Materials Scientists
19-2032.00Median wage $117,790/yr8,470 employed (US)Rank #410 of 923 scored · top 44% by substitution
Research and study the structures and chemical properties of various natural and synthetic or composite materials, including metals, alloys, rubber, ceramics, semiconductors, polymers, and glass. Determine ways to strengthen or combine materials or develop new materials with new or specific properties for use in a variety of products and applications. Includes glass scientists, ceramic scientists, metallurgical scientists, and polymer scientists.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution29
Exposure27
Augmentation67
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
16 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%27
panel mean rating 2.1/5 → substitution pressure 27/100
Technical feasibility todayw 20%27
panel mean rating 2.1/5 → substitution pressure 27/100
Cost vs. human wagew 15%28
panel mean rating 2.1/5 → substitution pressure 28/100
Adoption barriersw 20%inverted — strong barriers lower the score38
panel mean rating 3.5/5 (barrier strength) → substitution pressure 38/100
Sector adoption velocityw 10%28
panel mean rating 2.1/5 → substitution pressure 28/100
Task breakdown (16 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.
Write research papers for publication in scientific journals.
69CI 50–87 · exposure 70 · augmentation 100 · importance 3.5/5 · click for rater detail
Write research papers for publication in scientific journals.
69| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Materials science and academic research sectors show rapid, visible adoption of AI writing tools; major universities, national labs, and research institutions integrate these tools into workflows. Displacement is measurable in reduced time-to-draft and widespread pilot/production adoption among researchers. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Academic and R&D sectors show growing but uneven adoption of AI writing tools, with many researchers experimenting but institutional and journal policies lagging behind actual practice. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically accelerates the writing phase while researchers remain in control of technical accuracy, framing, and novelty claims. This creates substantial productivity lift—researchers can iterate faster, manage multiple drafts, and focus on scientific judgment rather than prose composition. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially speeds up drafting, editing, formatting, and literature summarization, while the scientist retains control over experimental design, interpretation, and final claims. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Large language models can generate complete, coherent research papers with methodology, results, and discussion sections from data and outlines, achieving substantial time savings (>50%) for many manuscript components. End-to-end paper writing including literature synthesis, structure, and prose is within current AI capabilities at production quality. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft literature reviews, structure sections, and generate prose from provided data and results, but synthesizing novel findings, ensuring scientific accuracy, and framing contributions still require substantial human expertise and revision. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No legal requirement mandates human authorship, though journals increasingly require disclosure of AI use and human accountability for content accuracy. Peer review and journal submission policy still require human final approval, creating modest friction but not a blocking barrier. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but journal policies increasingly mandate disclosure of AI use and human accountability for scientific claims, and reputational/authorship norms create friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | API costs for generating a full research paper draft are typically $1–10 in inference, compared to 40+ hours of materials scientist labor at $60–100/hour loaded cost. AI is easily an order of magnitude cheaper for the writing component alone. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI writing tools are cheap per use, but the overall task still requires significant scientist time for data interpretation, verification, and revision, making net savings moderate rather than order-of-magnitude. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed systems (ChatGPT, Claude, specialized research assistants) are actively used to draft, revise, and structure research papers in real workflows. While human review and correction of technical details remain standard practice, the AI-generated output is reliable enough for material reuse without complete rewriting. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | LLM-based writing assistants (e.g., ChatGPT, Grammarly, Writefull) are widely used by scientists for drafting and editing, but fully autonomous paper generation without expert oversight is not reliable in production due to hallucination and citation risks. |
Prepare reports, manuscripts, proposals, and technical manuals for use by other scientists and requestors, such as sponsors and customers.
53CI 48–59 · exposure 50 · augmentation 88 · importance 4.1/5 · click for rater detail
Prepare reports, manuscripts, proposals, and technical manuals for use by other scientists and requestors, such as sponsors and customers.
53| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials science and academic/research sectors show cautious, slow adoption of AI writing assistants compared to information technology or professional services. Pilots exist, but production deployment at scale remains limited by researcher skepticism and institutional conservatism around scientific authorship. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | R&D and scientific documentation workflows are adopting AI writing tools moderately, with pilots and partial integration common but full-scale reliance still limited by accuracy concerns. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI tools meaningfully assist scientists by accelerating first-draft creation, formatting, literature summarization, and editing feedback, substantially raising productivity when the human scientist retains final judgment on technical content and claims. This assistance is now widely observed in practice. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI is highly effective at drafting outlines, improving clarity, summarizing data, and formatting technical documents, substantially speeding up the writing process while scientists retain oversight of technical accuracy. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can draft significant portions of technical reports and manuscripts—literature review sections, methodology descriptions, results summaries—but typically requires expert review and revision of conclusions, implications, and novel findings. The task involves both routine documentation (automatable) and judgment-dependent synthesis (not automatable to 50% time-saving at equal quality without human oversight). |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft substantial portions of reports and proposals from provided data and outlines, but accurate technical synthesis, novel scientific claims, and correct interpretation of experimental data still require significant human authorship and verification. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | No legal requirement exists for a licensed scientist to author reports, but organizational norms, accountability expectations, and sponsor/customer preference for human expertise create moderate friction. Scientific reputation and error liability also slow adoption despite technical feasibility. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI-assisted writing, though sponsors/customers may require credentialed scientist sign-off and proprietary data handling adds some friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Inference cost for generating technical drafts is minimal (cents per use), and even with integration overhead and human review time, the cost per task-equivalent is substantially lower than employing a scientist to write reports from scratch. The ratio favors AI by a meaningful margin, though not yet an order of magnitude. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI drafting tools are cheap per query, but the need for expert review, fact-checking, and iterative revision by a materials scientist keeps overall cost comparable rather than dramatically lower. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Large language models and generative AI tools can produce competent technical drafts, but deployed products show notable limitations in accuracy of technical detail, proper citation, novel scientific claims, and integration of complex experimental data. Production use is growing in some organizations but remains heavily human-supervised. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | LLM writing assistants are widely used in R&D organizations for drafting and editing technical documents, but reliability drops for highly technical, data-dense materials science content requiring domain accuracy and citation integrity. |
Supervise and monitor production processes to ensure efficient use of equipment, timely changes to specifications, and project completion within time frame and budget.
