Nuclear Engineers

17-2161.00

Median wage $133,970/yr15,280 employed (US)Rank #789 of 923 scored · top 85% by substitution

Conduct research on nuclear engineering projects or apply principles and theory of nuclear science to problems concerned with release, control, and use of nuclear energy and nuclear waste disposal.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution15

Exposure17

Augmentation56

Substitution — the headline: capability discounted by cost, barriers and adoption.

Exposure — technical capability alone, regardless of whether anyone deploys it.

Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.

Tasks on the substitution scale

20 rated tasks, binned by substitution score.

Position among all scored occupations

Distribution of 923 occupation scores; the marker is this occupation.

Tasks with substitution ≥ 70

Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.

Why this score

The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.

Task automatabilityw 35%19

panel mean rating 1.8/5 → substitution pressure 19/100

Technical feasibility todayw 20%14

panel mean rating 1.6/5 → substitution pressure 14/100

Cost vs. human wagew 15%19

panel mean rating 1.7/5 → substitution pressure 19/100

Adoption barriersw 20%inverted — strong barriers lower the score9

panel mean rating 4.6/5 (barrier strength) → substitution pressure 9/100

Sector adoption velocityw 10%13

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

Task breakdown (20 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.

Keep abreast of developments and changes in the nuclear field by reading technical journals or by independent study and research.

CI 52–76 · exposure 62 · augmentation 100 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Professional services and engineering sectors show moderate AI adoption in knowledge management and research monitoring; however, conservative institutional practices in nuclear engineering and preference for human expertise review slow deep penetration compared to faster-adopting sectors.
Sector adoption velocity	claude-sonnet-5	2/5	Nuclear engineering is a specialized, safety-conscious, and historically slow-to-digitize sector, with AI research tool adoption still nascent compared to fast-moving information/finance sectors.
Augmentation potential	claude-haiku-4-5-20251001	5/5	AI significantly amplifies a nuclear engineer's ability to stay current by filtering, summarizing, and cross-referencing vast journal volumes, allowing the human to focus on critical interpretation and application rather than raw information consumption.
Augmentation potential	claude-sonnet-5	5/5	AI-powered literature search, summarization, and alerting tools substantially speed up staying current with technical developments, making this a strong augmentation use case even without full automation.
Task automatability	claude-haiku-4-5-20251001	4/5	AI systems can automatically monitor, summarize, and synthesize technical journals, research papers, and field developments with minimal human intervention, achieving substantial time savings. However, the task inherently requires human judgment to interpret significance and integrate findings into professional practice, preventing a full end-to-end automation rating.
Task automatability	claude-sonnet-5	3/5	AI can summarize, search, and synthesize technical literature at scale, but genuine expert-level assessment of novel nuclear developments still requires human judgment, so only partial time savings are realized.
Adoption barriers	claude-haiku-4-5-20251001	2/5	While professional standards expect engineers to maintain current knowledge, there are minimal legal barriers, licensing requirements, or mandatory human oversight specific to the act of reading journals. Organizational culture may prefer human judgment but does not legally mandate it.
Adoption barriers	claude-sonnet-5	2/5	No licensing requirement mandates a human perform literature review, though professional norms and accuracy expectations in a safety-critical field create some friction against pure automation.
Cost vs. human wage	claude-haiku-4-5-20251001	5/5	AI-powered journal monitoring and summarization services cost a fraction of the loaded hourly wage of a nuclear engineer, offering orders-of-magnitude cost advantage compared to manual reading and synthesis time.
Cost vs. human wage	claude-sonnet-5	3/5	AI tools for literature review and summarization are inexpensive relative to engineer time, but the need for expert verification of technical accuracy narrows the cost advantage to roughly comparable when accounting for oversight.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	Mature products exist today (academic paper aggregators, AI-powered research monitoring tools, journal alert systems with summarization) that reliably perform large portions of this task in production. Limitations remain in real-time filtering for domain-specific relevance and context-dependent prioritization.
Technical feasibility today	claude-sonnet-5	3/5	Deployed AI literature-summarization and research-assistant tools exist and are used in technical fields, but domain-specific nuclear engineering nuance and accuracy checking still require expert review, limiting reliability.

Recommend preventive measures to be taken in the handling of nuclear technology, based on data obtained from operations monitoring or from evaluation of test results.

CI 15–78 · exposure 53 · augmentation 88 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	4/5	Nuclear facilities have substantial incentives to deploy monitoring and predictive maintenance systems, and such tools are increasingly integrated into operations. However, regulatory conservatism and sector-wide cautious modernization mean adoption is deliberate rather than rapid.
Sector adoption velocity	claude-sonnet-5	2/5	The nuclear energy sector is historically slow to adopt new technologies due to regulatory conservatism and safety culture, with AI tools seeing only limited pilot use for data analysis rather than decision-making.
Augmentation potential	claude-haiku-4-5-20251001	5/5	AI dramatically assists nuclear engineers by synthesizing vast amounts of real-time operational data, identifying subtle anomalies, and surfacing candidate preventive measures—dramatically raising their ability to detect emerging risks while they remain the decision-maker and validator.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist by rapidly analyzing monitoring data, flagging anomalies, and drafting preliminary reports, substantially speeding up the engineer's evaluation process while the engineer retains final judgment.
Task automatability	claude-haiku-4-5-20251001	5/5	AI can analyze operational data and test results against established safety protocols and historical incident patterns to recommend preventive measures with high speed and consistency. This is predominantly a data-to-recommendation task that meets or exceeds 50% time savings at equal quality compared to manual review.
Task automatability	claude-sonnet-5	2/5	This requires synthesizing engineering judgment, safety-critical risk assessment, and regulatory context from complex monitoring data; AI can assist analysis but cannot autonomously generate reliable, accountable safety recommendations end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory frameworks (NRC, IAEA) typically require that a licensed nuclear engineer review, validate, and sign off on safety recommendations; the final authority and liability rest with the licensed professional, creating a hard requirement for human oversight and sign-off.
Adoption barriers	claude-sonnet-5	5/5	Nuclear safety recommendations are subject to strict regulatory oversight (e.g., NRC) requiring credentialed professional engineers to sign off, making this a hard-barrier task with severe liability exposure.
Cost vs. human wage	claude-haiku-4-5-20251001	5/5	AI inference for data analysis is extremely cheap relative to the loaded cost of a nuclear engineer's time; continuous monitoring with automated alerts cost orders of magnitude less than employing humans for the same task.
Cost vs. human wage	claude-sonnet-5	2/5	While data analysis tools are cheap, the liability and expertise required mean a human engineer must still validate any AI output, keeping overall cost close to or above human-only baselines when oversight is factored in.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	Deployed systems (anomaly detection, predictive maintenance, and rule-based expert systems) exist in nuclear operations and can reliably identify patterns in monitoring data and flag safety recommendations. While some nuclear plants use such systems in production, full end-to-end recommendation generation across heterogeneous data sources and regulatory contexts is still maturing.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product independently issues nuclear safety recommendations in production; this remains firmly in the domain of licensed engineers with AI only as a research-stage analytic aid.

