Occupational Health and Safety Technicians

19-5012.00

Median wage $61,560/yr30,590 employed (US)Rank #539 of 923 scored · top 58% by substitution

Collect data on work environments for analysis by occupational health and safety specialists. Implement and conduct evaluation of programs designed to limit chemical, physical, biological, and ergonomic risks to workers.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution25

Exposure24

Augmentation55

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

26 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%25

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

Technical feasibility todayw 20%22

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

Cost vs. human wagew 15%29

panel mean rating 2.1/5 → substitution pressure 29/100

Adoption barriersw 20%inverted — strong barriers lower the score29

panel mean rating 3.8/5 (barrier strength) → substitution pressure 29/100

Sector adoption velocityw 10%20

panel mean rating 1.8/5 → substitution pressure 20/100

Task breakdown (26 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.

Maintain all required environmental records and documentation.

CI 60–80 · exposure 78 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Environmental, health, and safety functions are increasingly digitizing and adopting RPA and document management platforms, but adoption remains uneven across small versus large firms and varies by regulatory regime; pilots are common but production-at-scale remains moderate.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety functions are typically embedded in industrial/physical worksites with slower digitization and AI adoption compared to pure information-sector roles.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI systems substantially assist technicians by auto-populating templates, flagging missing documentation, organizing records by compliance category, and generating compliance reports, significantly raising the technician's productivity while they retain oversight and judgment roles.
Augmentation potential	claude-sonnet-5	4/5	AI tools can significantly speed up record organization, templating, data entry, and flagging inconsistencies, meaningfully boosting technician productivity while they remain responsible for accuracy and compliance.
Task automatability	claude-haiku-4-5-20251001	5/5	Maintaining environmental records and documentation is highly routine, rule-based work involving data entry, file organization, and compliance checklist management—all tasks that current AI systems and RPA tools can handle end-to-end with well over 50% time savings once templates and data sources are mapped.
Task automatability	claude-sonnet-5	4/5	Records maintenance and documentation compilation is largely structured data entry, tracking, and formatting work that AI-assisted systems can handle end-to-end with substantial time savings, though initial data collection and verification often still requires human input.
Adoption barriers	claude-haiku-4-5-20251001	3/5	While environmental records have regulatory oversight requirements and may require technician sign-off or review, the actual maintenance and organization of records can be delegated to automated systems with human audit points; some industry regulations specify who may attest to accuracy, creating modest friction.
Adoption barriers	claude-sonnet-5	3/5	Environmental compliance records often require certified accuracy and may be subject to regulatory audit or sign-off by responsible personnel, creating moderate liability and procedural barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	5/5	Automating record maintenance through RPA, document management systems, or AI-powered data extraction costs a fraction of a full technician's loaded wage, often an order of magnitude cheaper when amortized across organizations with substantial compliance volumes.
Cost vs. human wage	claude-sonnet-5	4/5	Automated record-keeping and documentation tools are inexpensive to run relative to a technician's loaded wage for routine record maintenance, though some oversight cost remains.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	Deployed document management, RPA, and data-capture systems reliably perform record creation, organization, and compliance tracking in production environments across industries; minor limitations exist around novel formats or ambiguous inputs, but mainstream products handle standard environmental documentation reliably.
Technical feasibility today	claude-sonnet-5	3/5	Document management and compliance record-keeping software with AI features exist and are used in production, but full automation of accurate, audit-ready environmental documentation across varied regulatory schemes still requires human review and correction.

Maintain logbooks of daily activities, including areas visited or activities performed.

CI 65–72 · exposure 70 · augmentation 75 · 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	3/5	Adoption is moderate; many health and safety departments use digital logbooks and some employ AI transcription or calendar integration, but widespread end-to-end automation of activity logging remains incomplete in smaller and lower-digitization organizations.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety fields are moderately digitized but not leading adopters of AI tools; documentation automation is emerging but not yet widespread in this specific technician role.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI significantly enhances technician productivity by auto-populating logbooks from voice memos, camera feeds, or calendar data, freeing the technician to focus on actual site work rather than post-visit administrative tasks. The human remains responsible for review and sign-off.
Augmentation potential	claude-sonnet-5	4/5	AI can substantially assist technicians by auto-generating structured log entries from voice notes, photos, or sensor data, letting them focus on fieldwork while still reviewing and finalizing records.
Task automatability	claude-haiku-4-5-20251001	4/5	AI can capture and organize daily activity descriptions from various inputs (voice memos, emails, calendar data) and populate logbooks with high consistency, achieving significant time savings over manual entry. However, some context-dependent judgment about which activities warrant detailed logging may still require human review.
Task automatability	claude-sonnet-5	4/5	Logging daily activities is a structured documentation task that AI can largely handle via voice-to-text, templated entry generation, and integration with scheduling/location data, saving significant time versus manual entry.
Adoption barriers	claude-haiku-4-5-20251001	2/5	Minimal barriers exist: logbook maintenance is largely clerical and internal-facing, with no legal mandate that a licensed technician personally maintain records by hand. Most organizations face only light organizational friction in switching to AI-assisted logging.
Adoption barriers	claude-sonnet-5	2/5	No licensing requirement for logbook maintenance itself, though organizational recordkeeping standards and audit requirements may require human verification of entries for accuracy and accountability.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	AI transcription and automated log aggregation cost a fraction of technician time spent on manual entry and organization; integration costs are modest for standard enterprise tools. The ratio strongly favors AI, typically an order of magnitude cheaper than paying technician labor for this clerical work.
Cost vs. human wage	claude-sonnet-5	4/5	Automated logging via mobile apps, transcription, and simple NLP summarization is inexpensive compared to technician time spent manually writing entries, though some human verification is needed.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	Deployed systems (AI transcription, calendar parsing, log-management platforms with AI assistance) reliably extract and organize activity data in production environments across multiple sectors. Minor gaps exist in handling ambiguous or context-sensitive categorizations, but core logbook maintenance is well-established.
Technical feasibility today	claude-sonnet-5	3/5	Digital logging tools and AI transcription/summarization products exist and are used in field safety contexts, but full automated logbook maintenance integrated with technician workflows is not yet universally deployed or fully reliable without human review.

Examine credentials, licenses, or permits to ensure compliance with licensing requirements.

CI 51–71 · exposure 62 · augmentation 88 · 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	3/5	Adoption is middling; large regulated organizations (construction, healthcare, manufacturing) have begun piloting automated compliance checks, but many smaller firms and regional operators still rely on manual review, and full displacement remains uncommon in production.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety compliance functions are typically embedded in slower-moving, compliance-heavy organizational environments with limited AI deployment so far.
Augmentation potential	claude-haiku-4-5-20251001	5/5	AI dramatically augments human technicians by instantly retrieving, parsing, and flagging expired or non-compliant credentials, freeing them to focus on investigation, remediation, and judgment calls on borderline cases, significantly raising their verification throughput.
Augmentation potential	claude-sonnet-5	4/5	AI tools can significantly speed up credential verification, flagging expired or missing licenses for human review, improving efficiency while the technician retains final judgment.
Task automatability	claude-haiku-4-5-20251001	4/5	Current AI systems can reliably extract, parse, and validate credentials, licenses, and permits against regulatory databases with high accuracy. Document analysis and cross-referencing against compliance rules can be largely automated, though edge cases or novel credential formats may require human review, achieving the 50%-time-saving threshold in routine compliance checking.
Task automatability	claude-sonnet-5	3/5	AI can verify document validity, extract data, and cross-check against databases or rules, but flagging nuanced compliance issues or ambiguous cases still requires human judgment and follow-up investigation.
Adoption barriers	claude-haiku-4-5-20251001	3/5	Moderate barriers exist: some jurisdictions may legally require a licensed technician or manager to sign off on compliance findings, and liability concerns around false negatives create organizational friction. However, the task itself is not inherently gated by professional licensing.
Adoption barriers	claude-sonnet-5	3/5	While no law requires a human to personally review permits, occupational safety compliance often carries liability and regulatory accountability that push organizations to keep a human in the loop for sign-off.
Cost vs. human wage	claude-haiku-4-5-20251001	5/5	AI-based credential verification via automated scanning, database queries, and compliance rule-checking costs pennies to dollars per verification, vastly cheaper than paying a technician $20–$50/hour to manually review, cross-reference, and document each credential.
Cost vs. human wage	claude-sonnet-5	4/5	Automated document scanning and database lookups are far cheaper than manual review per unit, though integration with disparate state/local licensing systems adds some overhead.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	Deployed OCR and document-verification products exist in regulated industries (financial compliance, HR verification) and perform credential validation reliably in production at scale. However, some jurisdictions and credential types still require manual verification steps, preventing a perfect 5.
Technical feasibility today	claude-sonnet-5	3/5	Document verification and OCR-based credential-checking tools exist in production (e.g., background check platforms, licensing verification services), but full end-to-end compliance determination for safety-specific permits is narrower and less mature.