31CI 25–37 · exposure 30 · augmentation 75 · importance 4.0/5 · click for rater detail
Supervise and monitor production processes to ensure efficient use of equipment, timely changes to specifications, and project completion within time frame and budget.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Manufacturing and materials sectors show moderate AI adoption in monitoring and predictive maintenance, but primarily as assistive tools within human-led supervision. Full autonomous production oversight remains rare; pilots outpace production deployments in this domain. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Manufacturing and materials science sectors are generally slower adopters of AI compared to information-based industries, with pilots for predictive monitoring more common than full deployment of autonomous supervision. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-powered dashboards, anomaly detection, and predictive alerts substantially boost a materials scientist supervisor's ability to catch problems early and optimize resource allocation. The human retains final authority while AI provides real-time intelligence, creating strong productivity uplift. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered analytics, predictive maintenance, and process monitoring tools can meaningfully assist materials scientists in tracking equipment efficiency and flagging issues, improving decision-making while humans remain in control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can monitor equipment sensors and flag specification deviations in real-time, supervising a production process end-to-end requires dynamic decision-making, resource allocation, and personnel coordination that remain largely manual. AI might automate ~20-30% of the monitoring component but cannot independently manage timeline/budget trade-offs or handle unexpected process failures. |
| Task automatability | claude-sonnet-5 | 2/5 | This is a hands-on supervisory and coordination role requiring physical presence on production floors, real-time judgment, and accountability that current AI cannot fully replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Production supervision carries significant liability for equipment damage, safety violations, and budget overruns, creating strong organizational and legal incentive to retain human sign-off. Regulatory frameworks (e.g., FDA for pharmaceutical manufacturing) often mandate human accountability for process control decisions. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Production oversight often involves organizational accountability, safety and quality compliance, and liability for decisions, creating moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Real-time monitoring infrastructure (sensors, inference, dashboards, integration) is expensive to implement and maintain, while the human supervisor's salary is often justified by liability and decision-making authority. Cost parity exists only for narrow routine-monitoring subtasks, not the full supervisory role. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While sensors and software can reduce some monitoring labor, the human supervisory and decision-making component still requires significant staffing, keeping AI cost savings modest relative to the full task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Production monitoring software and IoT dashboards exist and are deployed, but they function as alert systems for human supervisors rather than autonomous supervisors. AI can detect anomalies and suggest parameter adjustments with material error rates in novel scenarios; no mature system replaces the supervisory role end-to-end. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Manufacturing execution systems and analytics dashboards exist to support monitoring, but no deployed product autonomously supervises production processes and makes specification changes reliably without human oversight. |
Test material samples for tolerance under tension, compression, and shear to determine the cause of metal failures.
30CI 25–35 · exposure 30 · augmentation 63 · importance 4.2/5 · click for rater detail
Test material samples for tolerance under tension, compression, and shear to determine the cause of metal failures.
30| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials science remains a traditional, highly specialized field with limited digital transformation and slow AI adoption outside large manufacturing firms. Most materials testing is still performed in academic and smaller industrial labs where capital investment in automation lags finance or software sectors. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science and manufacturing sectors adopt AI more slowly than information/finance sectors, with physical testing remaining largely manual or semi-automated with legacy equipment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-powered image analysis and statistical tools can assist materials scientists in identifying patterns in microscopy images and test data, but the scientist must remain central to interpreting failure modes and proposing causal mechanisms. AI functions as a data accelerant rather than a productivity transformer. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly assist in analyzing test data, predicting failure modes, and correlating results with material properties, boosting scientist productivity in interpretation and reporting. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While automated testing machines can perform standardized tension, compression, and shear tests on material samples, the root cause analysis of metal failures requires expert judgment, visual metallurgical inspection, and interpretation of complex failure modes that current AI cannot reliably diagnose end-to-end. The testing itself is automatable but only represents a fraction of the full task. |
| Task automatability | claude-sonnet-5 | 2/5 | Physical testing (tension, compression, shear) requires lab equipment, sample preparation, and instrumented machines; AI can assist in analysis but cannot perform the physical testing itself end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Materials testing for critical applications (aerospace, medical devices, structural) faces regulatory requirements, liability constraints, and ISO/ASTM standards that mandate qualified human judgment and sign-off on failure root cause determinations. Testing results inform safety-critical decisions where human expertise and professional responsibility are legally and ethically required. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement blocks automation, but liability for failure analysis in critical applications (aerospace, construction) creates strong incentive for human oversight and validation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Automated testing equipment requires significant capital investment and specialized integration into laboratory workflows. Data analysis overhead and the need for expert human review of results means total cost of AI-assisted testing remains comparable to, or potentially higher than, traditional hands-on materials scientist labor. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Physical testing equipment, materials, and lab time are costly regardless of AI involvement, and AI only reduces a portion of the analysis time, not the equipment/materials cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Mechanical testing equipment with digital data collection exists and is widely deployed in labs, but AI systems for autonomous failure diagnosis remain experimental and unreliable in production. Existing products can capture and organize test data but lack the capability to independently determine failure causes at acceptable accuracy. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Automated testing rigs exist but require human setup and calibration; AI-driven failure analysis exists in research settings but isn't broadly deployed as a full replacement. |
Conduct research on the structures and properties of materials, such as metals, alloys, polymers, and ceramics, to obtain information that could be used to develop new products or enhance existing ones.