Prepare environmental impact statements, reports, or presentations for regulatory or other agencies.

CI 25–37 · exposure 33 · augmentation 63 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Nuclear engineering is a highly regulated, low-volume sector with long approval cycles and conservative adoption practices. While some firms pilot AI writing tools, production deployment of AI for regulatory documents remains limited and cautious.
Sector adoption velocity	claude-sonnet-5	2/5	The nuclear engineering sector is conservative and highly regulated with slow technology adoption cycles, so AI tools are used cautiously and mainly for drafting support rather than deployed at scale.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist by auto-generating first drafts, organizing data tables, suggesting literature references, and formatting reports—materially speeding the preparation phase. However, the engineer must validate technical accuracy and regulatory compliance, so augmentation is meaningful but partial.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully speed up literature review, data synthesis, and drafting of report sections, significantly boosting engineer productivity while they retain responsibility for technical accuracy and final approval.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can draft sections of environmental reports (literature review, data summaries), nuclear regulation requires precise technical language, site-specific analysis, and legally binding assertions that demand expert human judgment and accountability. Current systems cannot reliably produce end-to-end compliant impact statements without substantial expert revision.
Task automatability	claude-sonnet-5	3/5	AI can draft substantial portions of environmental impact statements by synthesizing data and templates, but accurate technical content requires specialized nuclear domain data, site-specific analysis, and expert verification that current systems cannot fully automate end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory agencies (NRC, EPA) require licensed nuclear engineers or authorized environmental professionals to sign and certify environmental statements; liability for false or incomplete disclosures is severe and personal. These legal and accountability barriers prevent full substitution.
Adoption barriers	claude-sonnet-5	4/5	Regulatory submissions to agencies like the NRC require sign-off by qualified, often licensed professionals, and errors carry significant legal and safety liability, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI tools for drafting assistance are inexpensive, but nuclear engineers' highly specialized labor (typically >$100k/year loaded cost) remains essential for compliance review, technical validation, and liability sign-off, so the all-in cost of AI+engineer remains comparable to engineer-alone.
Cost vs. human wage	claude-sonnet-5	3/5	AI can reduce drafting time and cost substantially, but the need for licensed engineers and specialists to verify technical accuracy and regulatory compliance keeps overall costs closer to human-comparable levels.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI writing assistants exist and can help structure documents, but no deployed product reliably generates a complete, regulatory-grade environmental impact statement for nuclear facilities without human engineer oversight and rewriting. Products demonstrate capability only on narrower subtasks like data visualization or summary generation.
Technical feasibility today	claude-sonnet-5	2/5	Generic AI writing/drafting tools are used for report generation in many industries, but no deployed product reliably produces regulatory-grade nuclear environmental impact statements without extensive expert revision and validation.

Prepare technical reports of findings or recommendations, based on synthesized analyses of test results.

CI 23–34 · exposure 33 · augmentation 63 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear engineering operates in highly regulated, risk-averse sectors with strict compliance requirements; adoption of unsupervised AI for technical reporting is minimal, and organizational and regulatory culture strongly favors human expert authorship and accountability.
Sector adoption velocity	claude-sonnet-5	2/5	The nuclear energy sector is conservative, highly regulated, and slow to adopt new digital tools compared to information or finance sectors, with AI use largely limited to pilot projects.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist by organizing data, drafting initial sections, and suggesting phrasing, but the nuclear engineer must retain control over technical interpretation, conclusions, and recommendations, making it a supportive rather than transformative tool.
Augmentation potential	claude-sonnet-5	4/5	AI can significantly speed up drafting, summarizing data trends, and organizing findings into report structures, meaningfully boosting engineer productivity while they retain responsibility for accuracy and compliance.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist in synthesizing test data and drafting report sections, the task requires expert judgment to interpret complex nuclear engineering findings and formulate defensible safety-critical recommendations. Current systems cannot reliably perform the full end-to-end task with sufficient accuracy for regulatory contexts.
Task automatability	claude-sonnet-5	3/5	AI can draft report sections and synthesize structured test data into narrative form, but nuclear engineering reports require domain-specific validation, regulatory precision, and interpretive judgment that current AI cannot fully replicate end-to-end.on average.Half of drafting/synthesis is automatable but expert review remains essential.this.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Nuclear engineering reports often must be signed by licensed Professional Engineers (PEs) and comply with regulatory standards (NRC, ASME, etc.); the engineer cannot delegate the professional certification responsibility, creating a hard legal barrier to full automation.
Adoption barriers	claude-sonnet-5	4/5	Nuclear engineering reports feed into heavily regulated safety and licensing processes (e.g., NRC oversight), requiring credentialed engineers to review, sign off, and take liability for findings, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI infrastructure and oversight costs are comparable to or exceed the cost of a nuclear engineer's time spent on initial drafting, particularly when accounting for mandatory expert review, revision cycles, and liability concerns in safety-critical domains.
Cost vs. human wage	claude-sonnet-5	2/5	While drafting assistance is cheap, the highly specialized, safety-critical nature of nuclear reports requires extensive expert review, keeping overall costs closer to human-level rather than an order of magnitude cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably performs independent technical report generation for nuclear engineering at the quality required for regulatory submission or decision-making. AI writing tools exist but lack domain expertise and produce work requiring substantial expert review and modification.
Technical feasibility today	claude-sonnet-5	2/5	General-purpose LLMs and report-drafting tools exist and are used for technical writing assistance, but no deployed product reliably handles nuclear-specific test synthesis and regulatory-grade reporting without heavy human oversight.

Conduct environmental studies on topics such as nuclear power generation, nuclear waste disposal, or nuclear weapon deployment.