Review records or reports concerning laboratory results, staffing, floor plans, fire inspections, or sanitation to gather information for the development or enforcement of safety activities.

CI 34–54 · exposure 38 · augmentation 75 · 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	3/5	Safety-focused organizations and larger firms have begun deploying AI for document analysis and compliance monitoring, but adoption remains pilot-heavy rather than deeply embedded. Smaller facilities and traditional sectors lag significantly.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational safety functions are typically embedded in manufacturing, construction, and facilities sectors that have historically slower AI adoption compared to information/finance sectors, with pilots more common than production deployment.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI is already proving valuable in assisting technicians by automating record retrieval, flagging anomalies in lab results, and cross-referencing inspection data against regulatory baselines. This substantially accelerates the information-gathering phase while the technician retains judgment on enforcement and intervention.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully speed up gathering, organizing, and flagging relevant information across multiple record types (lab results, inspections, floor plans), letting technicians focus on judgment and enforcement decisions.
Task automatability	claude-haiku-4-5-20251001	2/5	AI can extract and summarize data from structured records and reports, but the task requires contextual judgment about safety implications, correlation across multiple document types, and identification of non-obvious patterns. End-to-end automation would still require significant human review and decision-making.
Task automatability	claude-sonnet-5	3/5	AI can extract and summarize structured/unstructured data from lab reports, inspection records, and floor plans, but synthesizing this into safety enforcement decisions requires domain judgment and physical-site context that limits full automation.rst
Adoption barriers	claude-haiku-4-5-20251001	3/5	Occupational health and safety work is regulated and documented, but technicians are not legally required to sign off on record reviews in most jurisdictions. However, organizational liability concerns and the need for human judgment on safety interventions create moderate friction to full automation.
Adoption barriers	claude-sonnet-5	3/5	No strict licensing requirement for record review itself, but safety enforcement decisions carry liability implications and often require sign-off by a qualified technician or officer, creating moderate friction.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	AI-based document review and summarization is substantially cheaper than manual record review once systems are deployed, with inference costs near commodity rates. Integration and oversight overhead is modest compared to the labor cost of a technician manually reading and synthesizing multiple reports.
Cost vs. human wage	claude-sonnet-5	3/5	AI-assisted document review can cut analyst time on report synthesis, but the need for human verification of safety-critical conclusions and integration across disparate record types (floor plans, lab data) keeps overall costs closer to comparable rather than order-of-magnitude cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Document processing and data extraction tools are mature (OCR, NLP, information extraction), but no production system fully performs this task autonomously. Deployed products handle individual document types reliably but struggle with cross-document synthesis and interpreting ambiguous safety contexts that technicians navigate.
Technical feasibility today	claude-sonnet-5	2/5	Document review and summarization tools exist and are used in compliance contexts, but no mature deployed product specifically integrates lab results, floor plans, fire inspections, and sanitation records into safety enforcement workflows reliably at scale.

Plan emergency response drills.

CI 25–60 · exposure 45 · augmentation 75 · 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	2/5	Occupational health and safety remains a laggard sector with limited digitization and conservative adoption patterns; most organizations still rely on manual planning and spreadsheets rather than AI-driven automation.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety is a slower-adopting, compliance-heavy field with limited AI agent deployment for physical-site emergency planning tasks.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI can significantly assist technicians by auto-generating drill schedules, suggesting scenarios based on regulatory and incident data, and automating compliance documentation, allowing humans to focus on scenario refinement and stakeholder engagement.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully help draft drill scenarios, checklists, schedules, and regulatory documentation, significantly speeding up the technician's planning work while they retain oversight.
Task automatability	claude-haiku-4-5-20251001	4/5	Scheduling, scenario design, communication logistics, and post-drill reporting can be largely automated with current systems; however, real-time adaptation and complex decision-making during drill execution typically require human oversight, preventing a full 5.
Task automatability	claude-sonnet-5	2/5	Planning emergency response drills requires site-specific knowledge, physical walkthroughs, coordination with personnel, and regulatory compliance judgment that AI cannot fully replicate end-to-end today.HTTP AI can assist with templates and scheduling but not the full task.
Adoption barriers	claude-haiku-4-5-20251001	3/5	No strict legal requirement that a human must perform drill planning in most jurisdictions, but organizational liability concerns, union agreements, and the expectation that safety leadership sign off on drills create moderate friction to full automation.
Adoption barriers	claude-sonnet-5	4/5	OSHA and similar regulations often require qualified personnel to design and validate emergency procedures, and liability for inadequate drills falls on certified professionals, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	Automation of planning, scheduling, documentation, and analysis can reduce labor by 60–70% compared to manual coordination; AI inference and oversight costs are substantially lower than the loaded wage for repetitive planning tasks.
Cost vs. human wage	claude-sonnet-5	2/5	Human safety technicians must inspect physical sites and coordinate real people, so AI only reduces some documentation time while the core planning labor cost remains largely human-driven.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed tools can generate drill schedules, populate communication templates, and produce compliance reports, but material gaps remain in scenario customization, resource allocation validation, and integration with organizational safety management systems in production.
Technical feasibility today	claude-sonnet-5	2/5	No deployed product autonomously plans and executes site-specific emergency drills; existing tools are generic checklist generators requiring heavy human customization.

Prepare or review specifications or orders for the purchase of safety equipment, ensuring that proper features are present and that items conform to health and safety standards.