29CI 25–32 · exposure 25 · augmentation 75 · importance 4.4/5 · click for rater detail
Conduct research on the structures and properties of materials, such as metals, alloys, polymers, and ceramics, to obtain information that could be used to develop new products or enhance existing ones.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While computational tools are widely used in materials science, adoption of AI agents to conduct or lead research is still nascent and predominantly in academic/pilot settings. Most industrial materials development remains human-driven with AI as a supporting tool rather than a replacement agent. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Materials science and R&D-intensive manufacturing sectors are adopting AI/ML tools (generative design, property prediction) at a moderate pace, with pilots more common than full production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially augments materials scientists through computational modeling, property prediction, literature synthesis, and data analysis—allowing researchers to explore larger design spaces and reduce experimental iterations. The human remains central to hypothesis formation, experimental design, and validation, but AI meaningfully amplifies their productivity. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids materials scientists via machine learning-based property prediction, simulation acceleration, literature synthesis, and generative design of candidate materials, substantially boosting research productivity while humans validate and interpret results. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with literature review, data analysis, and computational modeling of material properties, the core task of hands-on experimental research, lab work, and novel discovery synthesis requires human experimentation and physical testing that cannot be end-to-end automated today. Current systems cannot reliably design, execute, and interpret new experimental protocols at the speed and quality needed for genuine R&D. |
| Task automatability | claude-sonnet-5 | 2/5 | Core research involves physical experimentation, hypothesis generation, and novel material characterization that current AI cannot perform end-to-end, though it can accelerate literature review and data analysis subcomponents. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Materials research is conducted primarily within organizations with established R&D governance, institutional review boards, and safety protocols. The requirement for novel experimental design, liability for material safety/performance, and organizational culture around human expertise create substantial friction against full automation or substitution. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement per se, but organizational reliance on validated experimental results, safety-critical applications (aerospace, medical materials), and IP/quality control create moderate friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for materials research (simulation software, ML models) have significant upfront and ongoing costs, while still require highly trained PhD-level researchers to direct and validate results. The loaded cost of a materials scientist remains lower per unit of novel discovery than current AI infrastructure can deliver, especially when integration and human oversight overhead is included. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Physical testing, synthesis, and lab equipment costs dominate and AI does not replace these; software assistance offers modest cost savings on data analysis but not the full research cycle. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some deployed tools exist for computational materials prediction and literature mining, but they lack the breadth and integration to handle the full research pipeline reliably. Most materials scientists still rely on manual experimentation, literature review, and judgment; no mature, production-scale system performs the complete task of novel materials research without substantial human oversight and direction. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools (e.g., materials informatics platforms) assist with property prediction and screening but no deployed product autonomously conducts full materials research programs reliably in production. |
Determine ways to strengthen or combine materials or develop new materials with new or specific properties for use in a variety of products and applications.