CI 20–25 · exposure 20 · augmentation 63 · importance 3.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Nuclear engineering is a specialized, highly regulated sector with slow digitization and conservative adoption patterns. Few nuclear facilities are deploying autonomous AI agents for environmental work; pilots are nascent and deployment remains at the margin due to regulatory constraints and organizational risk aversion.
Sector adoption velocity	claude-sonnet-5	2/5	Nuclear engineering is a highly regulated, low-digitization niche sector with limited AI tool adoption compared to fast-moving sectors like finance or general professional services.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist nuclear engineers by automating literature searches, summarizing regulatory documents, and preprocessing environmental data for analysis. These augmentation opportunities are useful but partial—the core scientific judgment, stakeholder engagement, and regulatory interpretation remain human-driven.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist with data analysis, literature synthesis, report drafting, and simulation modeling, significantly boosting productivity while engineers retain responsibility for judgment and sign-off.
Task automatability	claude-haiku-4-5-20251001	2/5	Environmental studies require extensive domain expertise, original data collection, regulatory interpretation, and nuanced judgment about site-specific impacts. While AI can assist with literature review, data analysis, and report drafting, the full end-to-end task of conducting original environmental studies demands human expertise and cannot achieve 50% time savings at equal quality today.
Task automatability	claude-sonnet-5	2/5	AI can assist with literature review, data synthesis, and drafting portions of environmental studies, but the core work involves original research design, site-specific data collection, regulatory judgment, and expert analysis that current AI cannot autonomously perform end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Environmental impact assessments for nuclear projects are heavily regulated by federal and state agencies (NRC, EPA, NEPA) and typically require licensed professionals to take responsibility for findings. Liability and sign-off requirements create hard barriers; regulatory frameworks mandate human expertise and accountability, preventing full substitution.
Adoption barriers	claude-sonnet-5	4/5	Nuclear-related environmental studies typically require licensed professional engineers, regulatory agency review (e.g., NRC), and legal accountability for safety and compliance, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI tools (LLMs, data analytics) cost less per unit than a senior environmental engineer, but integration overhead, multiple human reviews, and the need for expert validation to ensure regulatory and technical adequacy means the total cost of AI-assisted workflow remains comparable to or higher than direct human performance.
Cost vs. human wage	claude-sonnet-5	2/5	While AI can cut costs on literature review and report drafting, the specialized modeling, regulatory compliance, and expert validation required keep human engineers central, limiting overall cost savings.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably conducts original environmental studies autonomously. AI tools exist for text generation and data analysis, but the specialized knowledge required for nuclear engineering studies, regulatory compliance (NRC, EPA requirements), and site assessment remains beyond current system capability; deployed solutions are narrow and require heavy human oversight.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product conducts full environmental studies on nuclear topics; this remains a specialized, research-stage application requiring domain expertise and physical/field data collection.

Monitor nuclear facility operations to identify any design, construction, or operation practices that violate safety regulations and laws or could jeopardize safe operations.

CI 18–20 · exposure 25 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	The nuclear industry is highly conservative, heavily regulated, and adoption of AI in safety-critical roles has been minimal. Pilots exist but production deployment of AI-driven safety monitoring remains rare, and organizational and regulatory friction strongly constrains velocity.
Sector adoption velocity	claude-sonnet-5	2/5	The nuclear sector is conservative and heavily regulated, with slow uptake of AI tools compared to information or finance industries.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can usefully augment engineers by flagging anomalies in sensor data, highlighting regulatory mismatches, or summarizing trends in operational logs. However, the final judgment of safety violations and jeopardy requires expert human interpretation, limiting the augmentation impact to decision support rather than transformation of productivity.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist by parsing regulations, analyzing sensor data, and flagging potential anomalies for human engineers to investigate further, improving efficiency while human judgment remains central.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with data review and pattern detection in operational logs, this task requires integrated judgment across complex regulatory frameworks, facility-specific context, and real-time anomaly detection. Current systems cannot autonomously identify latent safety violations or jeopardize conditions with the reliability needed to replace human expert monitoring, though they could flag potential issues for review.
Task automatability	claude-sonnet-5	2/5	AI can assist in flagging anomalies or cross-checking documentation against regulations, but real-time safety-critical monitoring requiring judgment, physical inspection, and accountability cannot be fully automated today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear facility oversight is heavily regulated by agencies like the NRC, which mandate that licensed nuclear engineers perform or directly oversee safety-critical monitoring. Liability, regulatory sign-off requirements, and the legal necessity of human professional accountability create hard barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Nuclear safety compliance is heavily regulated (e.g., NRC requirements) and legally mandates qualified licensed engineers to review and certify safety practices, making full substitution essentially prohibited.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI monitoring systems require significant integration, customization per facility, ongoing oversight by qualified engineers, and regulatory validation. The loaded cost of this integrated human-supervised system likely approaches or exceeds the cost of traditional expert monitoring, especially given the need for high-confidence outputs and liability.
Cost vs. human wage	claude-sonnet-5	2/5	Given the extreme liability and required certification, human expert oversight remains necessary, so AI mainly supplements rather than replaces cost, keeping the ratio close to comparable rather than dramatically cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Some regulatory monitoring systems exist that flag deviations from parameters, but no deployed AI product reliably performs comprehensive violation detection across design, construction, and operational practices at scale. Systems in use are mostly narrowly scoped (e.g., sensor threshold alerts) rather than integrated safety oversight, and nuclear facilities remain heavily dependent on expert human review.
Technical feasibility today	claude-sonnet-5	2/5	Some monitoring/anomaly-detection software exists in nuclear plants, but comprehensive AI systems that independently identify safety violations across design, construction, and operations are not deployed at scale.

Examine accidents to obtain data for use in design of preventive measures.

CI 15–20 · exposure 20 · augmentation 63 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Nuclear engineering is a slow-moving, heavily regulated sector with conservative practices. While some organizations may pilot AI-assisted analysis tools, production adoption of AI for accident investigation and design decisions remains limited due to safety culture and regulatory constraints.
Sector adoption velocity	claude-sonnet-5	2/5	The nuclear energy sector is conservative and slow to adopt AI for safety-critical functions, with heavy regulatory oversight limiting rapid deployment.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist nuclear engineers by automating data aggregation from accident reports, identifying statistical patterns, and generating preliminary analyses. However, the specialized judgment required means augmentation is partial—AI enhances data preparation and pattern discovery rather than transforming the core investigative task.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist by analyzing sensor logs, simulating failure scenarios, and organizing incident data, helping engineers identify patterns faster while they retain final judgment.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with data extraction and pattern recognition from accident reports, the task requires domain expertise in nuclear engineering, safety systems, and contextual judgment to identify root causes and inform design implications. Current AI cannot reliably perform the full investigative and synthesis work end-to-end.
Task automatability	claude-sonnet-5	2/5	Investigating nuclear accidents requires site inspection, physical evidence gathering, expert judgment, and synthesis of complex causal chains that AI cannot perform end-to-end; AI can assist with data analysis but not the full investigative task.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear engineering is heavily regulated by agencies like the NRC; design changes following accidents must be signed off by licensed nuclear engineers and undergo formal regulatory review. Legal liability for accident prevention is extremely high, creating hard barriers against autonomous AI-driven decisions.
Adoption barriers	claude-sonnet-5	5/5	Nuclear safety investigations are heavily regulated (e.g., NRC requirements) and require licensed engineers to sign off on findings, with severe liability for errors.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI-assisted tools reduce some data processing costs, but the task requires specialized nuclear engineers whose expertise commands high wages. Integration costs and necessary human oversight mean overall cost savings are modest; AI remains a supporting tool rather than a cost-reducing substitute.
Cost vs. human wage	claude-sonnet-5	2/5	The human expertise, safety-critical judgment, and physical investigation required mean AI can only reduce some analysis time, not replace the overall cost of skilled investigators.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably conducts independent accident investigation and preventive measure design in nuclear contexts. AI tools can support document analysis and data compilation, but human nuclear engineers must lead investigation, interpret findings, and determine design changes due to safety-critical liability.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product autonomously conducts nuclear accident investigations; this remains a highly specialized, research-stage application at best.