CI 25–34 · exposure 25 · augmentation 63 · 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	Adoption in occupational health and safety remains slow; sectors are conservative due to liability exposure, and many organizations rely on human specialists and external consultants. Digital transformation in this domain lags compared to finance or information technology, with human expertise strongly preferred for risk-critical decisions.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety is a compliance-heavy, often lower-digitization field where AI tools are used piecemeal (e.g., for documentation) rather than deeply integrated into procurement workflows.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist by automating standard lookups, flagging relevant regulatory requirements, and generating initial draft specifications that technicians then review and refine. This assists productivity on the research and documentation aspects, though human judgment remains essential for final specification decisions.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist by drafting specification language, cross-referencing regulatory standards, and flagging missing safety features, significantly speeding up the technician's review process.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can help retrieve and cross-reference safety standards and generate draft specifications, the task requires domain expertise in occupational safety, context-specific risk assessment, and judgment about which features are appropriate for particular workplace hazards. Current AI systems cannot reliably perform end-to-end specification preparation and review with the quality assurance needed for legal and safety compliance.
Task automatability	claude-sonnet-5	2/5	AI can help draft or check specifications against known standards but verifying real-world equipment conformity, supplier quality, and site-specific hazards requires human judgment and physical inspection knowledge, limiting full automation.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Strong regulatory and organizational barriers exist: safety equipment purchases typically require sign-off by licensed occupational health and safety professionals, liability concerns are high given that improper specifications directly affect worker protection, and regulatory bodies like OSHA impose accountability requirements that typically necessitate human expert review and authorization.
Adoption barriers	claude-sonnet-5	3/5	While not always requiring a licensed professional, safety compliance sign-off often requires accountable personnel and organizational risk-management processes, creating moderate friction against full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The cost of integrating AI systems, training models on organizational safety standards, and maintaining oversight to ensure compliance would likely exceed the loaded wage of a technician performing this task, especially given the liability implications of errors in safety equipment specification.
Cost vs. human wage	claude-sonnet-5	3/5	AI-assisted drafting could reduce time spent on document preparation substantially, but the need for expert verification and liability review keeps overall costs only moderately below fully human-performed review.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI tools can assist with standard lookups and document generation, but no deployed product reliably performs full specification review and purchasing order preparation for safety equipment in production. Current systems lack the ability to verify conformance to multiple overlapping safety standards and make contextually appropriate equipment recommendations without significant expert oversight.
Technical feasibility today	claude-sonnet-5	2/5	General LLMs can assist in drafting purchase specs or checking documents against standards text, but no deployed product reliably handles full safety-equipment procurement review in production without significant human oversight.

Verify availability or monitor use of safety equipment, such as hearing protection or respirators.

CI 25–30 · exposure 25 · 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	Adoption of AI-driven monitoring in occupational safety remains pilot-stage in most sectors; traditional on-site inspection by certified technicians is still the standard, and regulatory conservatism and liability concerns slow AI deployment in safety-critical roles.
Sector adoption velocity	claude-sonnet-5	2/5	Industrial and manufacturing safety sectors adopt digital monitoring tools slowly compared to office-based domains, with pilots more common than widespread production deployment.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist technicians by flagging equipment that appears missing or improperly used via automated monitoring, generating inspection schedules, or maintaining digital logs, moderately improving productivity and compliance tracking while the technician retains decision authority and sign-off responsibility.
Augmentation potential	claude-sonnet-5	3/5	AI-powered camera systems and IoT sensors can flag potential PPE non-compliance for a technician to verify, meaningfully assisting monitoring efforts without replacing the human role.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI systems could monitor equipment usage via computer vision or sensor logs in controlled settings, this task requires physical verification, real-time presence on-site, and nuanced judgment about proper fit and condition that current AI cannot reliably perform end-to-end. Partial automation of logging or flagging non-compliance might save 20–30% of time, but falls short of the 50% threshold.
Task automatability	claude-sonnet-5	2/5	This requires physical presence to check whether equipment is present and being worn correctly on-site; current AI cannot physically inspect workplaces or verify PPE availability without extensive sensor/camera infrastructure.vitamin
Adoption barriers	claude-haiku-4-5-20251001	4/5	OSHA and workplace safety regulations typically require a qualified human (often a certified technician or supervisor) to inspect and certify safety equipment and compliance; liability for equipment failure rests with the organization and a responsible human, creating hard legal and regulatory barriers to full automation.
Adoption barriers	claude-sonnet-5	3/5	OSHA and safety regulations often require documented human verification and accountability for compliance monitoring, and privacy/legal concerns around continuous employee video monitoring add friction.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The cost of camera systems, sensor networks, integration, and ongoing human oversight for verification and exceptions often rivals or exceeds the loaded cost of a single technician performing periodic manual checks, especially in smaller or distributed work environments.
Cost vs. human wage	claude-sonnet-5	2/5	Deploying camera-based monitoring systems requires significant capital investment in cameras, integration, and human oversight of alerts, making it costly relative to a technician performing spot checks manually.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Deployed products exist for video monitoring and asset tracking but with significant limitations: computer vision struggles with occlusion and equipment variability, and falsely flagged non-compliance creates liability risk. No mature production system reliably performs the full task of verifying safety equipment availability and proper use at scale without substantial human oversight.
Technical feasibility today	claude-sonnet-5	2/5	Computer vision PPE-detection products exist (e.g., hard hat/vest detection in construction), but hearing protection and respirator fit verification are less mature and not widely deployed for compliance-grade monitoring.

Collect data regarding potential hazards from new equipment or products linked to green practices.

CI 25–30 · exposure 25 · 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	2/5	Occupational health and safety is a conservative, compliance-driven field with limited digitalization and slow technology adoption. Most organizations rely on traditional audits and in-person inspections, and regulatory inertia discourages rapid substitution of human expertise with AI systems.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety is a moderately digitized but physically-grounded field with slow, cautious AI adoption focused on compliance documentation rather than hazard detection.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can usefully assist by organizing product documentation, flagging known hazard patterns, and preparing checklists for technicians to review. However, augmentation is limited to pre-analysis and organization; the core work of judgment, physical inspection, and hazard validation remains firmly human-centered.
Augmentation potential	claude-sonnet-5	3/5	AI can help research green product specifications, summarize safety data sheets, and organize hazard information, meaningfully aiding technicians even though it can't replace on-site data collection.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can help summarize product specifications and flag known hazard categories, collecting comprehensive hazard data requires hands-on inspection, testing, and evaluation of physical equipment—domains where current AI systems have limited capability. The task also requires judgment about novel or context-specific hazards that go beyond pattern-matching to known categories.
Task automatability	claude-sonnet-5	2/5	Data collection on physical equipment hazards requires site visits, sensor readings, and physical inspection that AI cannot perform end-to-end; AI can assist with research and documentation but not the core data-gathering activity.itest
Adoption barriers	claude-haiku-4-5-20251001	4/5	Occupational health and safety data collection is heavily regulated (OSHA, industry standards, liability frameworks). Any automation or AI-assisted hazard assessment would require human validation and sign-off, and workers' safety depends on accuracy, creating strong organizational and legal barriers to unsupervised AI deployment.
Adoption barriers	claude-sonnet-5	3/5	Occupational safety assessments often require certified professionals and carry liability implications if hazards are missed, though data collection itself isn't strictly licensed in all jurisdictions.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Collecting hazard data on-site and in real equipment contexts requires human technicians (labor cost ~$30–50/hour loaded). AI assistance for document review and data organization offers modest savings, but the core work—inspection and evaluation—still requires human field time, keeping overall cost comparable or slightly favoring humans.
Cost vs. human wage	claude-sonnet-5	2/5	Human technicians must physically inspect and measure hazards on-site, so AI cannot substitute at lower cost for the core data collection, though it may reduce time on research/documentation.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No production system reliably performs end-to-end hazard data collection and assessment for new equipment. AI tools can assist with documentation review and literature searches, but actual hazard identification demands physical inspection, measurements, and expert judgment that current deployed systems cannot execute independently.
Technical feasibility today	claude-sonnet-5	2/5	No deployed product autonomously collects field hazard data on new equipment; existing tools support document search and report drafting but not physical hazard identification.

Train workers in safety procedures related to green jobs, such as the use of fall protection devices or maintenance of proper ventilation during wind turbine construction.