28CI 25–30 · exposure 25 · augmentation 75 · importance 4.2/5 · click for rater detail
Determine ways to strengthen or combine materials or develop new materials with new or specific properties for use in a variety of products and applications.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI tools in materials science remains in the pilot and research phase, concentrated in large tech and advanced manufacturing firms. Most materials development workflows are still human-driven and laboratory-based, with AI integration slow outside elite research institutions. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science R&D is a specialized, historically slower-adopting sector for AI compared to information/finance industries, though AI-driven discovery pilots are growing in national labs and large firms. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems provide significant augmentation through property prediction, literature synthesis, and high-throughput computational screening, accelerating hypothesis generation and narrowing the experimental search space. Materials scientists using these tools can explore larger chemical spaces and iterate faster, raising productivity substantially while remaining central to decision-making and experimental design. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids literature review, property prediction, simulation, and candidate screening, substantially boosting researcher productivity even though final determination and validation remain human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in predicting material properties and suggesting compositions via machine learning models trained on materials databases, the task requires hands-on experimentation, iterative testing, and creative synthesis of novel materials—processes that cannot be fully automated end-to-end with current systems. AI can accelerate hypothesis generation but not replace the experimental validation and intuitive materials engineering judgment that dominates the work. |
| Task automatability | claude-sonnet-5 | 2/5 | AI (materials informatics, generative models, molecular simulation) can accelerate hypothesis generation and screening, but the core task requires physical experimentation, validation, and novel scientific judgment that current systems cannot fully replace. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Materials development is typically embedded in R&D departments with significant organizational and regulatory friction, intellectual property considerations, and safety testing requirements. The work often requires specialized equipment access, professional judgment on safety and feasibility, and sign-off by experienced researchers, creating substantial adoption barriers. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement per se, but significant organizational and scientific validation barriers exist since materials must be physically tested and verified by qualified scientists before use in products. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted discovery and screening can reduce computational costs, but materials scientists command high salaries and the full workflow—from ideation through synthesis, characterization, and validation—still requires human expertise and experimental infrastructure. Total cost savings remain modest relative to the loaded human wage. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can reduce computational screening costs, but overall task still requires expensive lab equipment, synthesis, and human expert oversight, keeping total cost comparable to or only modestly cheaper than human-led R&D. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI tools exist for materials property prediction and literature mining (e.g., machine learning potentials, high-throughput screening frameworks), but no deployed system reliably performs the full task of developing novel materials with specified properties in production settings. These remain primarily research-stage capabilities with high error rates when applied to genuinely new material classes. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some deployed tools (e.g., materials discovery platforms like Citrine, AI-driven DFT screening) assist researchers, but they are narrow-scope aids rather than end-to-end reliable production systems for discovering/combining materials. |
Plan laboratory experiments to confirm feasibility of processes and techniques used in the production of materials with special characteristics.
28CI 25–30 · exposure 25 · augmentation 63 · importance 4.1/5 · click for rater detail
Plan laboratory experiments to confirm feasibility of processes and techniques used in the production of materials with special characteristics.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials science remains a relatively traditional field with slow digital adoption for core research activities. While some labs use AI-assisted tools for data analysis and literature mining, the adoption of autonomous or semi-autonomous experimental planning systems is nascent and concentrated in research institutions rather than production environments. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | R&D-heavy materials science sectors adopt AI tools cautiously, mostly for data analysis and screening rather than full experiment planning, reflecting slower adoption than software/finance sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | Current AI tools can meaningfully augment experimental design through literature synthesis, parameter space exploration, and predictive modeling of material properties, helping scientists narrow hypotheses and reduce iterative cycles. However, augmentation remains partial—human creativity and domain judgment are still essential for novel process feasibility assessment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI models and materials informatics/ML tools can suggest candidate materials, predict outcomes, and help design experiment matrices, meaningfully boosting scientist productivity while humans retain final judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in literature review and initial experimental design, the task requires deep domain knowledge, creative hypothesis formation, and iterative refinement based on material properties and constraints that demand human expertise. AI systems today cannot reliably plan novel multi-stage experiments end-to-end with sufficient quality and novelty to meet the 50% time-saving bar. |
| Task automatability | claude-sonnet-5 | 2/5 | Planning novel experiments requires deep domain judgment, hypothesis generation grounded in physical intuition, and knowledge of lab constraints that current AI cannot reliably replicate end-to-end., |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Materials science research requires subject-matter expert judgment and often institutional/regulatory oversight of laboratory protocols. Organizations typically require a licensed scientist to design and sign off on experiments, particularly in regulated or safety-critical contexts, creating strong adoption friction. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically, but organizational and safety review processes, plus reliance on tacit lab expertise and equipment access, create moderate friction against automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI services (including specialized chemistry/materials models) require significant human oversight, validation, and integration costs. The loaded cost of a materials scientist's planning time remains cheaper than a full end-to-end AI system with necessary human review cycles for novel experimental work. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Significant human oversight, domain expertise, and validation are still required, so AI assistance reduces some cost but does not approach order-of-magnitude savings versus a materials scientist's time. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No mature deployed products perform full experimental planning for materials science reliably in production. Relevant tools (AI-assisted literature search, some design automation) exist but are narrow in scope, and critical decisions around feasibility, material selection, and process parameters still require human materials scientists to validate and iterate. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools exist for literature review and suggesting experimental designs (e.g., materials informatics platforms), but no deployed product autonomously plans full feasibility studies for novel material processes in production settings. |
Recommend materials for reliable performance in various environments.