Consult with other scientists to determine parameters of experimentation or suitability of analytical models.

CI 9–25 · exposure 20 · augmentation 63 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear engineering remains a highly specialized, highly regulated sector with strong organizational and cultural emphasis on human expert accountability. Adoption of AI for critical parameter and model decisions is minimal, and sectors show laggard adoption patterns for safety-critical automation.
Sector adoption velocity	claude-sonnet-5	2/5	Nuclear engineering and specialized scientific research are conservative, highly regulated fields with slower AI adoption compared to sectors like finance or general software, though computational modeling tools are used.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist by rapidly summarizing relevant literature, suggesting candidate models based on prior experiments, or generating preliminary parameter ranges—useful productivity aids that keep the human engineer in control of the final determination.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist by running simulations, summarizing literature, suggesting parameter ranges, and flagging model limitations, improving efficiency of the consultation process while humans retain final judgment.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can suggest experimental parameters and models based on literature and data, the task fundamentally requires domain expertise, contextual judgment, and collaborative scientific reasoning that current AI systems cannot reliably perform end-to-end. AI may assist in generating options, but the final determination requires human nuclear engineers to evaluate feasibility, safety, and suitability.
Task automatability	claude-sonnet-5	2/5	This requires real-time scientific judgment, negotiation, and domain expertise integration across specialists; AI can support but not replace the collaborative decision-making core to the task.dlg
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear engineering is heavily regulated; experimentation parameters and analytical models must be approved by qualified licensed engineers and nuclear authorities. Liability, safety certification, and legal requirements mandate that authorized human experts sign off on and take responsibility for these determinations.
Adoption barriers	claude-sonnet-5	4/5	Nuclear engineering involves safety-critical, regulated work where model suitability and experimental parameters typically require sign-off by qualified engineers, creating strong professional and regulatory barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	Current AI systems would require significant expert oversight and validation, making the all-in cost (inference, integration, expert review) likely higher than simply having qualified nuclear engineers collaborate directly without AI intermediation.
Cost vs. human wage	claude-sonnet-5	2/5	Human expert consultation is not easily replaced by AI inference costs; the value lies in expert judgment and trust, so AI does not yet offer a clear order-of-magnitude cost advantage for the full task.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably performs this task independently. AI tools can draft literature reviews and suggest parameters, but they lack the specialized nuclear engineering knowledge, safety certification requirements, and accountability needed to meaningfully 'consult' and 'determine' suitability in a production nuclear context.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously consults with scientists to set experimental parameters or validate models; this remains a human collaborative activity, with AI at best a background research aid.

Write operational instructions to be used in nuclear plant operation or nuclear fuel or waste handling and disposal.

CI 13–18 · exposure 20 · augmentation 38 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear power is a heavily regulated, risk-averse sector with low digitization of operational documentation workflows; adoption of AI for safety-critical instruction writing remains negligible despite decades of AI progress, reflecting deep institutional and regulatory resistance.
Sector adoption velocity	claude-sonnet-5	1/5	The nuclear power sector is heavily regulated, safety-conservative, and slow to adopt AI for safety-critical documentation, with adoption lagging far behind information-sector norms.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI can assist with formatting, organizing sections, or suggesting standard language from existing procedure libraries, but the bespoke technical and safety content requires domain expert authorship; augmentation is marginal because the expert must validate and often rewrite most material.
Augmentation potential	claude-sonnet-5	3/5	AI can help engineers draft initial text, check formatting consistency, or summarize regulatory references, providing useful but partial productivity gains while humans retain full responsibility for content accuracy.
Task automatability	claude-haiku-4-5-20251001	2/5	Writing operational instructions for nuclear plants requires deep domain expertise, regulatory knowledge, and safety-critical judgment that current AI cannot fully replicate. While AI can draft template text or organize existing procedures, the nuclear safety and liability stakes demand human expert review and sign-off, preventing >50% unattended time savings.
Task automatability	claude-sonnet-5	2/5	Drafting technical procedures requires deep engineering judgment, regulatory compliance, and site-specific safety analysis that current AI cannot reliably originate end-to-end, though it can assist with drafting boilerplate sections.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Strict NRC and international nuclear regulatory frameworks require licensed nuclear engineers to author and certify operational procedures; legal liability, security classification, and human sign-off mandates create hard barriers to full automation or unsupervised AI deployment.
Adoption barriers	claude-sonnet-5	5/5	Nuclear operational instructions are subject to strict regulatory oversight (e.g., NRC) requiring licensed engineer review and sign-off, with severe liability for errors, making this a hard-barrier task.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI-generated instruction drafts still require expensive nuclear engineer review, validation, and revision; the oversight cost nearly matches the full wage of writing from scratch, and liability risk prevents cost savings from materializing at scale.
Cost vs. human wage	claude-sonnet-5	2/5	While AI drafting assistance is cheap, the extensive verification, engineering review, and regulatory sign-off required keep overall costs comparable to or only modestly below fully human-authored procedures.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI product reliably generates production-ready operational instructions for nuclear facilities; vendors offer drafting aids and templates, but all output requires substantial expert review and modification. The regulatory compliance burden and zero-tolerance error culture mean even pilot deployments remain limited in scope.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product autonomously writes nuclear operational instructions in production; this remains firmly in human expert territory with only generic drafting tools available.

Design or develop nuclear equipment, such as reactor cores, radiation shielding, or associated instrumentation or control mechanisms.