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

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Safety training in construction and green energy sectors remains heavily regulated and human-centric; adoption of AI for core instruction is slow. While VR and simulation tools are emerging, most organizations still rely on traditional in-person or instructor-led online training.
Sector adoption velocity	claude-sonnet-5	2/5	Construction and renewable energy sectors are relatively slow adopters of AI-driven training tools compared to information/professional services, with physical safety training still predominantly in-person.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist instructors by generating customized training materials, quizzes, and compliance documentation, and by automating record-keeping. However, augmentation is limited to peripheral tasks; the core skill of teaching workers safe technique and behavior requires human expertise and presence.
Augmentation potential	claude-sonnet-5	3/5	AI can enhance training through interactive content, VR simulations, and personalized quizzes, improving efficiency of the instructional component even though hands-on elements remain human-led.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can generate safety training materials and content, the task requires real-time demonstration, hands-on practice feedback, and judgment about worker comprehension—elements that demand human instruction. Partial automation of content creation or assessment exists, but end-to-end replacement with 50% time savings is not achievable today.
Task automatability	claude-sonnet-5	2/5	Hands-on physical safety training involving equipment demonstration, worker observation, and real-time correction cannot be fully automated; AI could support content creation but not deliver the physical instruction and assessment.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	OSHA and regulatory bodies often require that safety training be conducted or verified by qualified instructors, and liability for inadequate safety training creates strong accountability requirements. Organizations typically require human sign-off on worker competency in high-risk tasks like fall protection.
Adoption barriers	claude-sonnet-5	4/5	OSHA and industry regulations often require documented, competent-person-led training and hands-on verification for fall protection and confined space/ventilation safety, creating strong regulatory and liability barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	While AI can reduce preparation time and produce templates, full training delivery—including instructor labor for live sessions, Q&A, and practical assessment—still requires human trainers. The all-in cost of AI-assisted training remains comparable to or higher than traditional instructor-led delivery.
Cost vs. human wage	claude-sonnet-5	2/5	AI-based training modules can be cheap to deploy at scale, but the need for hands-on demonstration, equipment fitting, and verification of competency requires human trainers, keeping overall costs comparable to or only modestly below human-led training.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI-generated training modules and videos exist, but no deployed system reliably conducts live safety instruction, corrects technique errors in real-time, or adapts to group dynamics. Products for content generation are mature; live training delivery and competency verification remain primarily human-delivered.
Technical feasibility today	claude-sonnet-5	2/5	Some e-learning and VR training modules exist for safety procedures, but they are supplementary tools rather than replacements for in-person, hands-on safety training with physical equipment.

Provide consultation to organizations or agencies on the workplace application of safety principles, practices, or techniques.

CI 25–25 · exposure 25 · augmentation 75 · 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	OHS is a safety-critical field with slower AI adoption; most organizations still rely on human consultants for customized advice. Pilot projects exist for compliance checking, but production-scale replacement of consultation services remains rare.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational safety is a moderately digitized but physically grounded field; AI adoption for consultation-type tasks remains in early pilot stages rather than widespread production use.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI can assist significantly by generating compliance checklists, summarizing regulations, identifying common hazard patterns, and drafting reports—allowing human OHS professionals to focus on workplace assessment and stakeholder communication. This assistive role is already emerging in practice.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist by synthesizing regulations, generating hazard checklists, and drafting reports, enhancing the consultant's efficiency even though the human remains central to judgment and delivery.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can help draft safety documentation and summarize regulations, consultation requires judgment about organizational context, culture, and specific hazard assessment that demands human expertise. Current AI cannot reliably conduct the holistic workplace assessment and custom recommendations that define effective consultation.
Task automatability	claude-sonnet-5	2/5	Consultation requires site-specific judgment, physical inspection, and contextual understanding of workplace hazards that current AI cannot fully replicate end-to-end.dev AI can support research and drafting but not replace the consultative engagement itself.
Adoption barriers	claude-haiku-4-5-20251001	4/5	OHS consultation often involves regulatory compliance advice where errors carry liability; many jurisdictions require a licensed safety professional to sign off on recommendations. Organizational liability concerns and professional licensing requirements create substantial barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Safety consulting often involves regulatory compliance (OSHA), professional certification, and liability exposure, creating meaningful barriers to full AI substitution, especially where sign-off or expert judgment is legally required.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	A typical OHS technician consult costs $150–300/hour loaded; AI assistance on parts of the task (drafting, research) may reduce preparation time, but the core consultation delivery remains human-centric. Full end-to-end automation would save costs, but feasibility is too low for meaningful comparison.
Cost vs. human wage	claude-sonnet-5	2/5	Given the need for human site visits, liability, and contextual judgment, AI alone cannot substitute for the human consultant, so cost savings are limited to partial task support rather than full replacement.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Some AI tools exist for safety compliance document generation and regulatory lookup, but no deployed product reliably performs end-to-end consultation at production scale. Consultation requires understanding nuanced workplace conditions and stakeholder needs that current systems handle inconsistently.
Technical feasibility today	claude-sonnet-5	2/5	No deployed product performs full safety consultation reliably; existing tools are limited to compliance checklists, document search, or drafting assistance rather than autonomous consultation.

Recommend corrective measures to be applied based on results of environmental contaminant analyses.

CI 25–25 · exposure 25 · augmentation 63 · 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	OHS technician roles are distributed across small and medium enterprises with slow digitization; adoption of AI-driven recommendation systems remains in pilot phases, with most organizations still relying on human expert judgment and established protocols.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational safety is a highly regulated, physically-grounded field with slow AI adoption; pilots exist but production-level agentic deployment for hazard mitigation decisions is rare.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can usefully assist by summarizing contaminant data, flagging regulatory thresholds, and suggesting candidate measures, allowing technicians to focus on context-sensitive vetting and implementation planning. The human remains the decision-maker but with productivity gains on data synthesis.
Augmentation potential	claude-sonnet-5	4/5	AI can effectively analyze contaminant data patterns, cross-reference regulatory thresholds, and draft recommendation options, substantially aiding technicians who retain final judgment.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can analyze environmental contaminant data and suggest generic corrective actions, the task requires contextual judgment about facility-specific constraints, regulatory nuances, and implementation feasibility that exceed current automation thresholds. End-to-end automation with 50% time savings at equal quality is not yet demonstrated.
Task automatability	claude-sonnet-5	2/5	Generating recommendations requires interpreting site-specific data, regulatory context, and physical hazard nuances that AI cannot reliably synthesize end-to-end without expert verification.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory frameworks (OSHA, EPA, state occupational health codes) often require a licensed or certified technician to sign off on corrective actions and ensure compliance; liability for inadequate recommendations creates high error-cost asymmetry, making legal authorization barriers substantial.
Adoption barriers	claude-sonnet-5	4/5	Safety-critical determinations often require sign-off by certified occupational health/safety professionals, and regulatory liability for faulty corrective measures creates strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Current AI systems require significant setup, domain expert oversight, and integration costs to validate outputs; this overhead often equals or exceeds the cost of a technician's direct analysis, especially for novel or complex contamination scenarios.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cheaply draft suggestions from data, but the liability and validation overhead by a qualified technician largely offsets savings, keeping costs comparable to human-only work.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No mature production system reliably generates actionable, organization-specific corrective recommendations from contaminant analyses. Research prototypes and lab tools exist, but deployed products that consistently produce reliable recommendations without human expert review are not established in real OHS workflows.
Technical feasibility today	claude-sonnet-5	2/5	No deployed product autonomously issues corrective action recommendations for contaminant analyses in production safety workflows; existing tools are decision-support only, requiring technician review.

Educate the public about health issues or enforce health legislation to prevent disease, to promote health, or to help people understand health protection procedures and regulations.

