrained by regulation, liability, and interoperability challenges. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could potentially assist with summarizing medical records or flagging relevant information to support collaboration, but the act of collaborating itself—building consensus, negotiating care plans, integrating expertise—remains primarily a human function that AI augments only peripherally. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI tools like shared EHR summaries, decision-support alerts, and documentation aids can meaningfully support communication and information-sharing among care teams, though the core collaboration remains human-driven. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Collaboration with multidisciplinary health teams requires nuanced interpersonal communication, shared decision-making, and consensus-building that AI cannot meaningfully perform autonomously. This is fundamentally a human coordination and judgment task where an AI system cannot substitute for professional engagement. |
| Task automatability | claude-sonnet-5 | 1/5 | Interdisciplinary collaboration requires real-time human judgment, negotiation, and relationship-based communication that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Regulatory and liability frameworks require licensed ophthalmologists to personally participate in clinical decision-making and team collaboration; no non-physician AI can legally assume this role. Professional responsibility and standard of care create hard barriers to substitution. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Medical licensure, liability for clinical decisions, and legal requirements for physician sign-off create hard barriers preventing AI substitution in care coordination. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | While AI tools might support documentation or information sharing, the core collaboration task itself cannot be substantially cost-reduced through automation. The human physician must remain the active collaborator, limiting economic displacement. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute for this collaborative role, so cost comparison favors the human by default. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product today can autonomously collaborate with multidisciplinary teams on behalf of a physician or independently coordinate care decisions. This requires legal accountability, professional relationships, and real-time human negotiation that remain outside current AI capabilities. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product autonomously coordinates or participates as a team member in multidisciplinary patient care decisions. |
Provide or direct the provision of postoperative care.
3CI 3–3 · exposure 0 · augmentation 50 · importance 4.8/5 · click for rater detail
Provide or direct the provision of postoperative care.
3| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Ophthalmology has modest AI adoption in diagnostic imaging and surgical planning, but postoperative care remains a human-directed, relationship-intensive domain with low velocity of substitution due to clinical and legal requirements. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially surgical specialties, has been slow to adopt AI for hands-on clinical care despite faster uptake in administrative and diagnostic support tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist through automated complication detection in imaging, documentation support, and clinical reminder systems, but the core task of directing and providing postoperative care remains physician-centric with only partial augmentation potential. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with symptom triage, patient education, appointment scheduling, and documentation of postoperative visits, improving efficiency without replacing the clinician's direct care role. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Postoperative care requires real-time clinical assessment, physical examination, medication titration, and judgment about complication detection—tasks demanding human sensory and decision-making capabilities that current AI cannot perform end-to-end. No current system can autonomously manage post-surgical patient care without direct human oversight. |
| Task automatability | claude-sonnet-5 | 1/5 | Postoperative care for ophthalmic surgery requires hands-on eye examination, judgment about healing complications, and physical adjustments (e.g., suture removal, medication changes) that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Postoperative care is a core clinical function that must be directed by and legally attributable to a licensed ophthalmologist; medical liability, patient safety regulations, and informed consent requirements create hard legal and regulatory barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Postoperative care after eye surgery legally requires a licensed physician's oversight and physical examination, with high liability for missed complications like infection or detachment. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The task requires physician-level expertise and legal accountability; AI systems cannot substitute for the physician's liability and judgment, making the cost comparison academic—the human physician remains mandatory. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this task, so cost comparison favors the human physician entirely; AI cannot replace the clinical exam and decision-making involved. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | While AI can assist with documentation and decision support, no deployed product reliably manages postoperative ophthalmologic care independently; care coordination, patient assessment, and clinical decision-making remain human-centric in all production settings. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product directs or performs postoperative surgical follow-up care; at most AI assists with documentation or scheduling reminders, not clinical care delivery. |
Refer patients for more specialized treatments when conditions exceed the experience, expertise, or scope of practice of practitioner.
1CI 0–3 · exposure 0 · augmentation 38 · importance 4.5/5 · click for rater detail
Refer patients for more specialized treatments when conditions exceed the experience, expertise, or scope of practice of practitioner.
1| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Ophthalmology practices remain cautious about ceding clinical judgment to AI; even diagnostic support tools see limited uptake outside academic and large healthcare systems, and referral decisions—a core fiduciary act—show minimal AI adoption. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare, especially specialist medical decision-making, is a slow-adopting, highly regulated sector where AI is used for decision support but not autonomous referral action. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could marginally assist by flagging cases with features suggesting specialist consultation (e.g., complex refractive or retinal pathology), but the task itself is already rapid and driven by expertise; augmentation value is limited. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI diagnostic tools and decision-support systems can help flag abnormal findings or suggest specialist referral thresholds, aiding but not replacing the ophthalmologist's judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Referral decisions require nuanced clinical judgment integrating patient history, examination findings, treatment goals, and knowledge of specialist expertise—judgments that demand human accountability and cannot be reliably automated to meet the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a clinical judgment and professional referral decision requiring licensed medical authority; AI cannot legally or practically execute this end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Ophthalmologists are licensed professionals whose scope and standard of care include determining when specialty referral is necessary; liability and regulatory expectations place this decision squarely on the physician, and malpractice exposure creates a hard barrier to substitution. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Referral decisions are a core licensed medical function with direct liability implications, requiring a physician's legal scope-of-practice judgment—among the strongest barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The task is brief (a referral decision and communication) with minimal cost; any AI system would require clinical validation, oversight, and integration overhead that would exceed the time and cost savings of automating a few minutes of physician judgment. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | The task must still be performed by a licensed physician regardless of AI assistance, so no cost substitution occurs; AI tools add cost as adjuncts rather than replacing the decision-maker. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system reliably makes clinical referral decisions in production ophthalmology practice; existing diagnostic aids support image analysis but do not replace the physician's discretionary judgment on specialist appropriateness. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently makes referral decisions for ophthalmologists; at best AI flags cases for physician review, but the referral act itself is not performed by AI in production. |
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.