CI 9–18 · exposure 20 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear engineering is a conservative, heavily regulated sector with long design cycles and stringent safety oversight; adoption of unsupervised AI automation is minimal, and organizational risk aversion combined with regulatory constraints means velocity remains very slow.
Sector adoption velocity	claude-sonnet-5	1/5	The nuclear engineering sector is a highly regulated, safety-critical, low-digitization-of-core-workflows field with minimal AI production deployment for actual equipment design.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI tools (generative design exploration, parametric analysis, documentation drafting) can usefully assist nuclear engineers in routine subtasks, reducing iteration time and routine calculations, but the core work of safety validation and regulatory justification remains human-centered.
Augmentation potential	claude-sonnet-5	3/5	AI tools can assist with simulations, literature review, code generation for modeling, and documentation drafting, providing moderate productivity gains while humans retain full design responsibility.
Task automatability	claude-haiku-4-5-20251001	2/5	Nuclear equipment design requires integrating complex physics, safety regulations, and novel configurations—AI can assist with routine calculations and code generation for control logic, but the novel synthesis, safety validation, and regulatory justification demand human engineering judgment that AI cannot autonomously complete end-to-end.
Task automatability	claude-sonnet-5	2/5	AI can assist with simulations, calculations, and drafting design documentation, but core engineering design of reactor systems requires deep domain judgment, safety-critical validation, and physical testing that current AI cannot autonomously perform end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear design is heavily regulated (NRC, IAEA standards) and requires licensed professional engineers (PE) to certify designs; liability for radiation safety failures is severe, and regulatory frameworks explicitly mandate human expert review and sign-off, creating hard legal barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Nuclear equipment design is subject to stringent regulatory oversight (e.g., NRC licensing), requiring certified professional engineers to sign off, making substitution by AI legally and practically barred.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Nuclear equipment design is high-stakes, low-volume work performed by credentialed engineers earning six figures; AI inference is cheap, but integration, validation, and the liability burden of AI-assisted nuclear design add substantial overhead that does not yet approach an order-of-magnitude cost advantage.
Cost vs. human wage	claude-sonnet-5	1/5	Given the necessity of extensive human expert validation, regulatory review, and physical safety analysis, AI does not reduce all-in costs below human engineering teams for this task.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI system autonomously designs reactor cores or radiation shielding at production quality; CAD and FEA tools exist but require expert setup, interpretation, and verification. Current AI excels at narrow subtasks (parameter optimization, documentation) but lacks the cross-domain integration and safety assurance production systems require.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product independently designs nuclear equipment; this remains highly specialized engineering work done by licensed professionals using traditional CAD/simulation tools with AI at most as a minor aid.

Conduct tests of nuclear fuel behavior and cycles or performance of nuclear machinery and equipment to optimize performance of existing plants.

CI 3–20 · exposure 13 · augmentation 50 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Nuclear energy sectors are conservative, heavily regulated, and slow to adopt novel automation due to safety and compliance requirements. Adoption of AI-assisted analysis is emerging but deployment of autonomous testing systems remains nascent across the industry.
Sector adoption velocity	claude-sonnet-5	2/5	Nuclear power is a highly regulated, safety-critical, low-digitization physical sector where AI adoption for hands-on testing remains nascent despite some data-analysis pilots.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully augment nuclear engineers by automating data reduction, anomaly detection in sensor streams, and simulation-based optimization of test parameters. However, the core work of designing experiments, interpreting complex physical phenomena, and making safety-critical decisions remains human-led.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with simulation, data analysis, predictive modeling of fuel behavior, and anomaly detection in test data, improving efficiency of the analytical component of this task.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist in data analysis and simulation of test results, the task inherently requires hands-on physical testing, equipment operation, and real-time monitoring of nuclear systems that cannot be fully automated remotely. Physical presence and direct control of complex, safety-critical nuclear equipment remains essential.
Task automatability	claude-sonnet-5	1/5	This requires physical testing of nuclear fuel and machinery, hands-on instrumentation, and specialized lab/plant access that AI cannot perform end-to-end; only data analysis portions are automatable.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear engineering and testing are heavily regulated by agencies like the NRC; federal licensing requirements mandate that qualified nuclear engineers must design, conduct, and sign off on fuel and equipment testing. Legal liability for safety performance creates hard barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Nuclear facilities are heavily regulated (NRC oversight), require licensed engineers, and have extreme liability/safety requirements mandating human sign-off and physical presence for testing.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Nuclear testing infrastructure, equipment, and the specialized expertise required remain expensive. AI's contribution is primarily in data post-processing rather than replacing the core testing activities, so the all-in cost of AI-assisted testing remains comparable to or higher than traditional engineering approaches.
Cost vs. human wage	claude-sonnet-5	1/5	The physical testing infrastructure, specialized equipment, and safety protocols mean AI cannot substitute for the human-run experimental process, so no cost advantage exists for the core task.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Current AI systems can support analysis of test data and simulation modeling, but no deployed products perform end-to-end nuclear fuel testing or machinery performance assessment autonomously. Regulatory oversight and safety criticality mean that AI tools remain primarily assistive rather than autonomous performers.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product conducts physical nuclear fuel or equipment tests; this remains a specialized engineering/experimental activity performed by trained personnel with regulatory oversight.

Develop or contribute to the development of plans to remediate or restore environments affected by nuclear radiation, such as waste disposal sites.

CI 0–16 · exposure 13 · augmentation 50 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear remediation is a heavily regulated, safety-critical sector with long project timelines and strong institutional resistance to automation. Adoption of AI for autonomous plan development is essentially non-existent; the sector prioritizes human expertise and regulatory compliance over efficiency gains.
Sector adoption velocity	claude-sonnet-5	2/5	Nuclear engineering is a highly regulated, low-digitization niche sector where AI adoption for safety-critical planning has been slow and cautious.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist nuclear engineers by analyzing contamination data, searching relevant literature, running scenario simulations, and organizing regulatory information, thus supporting faster and more comprehensive plan development. However, the core judgment and responsibility remain with the licensed engineer.
Augmentation potential	claude-sonnet-5	3/5	AI can help analyze environmental data, model contamination spread, and draft portions of technical documentation, aiding engineers without replacing their judgment.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires domain-specific expertise in nuclear science, environmental assessment, regulatory compliance, and site-specific decision-making that current AI systems cannot perform end-to-end. The task involves creating novel remediation strategies for complex, safety-critical situations where human judgment and professional accountability are irreplaceable.
Task automatability	claude-sonnet-5	2/5	This requires site-specific engineering judgment, regulatory knowledge, and physical assessment that current AI cannot perform end-to-end; AI can assist with data analysis and drafting but not replace the core planning work.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Strong regulatory and legal barriers protect this task: remediation plans must be developed and signed off by licensed nuclear engineers under NRC and EPA oversight, liability for failures is severe, and regulatory frameworks explicitly require qualified human professionals for nuclear remediation decisions. No automation can bypass these licensing requirements.
Adoption barriers	claude-sonnet-5	5/5	Nuclear remediation plans require licensed professional engineers and regulatory approval (e.g., NRC oversight), with severe liability and safety consequences for errors.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The cost and overhead of integrating AI systems, validating outputs, and maintaining human expert oversight for safety-critical nuclear remediation planning far exceeds the cost of direct expert labor. The liability and verification burdens make AI economically unfavorable compared to licensed nuclear engineers.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot independently produce a compliant remediation plan, so any cost comparison requires substantial human expert oversight and verification, making AI-alone deployment not cost-competitive.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can assist with data analysis, literature review, and preliminary modeling, no deployed system reliably performs independent remediation plan development for nuclear sites. Current systems lack the specialized nuclear engineering domain knowledge, safety certification requirements, and accountability mechanisms needed for production deployment in this safety-critical context.
Technical feasibility today	claude-sonnet-5	1/5	No deployed products autonomously generate nuclear remediation plans; this remains a highly specialized engineering task performed by licensed professionals with no production AI substitute.