CI 25–25 · exposure 25 · augmentation 50 · 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	Health and safety enforcement is embedded in public health agencies and occupational sectors with moderate digitization; adoption of AI for core enforcement and public-facing education remains in pilot stages rather than production deployment at scale.
Sector adoption velocity	claude-sonnet-5	2/5	Public health and regulatory enforcement sectors are traditionally slow to adopt AI due to legal, bureaucratic, and public trust constraints.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can usefully assist by drafting and translating educational content, summarizing regulatory changes, and generating inspection checklists, materially improving human technician productivity while they retain authority and judgment.
Augmentation potential	claude-sonnet-5	3/5	AI can help draft educational materials, answer common questions, and summarize regulations, meaningfully supporting technicians without replacing enforcement duties.
Task automatability	claude-haiku-4-5-20251001	2/5	AI can assist with creating educational materials and drafting health communications, but the task critically requires interpersonal engagement, judgment about audience-specific messaging, and real-time adaptation to public questions—elements that demand human presence and cannot achieve 50% time savings end-to-end with current systems.
Task automatability	claude-sonnet-5	2/5	Some educational content generation can be automated (e.g., drafting materials, FAQs), but enforcement of health legislation and interactive public education require human judgment, authority, and physical presence.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Health regulation enforcement often carries legal authority requirements—technicians may need certifications to inspect facilities or issue citations—and health education requires trust and human accountability, creating substantial organizational and regulatory friction against full automation.
Adoption barriers	claude-sonnet-5	4/5	Enforcement of health legislation typically requires legal authority vested in certified public officials, and liability for enforcement errors is high, creating strong regulatory and legal barriers.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Generating educational materials via AI is cheap, but oversight, fact-checking, regulatory compliance review, and human delivery of enforcement/education typically cost more to validate safely than hiring human technicians for the full task.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cheaply generate educational materials, but enforcement activities still require paid human technicians with authority, keeping overall cost comparable to human labor.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can draft health education content and summarize regulations, no deployed product reliably conducts multi-stakeholder public education or enforcement interactions at scale; such work typically requires human judgment, authority, and accountability that AI cannot currently substitute.
Technical feasibility today	claude-sonnet-5	2/5	AI chatbots and content tools exist for health education dissemination, but no deployed product performs legislative enforcement or comprehensive public health education reliably at scale.

Collect data related to ecological or human health risks at brownfield sites.

CI 20–25 · exposure 20 · augmentation 50 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Brownfield assessment remains a specialized, regulatory-constrained domain with slower digitization than mainstream sectors. Adoption of fully autonomous data collection is laggard; most sites still rely on certified technicians conducting field work with manual reporting, reflecting low organizational pressure and high compliance friction.
Sector adoption velocity	claude-sonnet-5	2/5	Environmental health and safety is a moderately digitized but physically-grounded sector with slow adoption of AI for fieldwork; software tools for data management are used but on-site collection remains manual.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI provides useful assistance in post-collection analysis—flagging anomalies in lab data, cross-referencing historical site records, generating preliminary risk profiles, and automating reporting—but does not meaningfully augment the technician's on-site hazard detection and sampling work itself.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with planning sampling strategies, analyzing lab results, mapping contamination patterns, and drafting reports, meaningfully boosting technician productivity in non-field aspects of the task.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with data collation and analysis of existing records, the task fundamentally requires on-site physical sampling, sensing, and real-time visual inspection of brownfield conditions that current autonomous systems cannot reliably perform unsupervised. Meaningful automation would require robotic field deployment with environmental sensors, which exists in research/pilot forms but not at production scale for this specific application.
Task automatability	claude-sonnet-5	2/5	Data collection at brownfield sites requires physical site visits, sampling of soil/water/air, and use of field instruments, which AI cannot perform end-to-end; only data logging, analysis, and reporting portions are automatable today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Strong regulatory and liability barriers exist: OSHA, EPA, and state environmental regulations often mandate documented human inspection, professional certification (e.g., industrial hygienist), and legal accountability for health/safety determinations. Insurance and liability asymmetry heavily favor human sign-off on hazard assessments at contaminated sites.
Adoption barriers	claude-sonnet-5	4/5	Environmental regulations (e.g., EPA brownfield assessment protocols) often require certified professionals to collect and certify samples, and liability for inaccurate risk data is high, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Occupational health and safety technicians earn substantial wages (~$40–50k/year loaded); deploying robotic systems, environmental sensors, and integration infrastructure for autonomous brownfield assessment currently exceeds this cost-per-site unless sites are numerous and standardized, making AI more expensive for typical operations.
Cost vs. human wage	claude-sonnet-5	2/5	Field sampling still requires trained technicians, protective equipment, and physical instrumentation; AI only reduces costs in the data processing/reporting phase, not the costly physical collection itself.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Specialized environmental monitoring products exist, but deployed autonomous systems for brownfield site assessment remain limited and typically require significant human oversight. Current AI excels at analyzing lab results and historical data post-collection, but field data acquisition itself—visual hazard identification, soil/water sampling protocols, sensor calibration—relies heavily on trained technicians.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product autonomously collects environmental samples or performs field surveys at contaminated sites; this remains a physical, human-executed task with sensors and lab work.

Examine practices at green building sites to determine whether adherence to green building standards alters risks to workers.

CI 14–30 · exposure 20 · augmentation 50 · importance 2.6/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Occupational safety is a traditional, compliance-driven field with established human expertise requirements. Adoption of AI tools for green building risk assessment remains in pilot phases at best; most construction firms still rely on in-person safety inspections by certified technicians.
Sector adoption velocity	claude-sonnet-5	1/5	Construction and physical safety inspection sectors are among the slowest to adopt AI tools, with low digitization and heavy reliance on in-person assessment.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist by pre-screening site images for code violations, summarizing green standards documentation, and flagging anomalies for technician review, materially reducing manual analysis burden. However, the human technician remains essential for contextual judgment and sign-off.
Augmentation potential	claude-sonnet-5	3/5	AI can help technicians research green building standards, generate checklists, analyze photos/sensor data, and draft reports, meaningfully aiding parts of the task even though the physical inspection remains human-led.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can analyze documentation and identify standards non-compliance patterns, the task requires on-site examination of physical conditions, worker behavior, and contextual risk assessment that demands human presence and judgment. Current AI lacks the multimodal sensing and real-time hazard discernment needed for end-to-end autonomy.
Task automatability	claude-sonnet-5	2/5	This requires physical site inspection, observation of worker practices, and contextual judgment about novel green-building materials/methods that AI cannot perform end-to-end without a human physically present.》Some report drafting and hazard checklist cross-referencing could be automated, but the core inspection task cannot.》
Adoption barriers	claude-haiku-4-5-20251001	4/5	Occupational health and safety determinations often trigger legal liability and regulatory reporting obligations (OSHA, state labor boards). A qualified, licensed safety professional is frequently required to certify findings, creating a hard barrier to full automation regardless of technical capability.
Adoption barriers	claude-sonnet-5	3/5	Occupational safety inspections often require qualified personnel and documentation for liability/regulatory compliance, though not always a specific license, creating moderate friction against full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	On-site examination requires human presence; AI can only support via analysis of photos or reports. The combined cost of human site visits plus AI analysis overhead makes displacement uneconomical compared to a single trained technician performing the integrated task.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot replace the physical presence and judgment needed on-site, so the human cost dominates; any AI use is a minor supplement, not a substitute reducing overall cost.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Computer vision and document analysis tools exist to flag standards violations and hazard indicators from images or text, but no production systems reliably perform the holistic risk assessment combining site conditions, worker practices, and green standards interpretation at the quality required for occupational health decisions.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously conducts physical safety inspections at construction sites and correlates green building standard adherence with worker risk; this remains a human field task.

Conduct interviews to obtain information or evidence regarding communicable diseases or violations of health or sanitation regulations.