28CI 25–30 · exposure 25 · augmentation 63 · importance 4.0/5 · click for rater detail
Recommend materials for reliable performance in various environments.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While materials informatics has grown in research and large R&D-heavy firms, adoption of AI for autonomous recommendation remains limited; most applications are assistive tools in specialized sectors rather than mainstream production displacement of materials scientists' decision-making. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science and engineering sectors are adopting AI tools for screening and simulation but production-scale autonomous decision-making is still rare and pilot-stage. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-powered materials databases, property prediction models, and literature search tools meaningfully assist materials scientists in narrowing candidate pools and exploring parameter spaces, though the core judgment and environmental-context matching still relies heavily on human expertise. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI/ML tools significantly speed up literature review, property prediction, and candidate screening, meaningfully boosting scientist productivity while humans retain final judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist in materials selection by scanning property databases and matching specifications to known environments, but reliable performance recommendations require deep domain expertise, understanding of failure modes, cost-benefit trade-offs, and context-specific constraints that current systems struggle with end-to-end. Humans currently remain essential for final judgment and validation. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires deep synthesis of material properties, environmental stressors, failure modes, and testing data combined with expert judgment, which current AI cannot fully replicate end-to-end at reliable quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Materials selection carries significant liability and performance consequences; industry standards, regulatory compliance (especially in aerospace, medical, energy sectors), and organizational risk management typically require a licensed or credentialed materials scientist to review and sign off on critical recommendations. |
| Adoption barriers | claude-sonnet-5 | 3/5 | While not licensed like medicine, recommendations affecting safety-critical engineering carry liability exposure, so organizations require qualified engineers to sign off on material choices. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Running materials simulation and database queries is relatively cheap, but the bulk of the cost lies in expert human review, validation testing, and liability assumption—making the total all-in cost comparable to or exceeding a materials scientist's time for non-routine recommendations. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply screen candidate materials computationally, but the human oversight, testing, and validation needed to trust a recommendation keeps overall cost comparable to or only modestly below expert labor. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Materials informatics tools exist (databases, ML models for property prediction), but no deployed product reliably makes independent material recommendations across diverse environments without expert review. Most systems function as decision-support tools requiring substantial human validation rather than autonomous recommenders. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some AI materials-informatics tools exist (e.g., for materials discovery or property prediction) but no deployed product reliably makes final material recommendations for real-world environmental performance without expert validation. |
Perform experiments and computer modeling to study the nature, structure, and physical and chemical properties of metals and their alloys, and their responses to applied forces.
28CI 25–30 · exposure 25 · augmentation 75 · importance 3.8/5 · click for rater detail
Perform experiments and computer modeling to study the nature, structure, and physical and chemical properties of metals and their alloys, and their responses to applied forces.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials science labs are typically university- or industry-R&D-based with slow digitization. While computational tools have been adopted, autonomous or AI-driven experimental systems remain rare in production. Adoption is pilot-stage, not deep deployment. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science R&D is adopting AI/ML tools steadily but remains a specialized, moderately-digitized field with slower uptake than software or finance sectors; most work remains pilot/lab-scale. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at augmenting materials scientists: predictive modeling narrows the experimental search space, data analysis accelerates pattern recognition, and simulation tools help design better experiments. A scientist using modern AI tools is substantially more productive than one without, keeping the human firmly in the loop. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids materials scientists via property prediction, simulation acceleration, literature synthesis, and experimental design optimization, meaningfully boosting productivity while humans retain control of physical experiments. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in designing experiments and running computational simulations of material properties, the task requires hands-on laboratory work (sample preparation, physical testing, equipment operation) that cannot be fully automated end-to-end. AI modeling tools support the computational fraction but cannot achieve 50% time savings on the integrated task today. |
| Task automatability | claude-sonnet-5 | 2/5 | Physical experimentation requires lab equipment, hands-on manipulation, and judgment that AI cannot perform end-to-end; only computational modeling and data analysis portions are automatable today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: novel materials research often requires institutional funding, equipment access, and publication/peer review oversight; liability for material failure is significant; and institutional and regulatory frameworks assume a licensed scientist is accountable. Autonomous experimentation without human validation is not yet accepted practice. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement blocks AI use, but experimental safety protocols, equipment access, and validation requirements for materials used in critical applications create meaningful organizational and procedural friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI computational tools reduce costs for the modeling portion, but the materials scientist's wage and the remaining experimental apparatus, sample preparation, and validation still dominate total cost. Integration overhead and the need for specialist oversight further offset AI savings. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Computational modeling with AI can be cheaper than traditional simulation, but the overall task still requires expensive physical experimentation, specialized equipment, and expert oversight, keeping costs comparable to human-driven work. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed systems exist for narrow aspects—computational materials modeling software and data analysis tools—but no integrated product reliably performs the full experimental-plus-modeling cycle independently. Physical experiments still require human handling, calibration, and decision-making in real labs. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI/ML tools for materials property prediction and simulation exist (e.g., DFT surrogate models, generative materials design), but physical experiment execution and interpretation still require human scientists in production labs. |
Devise testing methods to evaluate the effects of various conditions on particular materials.
28CI 25–30 · exposure 25 · augmentation 63 · importance 3.8/5 · click for rater detail
Devise testing methods to evaluate the effects of various conditions on particular materials.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials science remains a relatively conservative field with strong emphasis on expert judgment and regulatory compliance; while AI-assisted literature search is growing, adoption of AI-designed testing methods in production environments is minimal and slow. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | R&D and materials science sectors adopt AI more slowly for core experimental design work, though computational tools and AI-assisted simulation are gaining some traction in materials informatics. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by surveying published test methods, summarizing relevant standards, and suggesting candidate approaches, thereby reducing literature review burden and accelerating the ideation phase while the scientist retains design responsibility. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by suggesting relevant prior methods, simulating conditions, generating hypotheses, and helping design experimental parameters, boosting researcher productivity while humans retain judgment and validation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in literature review and suggest standard testing protocols, devising novel testing methods requires understanding material properties, experimental design, and creative problem-solving that current AI systems cannot reliably do end-to-end at equal quality with 50%+ time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | Designing novel testing methodologies requires deep domain expertise, creativity, and understanding of physical mechanisms that current AI cannot reliably originate end-to-end; AI can assist with literature review and protocol drafting but not devise validated experimental methods autonomously. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Professional licensing and regulatory standards (ISO, ASTM) often require qualified materials scientists to design and certify test methods; liability for faulty test designs that waste materials or produce incorrect data creates high error-cost asymmetry and organizational barriers to full substitution. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement for this specific task, but organizational and scientific rigor standards, safety protocols, and peer validation create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference costs for supporting literature analysis and method suggestion are modest, but the task still requires expert oversight, validation, and often physical lab design work, making total deployment cost comparable to or higher than hiring a materials scientist. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply generate draft ideas or summarize related methods, but a scientist must still validate, design experiments, and ensure safety/accuracy, so overall cost savings are modest relative to expert labor. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products reliably devise testing methods for novel materials scenarios; AI tools can retrieve existing methods and explain test procedures, but generating new, scientifically sound testing approaches remains in the research/prototype phase. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed product independently designs materials testing protocols in production; AI tools are used as research aids (literature synthesis, simulation suggestions) but the core devising of test methods remains human-led. |
Test individual parts and products to ensure that manufacturer and governmental quality and safety standards are met.