Design fuel cycle models or processes to reduce the quantity of radioactive waste generated from nuclear activities.

CI 0–15 · exposure 8 · augmentation 50 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	The nuclear engineering sector is conservative, heavily regulated, and slow to adopt automation for safety-critical design work. New fuel cycle models require multi-year validation and regulatory approval, precluding rapid AI-driven iteration.
Sector adoption velocity	claude-sonnet-5	2/5	The nuclear energy sector is conservative, highly regulated, and slow to adopt new technologies including AI, with pilots for AI-assisted modeling emerging but production-scale autonomous design tools not established.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI could assist with literature review, simulation data organization, or preliminary sensitivity analyses, but the core creative and safety-critical design work demands expert judgment that current AI augmentation does not substantially enhance.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist with complex simulations, data analysis, literature review, and iterative modeling scenarios, significantly speeding up parts of the fuel cycle design process while engineers retain final design authority.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires deep domain expertise, novel engineering design, physical and economic optimization tradeoffs, and regulatory compliance reasoning. Current AI cannot autonomously create safe, economically viable fuel cycle designs that satisfy nuclear engineering constraints and legal requirements.
Task automatability	claude-sonnet-5	2/5	This requires original engineering design, novel physics modeling, and creative optimization of fuel cycle processes that current AI cannot perform end-to-end reliably; AI can assist with simulations and calculations but not autonomously design validated fuel cycle processes.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear fuel cycle design is heavily regulated (NRC, IAEA standards) and requires licensed professional engineers to design and certify safety-critical processes. Legal liability for radioactive waste design falls on licensed humans, creating a hard authorization barrier.
Adoption barriers	claude-sonnet-5	5/5	Nuclear engineering design is subject to strict regulatory oversight (e.g., NRC licensing), requires certified professional engineers, and carries severe liability and safety consequences, making human sign-off legally mandatory.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	AI inference costs for such complex, novel design work would pale against the multi-year salaries of nuclear engineers and the domain expertise required; the human cost per design cycle far exceeds any plausible AI alternative.
Cost vs. human wage	claude-sonnet-5	2/5	AI tools can reduce some computational modeling costs, but the overall task still requires expensive human expert oversight, validation against safety standards, and specialized nuclear engineering judgment, keeping costs comparable to or only modestly below human-only approaches.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI system reliably generates novel fuel cycle process designs in production at nuclear facilities. This remains a specialized human expert task with no mature commercial product performing it end-to-end.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs nuclear fuel cycle design; this remains a highly specialized research and engineering task done by human experts using specialized simulation tools with AI playing at most a minor supporting role.

Discuss construction project proposals with interested parties, such as vendors, contractors, or nuclear facility review boards.

CI 0–14 · exposure 5 · augmentation 50 · importance 3.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	The nuclear industry is highly conservative, heavily regulated, and slow to adopt automation in high-stakes communication and decision-making contexts. Project discussion and vendor negotiation in nuclear construction remain firmly human-led processes with minimal documented AI displacement.
Sector adoption velocity	claude-sonnet-5	1/5	Nuclear engineering and construction are highly regulated, low-digitization, physical-infrastructure sectors with minimal AI agent deployment in stakeholder negotiation contexts.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist by pre-drafting talking points, analyzing vendor proposals, summarizing regulatory feedback, and preparing data-driven arguments—activities that improve engineer productivity before and after discussions. However, the interactive discussion itself remains human-centric, limiting overall augmentation impact.
Augmentation potential	claude-sonnet-5	3/5	AI can help engineers prepare briefing materials, summarize proposals, draft talking points, or analyze technical documents ahead of these discussions, offering meaningful but partial support.
Task automatability	claude-haiku-4-5-20251001	1/5	This task fundamentally requires nuanced negotiation, relationship-building, and real-time responsiveness to stakeholder concerns that current AI systems cannot perform autonomously. The task involves discussion and persuasion with multiple parties holding conflicting interests, which exceeds current AI capabilities for end-to-end autonomous execution.
Task automatability	claude-sonnet-5	1/5	This requires live, high-stakes negotiation and relationship-building with multiple stakeholders including regulatory boards, which involves judgment, trust-building, and real-time responsiveness that current AI cannot replicate end-to-end.},
Adoption barriers	claude-haiku-4-5-20251001	4/5	Nuclear facility projects are heavily regulated and stakeholder discussions often carry contractual, liability, and compliance weight that typically requires authorized personnel to be present and accountable. Regulatory frameworks and project governance structures create friction against full AI autonomy in discussions with official review boards and contractors.
Adoption barriers	claude-sonnet-5	5/5	Nuclear facility discussions involve licensed engineers, regulatory review boards, and legal/safety accountability, creating hard institutional and licensing barriers to any non-human substitution.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI assistance with preparation and follow-up documentation is inexpensive, but the core discussion task still requires human nuclear engineers whose fully-loaded cost far exceeds the marginal cost of AI-assisted drafting. Substitution is not economical because the human remains essential.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the human discussion itself, so there is no viable cost comparison for full task replacement.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can draft initial proposals or summarize discussion points, no deployed product reliably conducts live, multi-party technical negotiations involving vendors, contractors, and regulatory bodies. AI systems exist for meeting summarization and document drafting, but not for the interactive, real-time problem-solving inherent in stakeholder discussion.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product conducts substantive stakeholder discussions on nuclear construction proposals; this remains firmly in the human-judgment and interpersonal domain.

Direct environmental compliance activities associated with nuclear plant operations or maintenance.