CI 18–25 · exposure 20 · augmentation 50 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Public health and occupational safety sectors move slowly on automation; interviews for regulatory compliance remain inherently human-centric due to legal requirements and the need for credible witness testimony.
Sector adoption velocity	claude-sonnet-5	2/5	Public health and occupational safety inspection sectors are slow to adopt AI-driven interviewing tools due to legal, ethical, and trust considerations.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI could assist by preparing interview templates, suggesting follow-up questions, transcribing audio, or organizing findings after the human conducts the interview—moderately useful augmentation of technician productivity without full automation.
Augmentation potential	claude-sonnet-5	3/5	AI can help technicians prepare interview questions, transcribe and summarize responses, and flag inconsistencies, improving efficiency without replacing the human interviewer.
Task automatability	claude-haiku-4-5-20251001	2/5	Interviews require nuanced questioning, rapport-building, and real-time adaptation to responses—capabilities at which current AI struggles. While AI could draft questionnaires or transcribe responses, conducting the actual interview to extract reliable evidence would require human judgment and interpersonal skill that AI cannot fully replace today.
Task automatability	claude-sonnet-5	2/5	Interviewing requires in-person presence, adaptive follow-up questioning, and judgment about credibility and context that current AI cannot reliably replicate end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory and legal barriers are substantial: evidence gathered for health/sanitation violations typically requires human attestation, chain-of-custody documentation, and testimony that would not stand if an AI conducted the primary interview. Liability and admissibility concerns strongly protect human-conducted interviews.
Adoption barriers	claude-sonnet-5	4/5	Regulatory and legal evidentiary standards often require a qualified official to conduct interviews and document findings, creating strong institutional and legal barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The cost of AI infrastructure plus required human oversight (verifying interview quality, re-interviewing subjects, legal review) likely exceeds the loaded wage of a trained technician conducting the interview directly.
Cost vs. human wage	claude-sonnet-5	2/5	AI could assist with transcription or question prep cheaply, but the core interview and evidence-gathering still requires a paid human investigator, limiting overall cost savings.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably conducts investigative interviews end-to-end. AI chatbots exist but lack credibility with subjects, cannot detect evasion or unreliability, and would not satisfy legal/regulatory standards for evidence gathering in health violation cases.
Technical feasibility today	claude-sonnet-5	1/5	No deployed products conduct compliance or disease-investigation interviews autonomously; this remains a human field task requiring rapport-building and situational judgment.

Test or balance newly installed HVAC systems to determine whether indoor air quality standards are met.

CI 5–30 · exposure 13 · 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	2/5	HVAC technician work remains concentrated in smaller, less-digitized service sectors with slow tech adoption; while some firms use software for logging and scheduling, autonomous or AI-driven testing adoption is minimal and confined to pilot programs in large enterprises.
Sector adoption velocity	claude-sonnet-5	1/5	Building trades and facilities/HVAC sectors show low digitization and minimal AI agent adoption for physical inspection and balancing work.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist by automating data logging, flagging anomalies in readings against standards, and recommending balancing adjustments based on historical patterns, thereby speeding analysis and decision-making while the technician remains responsible for on-site measurement and system adjustment.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with analyzing collected sensor data, generating compliance reports, or flagging anomalies against IAQ standards, though the core physical testing remains manual.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist in data analysis and logging HVAC test results, the task requires hands-on equipment operation, physical measurement at multiple building locations, and real-time troubleshooting of complex mechanical systems—activities that demand physical presence and cannot be fully automated by current AI without significant human involvement in the testing itself.
Task automatability	claude-sonnet-5	1/5	This requires physical presence with instruments (anemometers, particle counters, balometers) to measure airflow and pollutant levels in a real building; no AI system can physically perform this measurement task.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Many jurisdictions require HVAC work and indoor air quality certification by a licensed professional; building codes and regulatory standards often mandate that a qualified technician certify compliance, creating a legal signature requirement that prevents full substitution regardless of technical capability.
Adoption barriers	claude-sonnet-5	3/5	While not always legally requiring a specific license, HVAC testing/balancing often follows industry certification standards (e.g., NEBB, AABC) and building code compliance, creating moderate professional barriers.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	HVAC balancing requires specialized equipment, on-site technical labor, and liability; even with AI-assisted data analysis, the core task still demands a licensed technician's time, making the total cost comparable to or higher than pure human execution, especially considering oversight and equipment costs.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical measurement work, so there is no AI cost basis to compare; a human technician with equipment remains the only option.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI system performs end-to-end HVAC testing and balancing autonomously today. Some software tools assist in data logging and standard interpretation, but the actual physical testing, sensor placement, equipment adjustment, and validation require human technicians on-site; capability exists only at research or narrow pilot stages.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product performs physical HVAC testing and balancing; this remains a hands-on technician task requiring on-site instrumentation.

Supply, operate, or maintain personal protective equipment.

CI 7–25 · exposure 13 · 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	2/5	While some sectors (manufacturing, healthcare) use AI-assisted inventory and maintenance tracking, on-site PPE supply, operation, and maintenance remain predominantly manual across most industries; pilot adoption of full automation is minimal.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational safety functions in physical workplace settings show slower AI adoption compared to information-based professional services, though some digitization of tracking/inventory exists.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI augmentation is useful for PPE inventory optimization, wear detection via computer vision, and maintenance scheduling reminders, assisting technicians in managing stock and planning; however, the core physical supply and operation tasks remain fundamentally human-centered.
Augmentation potential	claude-sonnet-5	2/5	AI can assist with inventory tracking, maintenance scheduling, and record-keeping for PPE programs, but offers minimal help with the physical acts of supplying, fitting, or maintaining equipment.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI could assist with inventory tracking and maintenance scheduling of PPE, the core task of physically supplying, operating (donning/doffing), and maintaining protective equipment requires human handling and real-time physical interaction that current systems cannot perform end-to-end with 50% time savings at equal safety quality.
Task automatability	claude-sonnet-5	1/5	This is a physical task involving handling, fitting, inspecting, and maintaining equipment like respirators, harnesses, and protective clothing, which requires manual dexterity and physical presence that current AI cannot replicate.
Adoption barriers	claude-haiku-4-5-20251001	4/5	OSHA regulations and workplace safety standards typically require that PPE be properly supplied, fitted, and maintained by trained human technicians; liability and error-cost asymmetry (improper PPE can cause serious injury) create strong regulatory and organizational barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	OSHA and similar regulations often require qualified personnel to inspect, fit-test, and certify PPE, creating both regulatory and liability barriers to full automation of this task.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI solutions for PPE management (tracking, predictive maintenance) involve significant integration and infrastructure costs that do not yet compete favorably with the labor cost of a technician handling these straightforward physical tasks at scale.
Cost vs. human wage	claude-sonnet-5	1/5	There is no AI substitute performing this physical supply/maintenance function, so the human cost is the only viable option and AI cannot reduce it.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Deployed products exist for PPE inventory management and predictive maintenance, but no current AI system reliably performs the physical supply, operation, or maintenance of PPE in production settings; robotic systems capable of this remain experimental and not widely deployed.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product supplies, fits, or physically maintains PPE; this remains entirely a human physical activity in workplace safety programs.

Test workplaces for environmental hazards, such as exposure to radiation, chemical or biological hazards, or excessive noise.