25CI 25–25 · exposure 25 · augmentation 50 · importance 3.8/5 · click for rater detail
Test individual parts and products to ensure that manufacturer and governmental quality and safety standards are met.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials testing is concentrated in manufacturing and specialized labs with slower digital transformation and stronger regulatory conservatism; while data analytics is adopting AI, the physical testing and certification steps remain largely human-controlled and change slowly. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Manufacturing and materials testing sectors adopt AI slowly due to physical process requirements, safety regulation, and capital-intensive equipment cycles. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by automating data logging, flagging outliers, suggesting which standards to check, and accelerating report generation, thereby improving a materials scientist's efficiency without replacing their judgment and hands-on testing work. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help analyze test data, predict failure modes, and automate report generation, meaningfully assisting scientists though not replacing physical testing steps. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with data analysis and documentation of test results, the core task requires physical handling of materials, precise measurements using specialized equipment, and judgment calls about safety standards that typically demand human oversight and hands-on execution today. |
| Task automatability | claude-sonnet-5 | 2/5 | Physical testing of materials and parts requires lab equipment, sample handling, and calibrated instruments that AI cannot perform end-to-end; AI can assist with data analysis and reporting but not the physical testing itself.dishonesty |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong regulatory and liability barriers exist: manufacturers and governmental quality standards (ISO, ASTM, FDA) typically require certified materials scientists or engineers to validate tests, sign off on results, and take responsibility for safety compliance. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Regulatory and safety compliance often requires certified human inspectors or engineers to sign off on test results, creating strong liability and licensing barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for analysis and documentation are relatively inexpensive, but the specialized equipment, human expertise required for test design, and the need for qualified personnel oversight mean the overall cost remains comparable to or higher than human-only testing in most real-world contexts. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Physical testing equipment, calibration, and compliance documentation still require significant capital and skilled human oversight, so AI-driven approaches don't yet undercut human-run testing labs substantially. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some deployed systems can analyze test data and flag anomalies, but no mature end-to-end product reliably performs the full spectrum of materials testing (mechanical, thermal, chemical, safety validation) without significant human intervention and physical manipulation. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some automated testing rigs and AI-driven data analysis tools exist, but full autonomous testing against regulatory standards without human oversight is not deployed at scale. |
Research methods of processing, forming, and firing materials to develop such products as ceramic dental fillings, unbreakable dinner plates, and telescope lenses.
25CI 20–30 · exposure 20 · augmentation 75 · importance 3.8/5 · click for rater detail
Research methods of processing, forming, and firing materials to develop such products as ceramic dental fillings, unbreakable dinner plates, and telescope lenses.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI in materials research is ongoing but still concentrated in informatics and simulation roles within research institutions and large materials companies; displacement of research scientists is minimal and adoption remains in pilot and exploratory phases rather than systematic production replacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science and advanced manufacturing sectors are adopting AI mainly for simulation and screening, but production-scale autonomous experimentation is still nascent compared to fast-moving digital sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments materials scientists by accelerating computational screening of material compositions, predicting properties before synthesis, automating literature analysis, and suggesting experiment designs—all while the scientist remains central to directing research, interpreting results, and making creative breakthroughs. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools (materials informatics, molecular simulation, ML-guided experiment design) meaningfully accelerate hypothesis generation, literature synthesis, and predictive modeling, substantially boosting researcher productivity while humans still perform and interpret physical experiments. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist with literature review, computational simulation of material properties, and experimental design optimization, but the hands-on synthesis, processing, forming, and firing of materials requires physical experimentation and iterative judgment that current systems cannot fully automate. The creative discovery and troubleshooting of novel processing methods remains heavily dependent on human intuition and direct laboratory work. |
| Task automatability | claude-sonnet-5 | 2/5 | This is experimental, hands-on materials R&D requiring physical processing, equipment operation, and empirical testing that AI cannot perform end-to-end; AI can assist with hypothesis generation and data analysis but not the physical experimentation itself. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: regulatory approval is often required for novel materials in regulated industries (medical, aerospace), liability and safety testing are non-delegable, and the task requires deep domain expertise and professional judgment that organizations typically require human scientists to provide and certify. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement mandates a human specifically, but quality/safety-critical applications (dental fillings) impose testing and validation standards, and organizational reliance on physical lab infrastructure creates friction against pure automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of deploying AI systems for materials research (computational infrastructure, data curation, expert integration) currently approaches or exceeds the cost of employing research scientists, especially when accounting for the need for human oversight and validation of experimental outputs. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Physical lab work, equipment, and iterative testing dominate the cost structure and cannot be replaced by cheaper AI inference; AI may reduce some literature review or simulation costs but the core task remains labor- and equipment-intensive. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI tools exist for materials informatics and computational modeling (e.g., machine learning models trained on materials databases), no deployed product reliably performs end-to-end research of new processing and forming methods in production. Lab automation and robotics assist specific steps, but integrating them into a coherent research pipeline that discovers novel methods remains largely in the research and pilot phase. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously researches and develops new material processing methods; this remains a research-stage capability limited to computational materials discovery tools that still require human experimental validation. |
Test metals to determine conformance to specifications of mechanical strength, strength-weight ratio, ductility, magnetic and electrical properties, and resistance to abrasion, corrosion, heat, and cold.