CI 4–9 · exposure 5 · augmentation 50 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear power is a highly regulated, capital-intensive sector with strong institutional resistance to removing human oversight from compliance activities. Adoption of AI for autonomous compliance direction is essentially non-existent; the industry uses AI only for supporting analytics and reporting.
Sector adoption velocity	claude-sonnet-5	1/5	Nuclear industry is highly regulated, safety-critical, and slow to adopt new technology for core compliance management functions.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can usefully assist by processing environmental data, automating routine report generation, flagging regulatory changes, and analyzing monitoring results, allowing the engineer to focus on decision-making and strategy. However, the directing role itself remains fundamentally human.
Augmentation potential	claude-sonnet-5	3/5	AI tools can assist with regulatory document analysis, monitoring data aggregation, and report drafting, improving efficiency while the engineer retains directive responsibility.
Task automatability	claude-haiku-4-5-20251001	1/5	Directing environmental compliance activities requires judgment about regulatory requirements, site-specific conditions, and organizational strategy that are tied to human authority and accountability. Current AI cannot autonomously make compliance decisions or direct staff in contexts where regulatory sign-off must come from a licensed professional.
Task automatability	claude-sonnet-5	1/5	Directing compliance activities requires judgment, accountability, regulatory interpretation, and interfacing with staff and regulators—AI cannot perform this end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear plant operations are heavily regulated by the NRC and EPA. Environmental compliance direction must be performed or signed off by a qualified nuclear engineer, and liability for non-compliance falls on the facility and licensed staff. These regulatory and legal requirements create hard barriers to automation.
Adoption barriers	claude-sonnet-5	5/5	Nuclear regulatory frameworks (e.g., NRC requirements) mandate licensed, accountable engineers to direct compliance activities, creating strong legal and liability barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Automating this task is not economically viable because it requires a licensed nuclear engineer to maintain legal and operational authority. AI tools that support compliance work are relatively inexpensive, but they cannot substitute for the human director, making the cost ratio unfavorable for replacement.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cut costs on document review or reporting subtasks, but the directive/managerial role still requires a highly paid engineer, keeping overall cost comparable to human-only execution.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can assist with data gathering, report generation, and flagging potential issues, no deployed system can independently direct compliance activities for nuclear facilities. Regulatory agencies require human engineers to oversee and sign off on compliance; AI has no production role in directing these activities.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product manages or directs environmental compliance programs at nuclear facilities; this remains a human management function.

Design or direct nuclear research projects to develop, test, modify, or discover new uses for theoretical models.

CI 0–7 · exposure 0 · augmentation 50 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear engineering remains a tightly regulated, human-expert-dependent field with slow digitization and minimal AI adoption in research direction or project management. Organizational and regulatory conservatism in this safety-critical domain strongly limits adoption velocity.
Sector adoption velocity	claude-sonnet-5	2/5	Nuclear engineering and research is a highly specialized, low-digitization physical science field where AI adoption for core research direction remains nascent and pilot-stage at best.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI can assist with data analysis, literature synthesis, and simulation visualization, but the core task of designing research direction and judging theoretical merit depends critically on human insight and accountability. Assistance is marginal rather than transformative for this task.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist with literature review, simulation, data analysis, and hypothesis exploration, substantially boosting researcher productivity even though humans must direct and validate the research.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires creating novel theoretical models, formulating research hypotheses, and making strategic decisions about uncharted scientific territory. Current AI cannot independently conceive new research directions or validate theoretical soundness at the level required to meet a 50% time-saving bar for the core work.
Task automatability	claude-sonnet-5	1/5	This task requires original scientific creativity, novel hypothesis generation, and directing complex physical experiments—capabilities far beyond current AI systems' end-to-end abilities.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Nuclear research is heavily regulated by agencies like the NRC and DOE, with strict requirements for qualified personnel to certify research design and outcomes. Regulatory frameworks mandate human experts sign off on nuclear research projects, creating hard legal barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Nuclear research is heavily regulated, requires credentialed expert oversight, safety certifications, and legal accountability for outcomes, creating hard institutional and regulatory barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	A nuclear engineer's loaded cost is substantial (>$150k annually), and the overhead of human expertise in validating research direction far outweighs current AI inference costs. The specialized domain knowledge and liability make any cost advantage negligible.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the human expert director/designer role here, so there is no viable AI-only cost basis; human expertise remains essential and costly but irreplaceable.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI system can autonomously design or direct research projects in nuclear science. While AI can assist with literature review or simulation, directing research—setting objectives, prioritizing approaches, validating theoretical merit—remains a human cognitive task without production-scale automation.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product designs or directs nuclear research programs; this remains squarely in the research-stage domain requiring deep domain expertise and physical experimentation.

Initiate corrective actions or order plant shutdowns in emergency situations.

CI 0–0 · exposure 0 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear power plants operate in one of the most heavily regulated and conservative sectors. Actual automation of emergency decisions is legally prohibited and would require fundamental regulatory change, making real-world adoption velocity effectively zero.
Sector adoption velocity	claude-sonnet-5	1/5	Nuclear power is a highly regulated, safety-critical, low-digitization-for-control-authority sector with extremely cautious and slow adoption of autonomous decision-making systems.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist by rapidly aggregating sensor data, flagging anomalies, and recommending actions to human operators, improving their situational awareness during emergencies. However, the human operator must retain full decision authority.
Augmentation potential	claude-sonnet-5	3/5	AI-based monitoring, anomaly detection, and predictive analytics can assist engineers in identifying issues faster and simulating responses, improving situational awareness even though humans retain control.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires real-time assessment of plant safety conditions, judgment about severity, and authorization to shut down critical infrastructure. Current AI systems cannot reliably assess novel emergency scenarios or take legal responsibility for shutdown decisions that have cascading consequences.
Task automatability	claude-sonnet-5	1/5	This requires real-time judgment under high-stakes uncertainty, legal accountability, and physical/organizational authority to halt operations; no current AI system can autonomously perform this end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear Regulatory Commission licensing requirements mandate that a qualified human operator must authorize any emergency action or shutdown. Federal law and safety regulations explicitly require human authority and accountability for this decision.
Adoption barriers	claude-sonnet-5	5/5	Nuclear regulatory frameworks (e.g., NRC licensing) require certified human operators to authorize shutdowns and corrective actions, making this a hard legal barrier to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	A licensed nuclear engineer performing this task in an emergency carries legal and financial responsibility that AI cannot assume. Human judgment and oversight are non-negotiable, making true cost replacement infeasible.
Cost vs. human wage	claude-sonnet-5	1/5	Given catastrophic failure costs and lack of any viable autonomous substitute, cost comparison is moot—human oversight is mandatory regardless of AI cost advantages.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI system performs autonomous emergency shutdown decisions in nuclear facilities today. Nuclear regulation requires human operators with licenses to make these calls; AI can only support data aggregation and alerting, not the decision itself.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product initiates nuclear plant shutdowns or corrective safety actions autonomously; decision support tools exist but final action authority remains strictly human.

Direct operating or maintenance activities of nuclear power plants to ensure efficiency and conformity to safety standards.