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

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Occupational health and safety remains heavily regulated and human-centric; while data analytics are adopted, autonomous field testing adoption is slow because regulatory bodies and employers prefer human professional judgment and certification.
Sector adoption velocity	claude-sonnet-5	1/5	Occupational safety inspection is a physical, on-site, low-digitization sector with minimal AI agent deployment for the core testing function.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist technicians by automating data analysis, generating preliminary reports, flagging anomalies, and recommending further testing—improving interpretation speed and consistency—while the technician remains responsible for field measurement and decision-making.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with data logging, hazard trend analysis, report generation, and interpreting sensor data, improving technician efficiency even though it cannot perform the physical testing itself.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can analyze hazard data and generate reports, the core task requires physical testing equipment deployment and real-time measurement in varied workplace environments—activities that demand robotics, field calibration, and on-site judgment that current AI systems cannot reliably perform end-to-end.
Task automatability	claude-sonnet-5	1/5	This requires physical presence, use of specialized sensing equipment, and hands-on sampling in real workplaces; no AI system can physically test environments today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory requirements (OSHA, EPA, industry standards) often mandate licensed or certified professionals conduct environmental testing and sign off on results; liability and legal defensibility of automated testing create significant adoption friction.
Adoption barriers	claude-sonnet-5	4/5	Regulatory frameworks (OSHA, EPA) often require certified technicians to conduct and certify hazard testing, and liability for missed hazards is high, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Even where AI can assist with data analysis, the upfront cost of specialized testing equipment, field deployment, calibration, and human oversight remains substantial; AI's cost advantage is modest compared to the loaded cost of a technician performing direct measurements.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical measurement equipment and technician labor required, so there is no viable cost comparison favoring AI for the core task.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI-powered analysis tools exist for interpreting environmental data (e.g., spectroscopy, noise measurements) after collection, but deployed products that autonomously conduct workplace hazard testing from setup through measurement and documentation are limited; human technicians typically conduct the actual field testing.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product performs physical environmental hazard testing; this remains a research-stage or non-existent capability for AI itself, though sensor hardware exists independently.

Prepare or calibrate equipment used to collect or analyze samples.

CI 5–25 · exposure 13 · augmentation 38 · 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	Adoption remains slow outside large pharmaceutical and clinical labs; most occupational health labs are in small-to-medium enterprises with legacy equipment, budget constraints, and low digital maturity, where full automation pilots remain uncommon.
Sector adoption velocity	claude-sonnet-5	1/5	Occupational safety fieldwork is a physical, hands-on sector with low AI/robotics penetration and no significant automation trend for equipment calibration.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-assisted tools (automated data logging, calibration reminders, anomaly detection in sample parameters) can improve technician productivity and reduce manual error, but the core task remains hands-on and judgment-dependent, limiting transformative potential.
Augmentation potential	claude-sonnet-5	2/5	AI can help track calibration schedules, log data, or flag anomalies in readings, but offers minimal assistance for the physical calibration process itself.
Task automatability	claude-haiku-4-5-20251001	2/5	Sample preparation and calibration have procedural elements that could be partially automated (e.g., automated liquid handlers, calibration scripts), but they typically require hands-on physical manipulation, real-time sensory feedback, and judgment about equipment state that current AI systems cannot reliably execute end-to-end without human oversight.
Task automatability	claude-sonnet-5	1/5	This requires physical manipulation of instruments (calibration gases, sensor adjustments, physical setup) which current AI systems cannot perform without robotic embodiment.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Significant barriers exist: regulatory requirements (ISO, GLP, OSHA standards) often mandate that qualified personnel oversee or sign off on sample preparation and calibration; liability for contamination or measurement error falls on the organization; and compliance audits expect documented human accountability.
Adoption barriers	claude-sonnet-5	4/5	Calibration often follows regulatory/manufacturer protocols requiring certified technician sign-off, and errors have safety/compliance consequences, creating strong barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Specialized lab automation equipment is capital-intensive and requires integration; total cost of ownership (hardware, maintenance, training, oversight) typically exceeds the wage of a single technician, especially for small-to-medium labs.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI substitute performing this physical task, so AI cost is not comparable—human labor remains the only option.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While some laboratory automation platforms exist for sample preparation, deployed products are narrow in scope, require extensive customization, and depend on human technicians for problem-solving, adjustment, and verification—not yet reliable autonomous operation at scale.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs physical equipment calibration for safety sampling instruments; this remains a manual technical task.

Conduct worker studies to determine whether specific instances of disease or illness are job-related.

CI 5–25 · exposure 13 · augmentation 50 · 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	Occupational health and safety remains a heavily regulated, compliance-driven sector with limited digital transformation. Adoption of AI for disease causality determination is laggard due to liability concerns and regulatory requirements for human professional involvement.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety is a slower-adopting, compliance-heavy, physically-grounded field with limited AI agent deployment in production compared to information/finance sectors.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist technicians by summarizing medical literature, organizing exposure data, flagging relevant precedents, and highlighting potential occupational connections, thereby accelerating the investigation process while the technician retains decision authority.
Augmentation potential	claude-sonnet-5	3/5	AI can help analyze exposure data, medical literature, and statistical correlations to support technicians' investigations, though it cannot replace on-site assessment and interviews.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires detailed investigation of worker medical history, workplace exposure, and causal inference linking illness to job conditions. Current AI systems cannot independently conduct the clinical interviews, exposure assessments, and complex epidemiological reasoning needed to make reliable job-relatedness determinations.
Task automatability	claude-sonnet-5	2/5	This requires physical site observation, worker interviews, data collection, and judgment about causation that current AI cannot autonomously execute end-to-end; only data analysis portions are automatable.
Adoption barriers	claude-haiku-4-5-20251001	4/5	This task involves medical diagnosis and occupational disease determination, which falls under healthcare and workers' compensation regulations. Many jurisdictions require a licensed occupational health professional or physician to conduct or certify such investigations, creating strong legal barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Determinations of work-relatedness often carry legal, workers' compensation, and regulatory (OSHA) implications requiring qualified professional judgment and documentation, creating liability-driven barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	Conducting worker studies requires specialized occupational health technicians with specific training and certification. AI systems today cannot replace this expertise, and the cost of AI infrastructure plus human oversight would exceed the labor cost of a qualified technician.
Cost vs. human wage	claude-sonnet-5	2/5	Since AI cannot perform the core fieldwork and judgment, human labor remains necessary, so cost savings are limited to marginal efficiency gains in data processing rather than full task substitution.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can assist with data organization and literature review, no deployed product reliably performs independent worker illness investigations at scale. Occupational health diagnosis and causality assessment remain expert-dependent activities with material error costs.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product conducts full occupational epidemiological studies or causation determinations; this remains a human-led investigative process with AI only aiding sub-tasks.

Perform tests to identify any potential hazards related to recycled products used at green building sites.

CI 5–25 · exposure 13 · augmentation 50 · importance 2.8/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption is slow; health and safety protocols in construction and green building remain conservative, with strong reliance on certified human expertise and regulatory compliance. Firms adopt AI for reporting and data management, not for replacing hazard testing itself.
Sector adoption velocity	claude-sonnet-5	1/5	Construction and occupational safety fields are traditionally slow adopters of AI, especially for physical inspection and testing tasks.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist by analyzing test results, cross-referencing material databases, predicting hazard patterns from historical data, and auto-generating compliance reports. These supports raise technician productivity on the analytical portion while the technician remains responsible for field testing and certification.
Augmentation potential	claude-sonnet-5	3/5	AI can help analyze test data, flag anomalies, generate reports, and cross-reference hazard databases, improving efficiency of the surrounding workflow even though it can't perform the physical test.
Task automatability	claude-haiku-4-5-20251001	2/5	Testing for hazards in recycled products requires physical sampling, specialized equipment calibration, and context-dependent interpretation of results tied to site-specific conditions. Current AI cannot autonomously perform physical inspections or operate testing equipment, though it could assist in data analysis and reporting after tests are conducted.
Task automatability	claude-sonnet-5	1/5	This requires physical sampling, on-site inspection, and hands-on testing of materials for hazards (e.g., chemical residues, structural integrity) that current AI cannot physically perform.;the analysis of results may benefit from AI but the core testing act is manual.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Occupational health and safety testing is often regulated by OSHA, EPA, or industry standards requiring certified human technicians to conduct and certify tests. Many jurisdictions legally require a qualified human to sign off on hazard assessments, creating a hard barrier to full automation.
Adoption barriers	claude-sonnet-5	4/5	Safety testing often requires certified technicians, adherence to regulatory protocols (OSHA, environmental standards), and liability for hazard identification, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	A trained technician commands significant labor cost (typically $25-35/hour loaded), and AI cannot replace the hands-on testing component. AI assistance for data analysis and report generation offers minor cost savings, but the bulk of the task cost is human labor that cannot be eliminated.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical test itself, so cost comparison favors the human technician who must be on-site with equipment.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI system performs hazard testing on physical products autonomously. While AI can analyze test data and support interpretation, the core task—physical testing with specialized equipment—remains firmly human-dependent in production environments.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product performs physical hazard testing of recycled building materials; this remains a human field/lab activity.