23CI 20–25 · exposure 20 · augmentation 50 · importance 4.3/5 · click for rater detail
Test metals to determine conformance to specifications of mechanical strength, strength-weight ratio, ductility, magnetic and electrical properties, and resistance to abrasion, corrosion, heat, and cold.
23| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials testing labs operate in conservative, highly regulated sectors where established procedures and human expertise are deeply embedded; while data logging is automated in modern labs, replacement of the materials scientist's judgment remains slow due to liability concerns and verification requirements. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science and manufacturing sectors adopt AI more slowly than information/finance sectors, with physical testing labs particularly resistant to full automation due to equipment and certification requirements. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by automating data reduction, flagging outliers, and suggesting conformance status, but the materials scientist remains responsible for interpreting ambiguous results, recommending retesting, and certifying the final decision—a meaningful but bounded augmentation role. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with data analysis, predictive modeling of material properties, anomaly detection in test results, and report generation, meaningfully speeding up the scientist's interpretive work even though physical testing remains manual. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can analyze spectroscopic and imaging data from material tests, the full task requires physical specimen preparation, calibration of specialized equipment, and interpretation of results in context of nuanced specifications that often demand human judgment about acceptable variance and failure modes. |
| Task automatability | claude-sonnet-5 | 2/5 | Physical testing of metal samples requires lab equipment, sample preparation, and hands-on operation of instruments (tensile testers, hardness testers, corrosion chambers) that AI cannot perform; AI can assist with data analysis but not the physical execution. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Material conformance testing often requires ISO/ASTM certification and may involve products for regulated industries (aerospace, medical, automotive); human-led testing and sign-off are frequently mandated by procurement or regulatory requirements, creating substantial legal and organizational barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Materials testing for specification conformance often ties to safety-critical industries (aerospace, construction, automotive) requiring certified testing procedures, documented chain of custody, and qualified personnel sign-off, creating strong regulatory and liability barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Specialized testing equipment is capital-intensive; while automated data analysis is cheap, the infrastructure for physical testing (tensile machines, furnaces, corrosion chambers) remains expensive to operate, and oversight by skilled materials scientists is necessary, keeping total cost per certified test relatively high. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Physical testing still requires human technicians and scientists to operate equipment and interpret results; AI only reduces some data-analysis time, so overall cost savings versus a trained human are limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed systems exist for automated data collection and basic compliance checking in some testing equipment, but no end-to-end product reliably handles the full spectrum of mechanical, thermal, electrical, and corrosion tests with the independent judgment required to certify conformance across diverse material types. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product autonomously performs physical materials testing end-to-end; automated lab equipment exists but requires human setup, calibration, and interpretation, with AI playing only a peripheral analytical role. |
Confer with customers to determine how to tailor materials to their needs.
21CI 11–30 · exposure 13 · augmentation 50 · importance 3.7/5 · click for rater detail
Confer with customers to determine how to tailor materials to their needs.
21| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Materials science remains a specialized, relationship-driven domain with slower digital transformation than information services; customer conferencing is typically high-touch and localized, with limited automation adoption signals in production settings. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science and manufacturing sectors show slower, more cautious AI adoption compared to information/finance sectors, with client-facing consultation being especially resistant. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by rapidly screening material databases, generating property comparisons, and flagging design trade-offs for the scientist to discuss with customers, moderately boosting the efficiency of the conferencing process itself. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help scientists prepare technical summaries, look up material properties, and draft proposals ahead of or after customer meetings, but doesn't materially change the live conferring process. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in analyzing material properties and generating tailoring suggestions, the task requires nuanced understanding of unspoken customer needs, iterative negotiation, and relationship judgment. Current AI systems struggle with the open-ended discovery and real-time adaptation central to customer conferencing, falling well short of 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires live, relationship-based dialogue to elicit unstated needs, negotiate trade-offs, and build trust—AI cannot conduct this end-to-end today with equal quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Customers typically expect and prefer direct engagement with credentialed materials scientists; liability for incorrect material recommendations falls on the organization; and regulatory contexts (aerospace, medical devices) often require documented expert review, creating organizational and legal friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically bars AI, but customer preference for human expert interaction, liability for wrong specifications, and organizational trust norms create real friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI infrastructure, integration, and mandatory human oversight for customer-facing technical consultation remains comparable to or higher than deploying a materials scientist directly, especially given the need to validate AI recommendations and manage customer relationships. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Human consultation costs are significant, but AI cannot yet substitute for the full task, so any AI cost is additive rather than replacing the human expert's time. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs customer conferencing and needs-discovery at scale in materials science. Chatbots exist but lack domain expertise and the ability to probe and clarify requirements the way human scientists do, making them unsuitable for production use without heavy human oversight. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously confers with customers to define custom materials specifications; this remains a human relationship-driven consultative process. |
Teach in colleges and universities.