CI 0–0 · exposure 0 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear energy is one of the most heavily regulated and conservative sectors; adoption of autonomous AI in operations is extremely limited and restricted to non-critical analytics. New plants still deploy traditional human-led control architectures.
Sector adoption velocity	claude-sonnet-5	1/5	The nuclear power sector is slow-moving, highly regulated, and conservative in adopting autonomous control technologies, with AI limited to advisory analytics rather than operational direction.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist engineers through anomaly detection, data analysis, predictive maintenance modeling, and compliance auditing—useful support that raises engineer productivity—but the human expert must remain in direct control and accountable for all safety-critical decisions.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with predictive maintenance analytics, anomaly detection, and compliance documentation, providing useful decision support while humans retain full directive control.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires real-time decision-making in safety-critical systems, hands-on physical maintenance, and accountability for catastrophic failure modes. Current AI cannot reliably replace the full chain of human judgment, coordination, and responsibility required in nuclear operations.
Task automatability	claude-sonnet-5	1/5	This is a supervisory, safety-critical directive role requiring real-time judgment, accountability, and physical presence; no AI system can autonomously direct plant operations or maintenance today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear Regulatory Commission (NRC) and equivalent international bodies legally require licensed human engineers to direct operations and sign off on safety decisions; licensing, liability, and statutory oversight make automation of human authority effectively illegal.
Adoption barriers	claude-sonnet-5	5/5	Nuclear plant operations are heavily regulated (NRC licensing, certified operator requirements) and mandate qualified human oversight and legal accountability, making substitution essentially prohibited.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	A nuclear engineer's loaded salary is high (>$100k), and the cost of AI oversight, integration, and liability insurance to potentially replace supervisory authority would far exceed the human cost, especially given regulatory requirements.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI substitute performing this directive role, so cost comparison favors the human engineer by default; any AI tooling only adds cost as an assistive layer.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed product performs autonomous direction of nuclear plant operations or real-time safety management. AI exists only as analytical support tools; actual plant control and maintenance direction remain entirely human-supervised in all operational contexts.
Technical feasibility today	claude-sonnet-5	1/5	No deployed products direct nuclear plant operations or maintenance activities; AI is used only for narrow decision-support or monitoring subtasks, not the directive function itself.

Design or oversee construction or operation of nuclear reactors, power plants, or nuclear fuels reprocessing and reclamation systems.

CI 0–0 · exposure 0 · augmentation 50 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear energy is a highly regulated, capital-intensive sector with long approval cycles. Adoption of autonomous AI in design or operational oversight is extremely slow; organizations prioritize human expert oversight and compliance rather than automation.
Sector adoption velocity	claude-sonnet-5	1/5	Nuclear power is a highly regulated, safety-critical, low-digitization physical infrastructure sector with minimal AI agent deployment in core engineering oversight roles.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist nuclear engineers through simulations, thermal modeling, data analysis of reactor parameters, and documentation review, raising productivity on sub-tasks. However, human judgment and accountability remain central to the role.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with simulations, data analysis, documentation, and design optimization subtasks, providing meaningful productivity gains while humans retain full oversight and decision authority.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires domain expertise in physics, safety protocols, and regulatory compliance that goes far beyond current AI capabilities. While AI can assist with calculations and simulations, the core responsibility of designing or overseeing construction/operation of nuclear systems—with catastrophic failure consequences—cannot be automated end-to-end by current systems.
Task automatability	claude-sonnet-5	1/5	This task requires physical construction oversight, complex multidisciplinary engineering judgment, safety-critical decision-making, and legal accountability that AI cannot perform end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Nuclear engineering is heavily regulated; only licensed Professional Engineers can legally design or sign off on reactor systems. Liability, safety certification, and regulatory requirements (NRC, IAEA) create hard legal and institutional barriers to automation or substitution.
Adoption barriers	claude-sonnet-5	5/5	Nuclear engineering design and oversight is subject to strict regulatory licensing (e.g., NRC requirements), professional engineer certification, and severe liability, making human sign-off legally mandatory.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	AI assistance tools (simulation software, optimization algorithms) cost far less than human engineers per hour, but the task itself cannot be delegated to AI—humans remain the primary cost and cannot be eliminated. All-in human costs remain the floor.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the human role here, so there is no viable cost comparison—human licensed engineers remain mandatory regardless of AI cost.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed product reliably performs nuclear reactor design or operational oversight autonomously. Current AI tools are research-stage for narrow sub-tasks like thermal modeling; no production system substitutes for licensed nuclear engineers in these critical functions.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product designs or oversees nuclear reactor construction or operation; this remains firmly in the domain of licensed human engineers with regulatory oversight.

Perform experiments that will provide information about acceptable methods of nuclear material usage, nuclear fuel reclamation, or waste disposal.

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

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Nuclear engineering is a highly regulated, safety-critical sector with slow adoption cycles and strong institutional resistance to replacing hands-on experimental work with automated systems due to safety, security, and compliance requirements.
Sector adoption velocity	claude-sonnet-5	1/5	Nuclear engineering is a highly regulated, physically-intensive field with minimal AI adoption for hands-on experimental work; digitization of core lab tasks remains low.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI could assist with data analysis, literature review, or simulation modeling to inform experiment design, but the core experimental work and safety-critical decision-making remain heavily human-dependent with limited augmentation potential.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with experimental design, data analysis, simulation modeling, and literature review to support the engineer's decision-making, though the physical experimentation itself remains human-driven.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires designing experiments, interpreting complex nuclear physics phenomena, and making safety-critical decisions about material handling. Current AI systems cannot autonomously conduct physical experiments involving nuclear materials or interpret novel experimental results at the level required for regulatory and safety compliance.
Task automatability	claude-sonnet-5	1/5	This task requires physical experimentation with hazardous nuclear materials, specialized lab equipment, and hands-on manipulation that current AI systems cannot perform end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Multiple hard barriers prevent substitution: nuclear engineering requires professional licensure, nuclear material handling is strictly regulated by agencies like the NRC, experiments must be conducted under security protocols, and liability for safety and environmental damage creates substantial legal constraints on automation.
Adoption barriers	claude-sonnet-5	5/5	Nuclear material handling is subject to strict regulatory oversight (e.g., NRC licensing), requiring credentialed human engineers and safety officers to legally perform and oversee experiments.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The overhead of supervision, verification, and regulatory compliance for any AI involvement would far exceed the cost of the nuclear engineer's labor, which is already expensive and irreplaceable given the expertise required.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical experimental apparatus, safety infrastructure, and specialized human expertise required, making cost comparison inapplicable to AI replacement of the core task.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI product performs nuclear material experimentation end-to-end. This demands hands-on laboratory work, real-time safety monitoring, and regulatory-grade documentation that remains firmly in the domain of human experts with specialized training and clearances.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product conducts physical nuclear experiments; this remains firmly in the domain of human researchers using specialized facilities.

Related occupations — Architecture & Engineering

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.