15CI 5–25 · exposure 17 · augmentation 63 · importance 3.6/5 · click for rater detail
Teach in colleges and universities.
15| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Universities are among the slowest-adopting sectors for labor displacement, with entrenched tenure systems, accreditation requirements tied to faculty qualifications, and cultural resistance to replacing teaching faculty. Adoption remains in the pilot/supplement phase, not production replacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Higher education adopts AI tools for support functions (grading, tutoring, content creation) but remains slow to replace core teaching roles due to accreditation and cultural resistance. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist with grading, generating lecture outlines, answering routine student questions, and providing supplemental explanations, raising faculty productivity in administrative and preparatory tasks. However, the human instructor remains essential for live teaching, feedback, mentoring, and assessment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI meaningfully augments teaching by helping draft lectures, generate examples, create assessments, and provide student tutoring support, while the human instructor retains primary responsibility. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Teaching in colleges and universities requires real-time interaction, dynamic responsiveness to student questions, interpersonal rapport-building, and contextual judgment that current AI systems cannot replicate end-to-end. While AI can assist with content creation or grading, it cannot replace the core pedagogical and mentoring functions that define university teaching. |
| Task automatability | claude-sonnet-5 | 2/5 | Teaching involves live interaction, mentoring, grading judgment, and adaptive instruction that current AI cannot fully replicate end-to-end despite being able to help with materials prep and grading assistance.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Strong legal and regulatory barriers exist: faculty must hold appropriate degrees and credentials, universities require human faculty for accreditation, and institutional governance (faculty contracts, intellectual property, student outcomes accountability) mandates human instruction. Stakeholders (students, parents, employers) expect and often legally require human educators. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Accreditation standards, degree-granting authority, and institutional/labor requirements typically mandate a qualified human instructor of record, creating strong regulatory and organizational barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The all-in cost of setting up, monitoring, and maintaining AI systems to approximate university instruction, plus liability and quality assurance overhead, exceeds the cost of a tenured or adjunct faculty member, especially at scale across multiple courses and institutions. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can cheaply generate content, replacing an accredited professor requires human oversight, credentialing, and institutional accountability, keeping all-in costs comparable to or only modestly below human costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs university-level teaching as a complete substitution; AI tutoring systems and content generators exist but are narrow in scope and require heavy human oversight. Current systems fail at the breadth, depth, and judgment required for accredited course delivery. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools (tutoring bots, lecture-generation, grading assistants) exist but no deployed product independently teaches a college course with instructor-level reliability and accreditation. |
Visit suppliers of materials or users of products to gather specific information.
13CI 5–21 · exposure 8 · augmentation 50 · importance 3.5/5 · click for rater detail
Visit suppliers of materials or users of products to gather specific information.
13| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Materials science operates in heavily regulated, relationship-driven sectors (manufacturing, aerospace, pharmaceuticals) where supply chain decisions require certified human expertise and legal accountability; adoption of autonomous supplier-visit automation is minimal. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Materials science and manufacturing sectors show moderate AI adoption for data analysis, but physical site visits remain untouched by current automation trends. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by pre-generating interview templates, organizing supplier data, translating technical specifications, and summarizing findings, which raises a scientist's preparation and documentation efficiency without replacing the human site visit and relationship-building. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help prepare visit agendas, summarize prior supplier data, analyze notes afterward, or draft follow-up reports, meaningfully assisting the surrounding workflow even though the visit itself is unaided. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can help draft supplier questions and organize information, the task fundamentally requires building relationships, understanding nuanced context-specific needs, and making real-time judgments during face-to-face or direct interactions. Current systems cannot conduct the full supplier visit, negotiate, or gather the tacit knowledge that defines this task. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires physical presence, in-person relationship building, and on-site observation that current AI cannot perform end-to-end.https requires travel and physical inspection which is not automatable by software. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Supplier and customer relationships require trust and legal accountability; the human scientist must sign off on material specifications and qualifications. Organizations will not substitute human judgment on critical supplier partnerships without regulatory and contractual liability concerns. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing barrier exists, but the inherent physical/relational nature of site visits and supplier trust creates strong practical friction against remote or automated substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The human cost of a materials scientist visiting suppliers and building relationships is already optimized for expertise; AI systems would require expensive human oversight and validation of information gathered, making the total cost exceed that of direct human engagement. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the physical visit itself, so there is no comparable AI cost basis; any AI cost is additive to human travel costs, not a replacement. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product reliably conducts supplier visits or user interviews independently. Chatbots can simulate Q&A but cannot navigate physical sites, read interpersonal cues, or make on-the-spot technical decisions needed for material evaluation. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product physically visits suppliers or customers to gather information; this remains a fully human, in-person activity. |
Related occupations — Life, Physical & Social Science
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.