Prepare documents to be used in legal proceedings, testifying in such proceedings when necessary.

CI 9–20 · exposure 17 · 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	1/5	Legal and regulatory sectors move slowly on AI integration for high-stakes litigation work. Adoption remains in the pilot and assistive-tool phase, with almost no displacement of human document preparation or testimony in actual court proceedings.
Sector adoption velocity	claude-sonnet-5	2/5	Legal and safety compliance functions are typically slower adopters of full AI automation due to liability concerns, though AI drafting tools are gaining some traction in adjacent legal support roles.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist by drafting document templates, organizing evidence summaries, and suggesting legal language that the technician reviews and refines. However, the core task—judgment about what evidence matters and credible courtroom testimony—remains fundamentally human, limiting augmentation upside.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist in organizing evidence, drafting reports, summarizing incident data, and preparing exhibits, significantly speeding up the preparation portion of the task even though testimony remains human-only.
Task automatability	claude-haiku-4-5-20251001	1/5	Legal document preparation and testimony require nuanced judgment about evidence relevance, strategic framing, and adherence to court rules that vary by jurisdiction. Current AI cannot reliably navigate the adversarial stakes or provide testimony that meets evidentiary standards and cross-examination pressure.
Task automatability	claude-sonnet-5	2/5	AI can help draft and organize documents, but preparing legally sound materials requires domain expertise, verification of facts specific to the case, and judgment that current systems cannot fully replace; testimony itself is entirely human and unautomatable.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Legal proceedings require a licensed or credentialed human to prepare documents and testify under oath. Courts expect accountability, cross-examination, and personal responsibility that cannot be delegated to an AI system. Regulatory and liability structures mandate human ownership of legal filings.
Adoption barriers	claude-sonnet-5	5/5	Testifying in legal proceedings legally requires a qualified human witness, and documents used in litigation typically require certification, expertise attestation, and accountability that only a licensed/qualified professional can provide.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI drafting assistance is cheap, but the human expert must substantially review, revise, and ultimately sign or testify under oath. The human cannot be eliminated, so total cost remains dominated by the technician's wage; AI provides only partial marginal savings.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cheaply assist with drafting and summarization, but the overall task still requires substantial expert human time for accuracy, legal defensibility, and in-person testimony, limiting overall cost savings.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can draft routine legal documents with prompting, no deployed product reliably prepares court-ready safety-related legal documents or performs testimony. AI tools exist for legal research assistance but lack the judgment, accountability, and courtroom presence needed for actual legal proceedings.
Technical feasibility today	claude-sonnet-5	2/5	Legal drafting assistants and document review tools exist and are used in litigation support, but they are not deployed as autonomous preparers of technical/legal documents for OSH proceedings specifically, and testifying is not something any product performs.

Inspect fire suppression systems or portable fire systems to ensure proper working order.

CI 5–18 · exposure 8 · augmentation 38 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Fire suppression is a heavily regulated, low-digitization domain with small firm prevalence and strong human-certification requirements. Adoption of AI for autonomous inspection is negligible; most innovation remains in data collection and reporting aids rather than full automation.
Sector adoption velocity	claude-sonnet-5	1/5	Occupational health/safety inspection work is physical, low-digitization field labor with minimal AI agent deployment in this specific task domain.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist by automating report generation, flagging anomalies in sensor data, scheduling inspections, and organizing compliance documentation. These aids boost technician productivity, but the core inspection judgment and hands-on work remain human-centered.
Augmentation potential	claude-sonnet-5	2/5	AI can assist with scheduling, checklist generation, record-keeping, or analyzing inspection logs and sensor data, but offers little help with the core physical inspection itself.
Task automatability	claude-haiku-4-5-20251001	2/5	Fire suppression system inspection requires physical access, hands-on testing of mechanical and electrical components, and real-time assessment of system pressure, flow rates, and activation mechanisms. While AI could assist with documentation or scheduling, current systems cannot reliably perform the complete physical inspection and functional testing end-to-end.
Task automatability	claude-sonnet-5	1/5	This requires physical presence to inspect equipment, check pressure gauges, test alarms, and visually verify hardware condition—no current AI system can perform hands-on physical inspection.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Strong regulatory and liability barriers exist: fire codes (e.g., NFPA 25) typically mandate inspection by certified technicians, and failures carry legal liability and safety consequences. Certification and sign-off requirements, plus customer trust in human expertise for life-safety systems, create substantial adoption friction.
Adoption barriers	claude-sonnet-5	4/5	Fire safety inspections are often subject to regulatory codes, certification requirements, and liability considerations that mandate qualified human inspectors and sign-off.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Fully autonomous inspection would require expensive robotics, specialized sensors, and integration overhead that currently exceeds the cost of a trained technician performing the work. Partial automation (data logging, report generation) may reduce costs by 20-30%, but does not yet achieve cost parity for the full task.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI substitute for physical inspection, so the AI cost is effectively infinite relative to a human technician's wage for this specific hands-on task.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI product reliably performs full fire suppression system inspections independently. Visual inspection via robotics exists in niche contexts, but functional testing of pressure systems, valve operation, and chemical status requires specialized equipment and human-interpreted sensory feedback not yet automated at production scale in this domain.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs physical fire suppression system inspections autonomously; this remains a manual, in-person technician task.

Confer with schools, state authorities, or community groups to develop health standards or programs.

CI 0–7 · exposure 0 · augmentation 38 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Adoption of AI for stakeholder engagement and policy development in health and safety is minimal; these sectors remain largely human-driven and relationship-dependent, with no evidence of production-scale AI agents replacing professional conferencing roles.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety functions, especially those involving community and government liaison, show slow AI adoption due to their relational and compliance-driven nature.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI might assist by drafting background materials, summarizing regulations, or preparing data for meetings, but the core task—conferring, negotiating, and consensus-building—remains fundamentally human-led with only marginal AI assistance value.
Augmentation potential	claude-sonnet-5	3/5	AI can help draft talking points, summarize regulations, prepare meeting materials, or synthesize community feedback, meaningfully aiding preparation even though the core conferring activity remains human-led.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires multi-stakeholder negotiation, relationship-building, and collaborative development of nuanced policy—activities that demand human judgment, trust, and contextual understanding. Current AI cannot reliably lead or replace the social and political coordination this task demands.
Task automatability	claude-sonnet-5	1/5	This is a relationship-based negotiation and consensus-building task requiring in-person or synchronous stakeholder engagement, judgment, and trust-building that AI cannot perform end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Significant legal and organizational barriers protect this task: state authorities require licensed professionals; standards development often requires regulatory sign-off and human accountability; schools demand direct human contact and representation from credentialed safety professionals.
Adoption barriers	claude-sonnet-5	4/5	Representing an organization to external authorities and community groups typically requires accountable human representation, institutional trust, and often regulatory or contractual authority to speak on behalf of the agency.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	AI has no cost advantage here—a human occupational health professional conducting stakeholder engagement is essential, and AI cannot substitute for their expertise, credibility, or legal standing in policy development.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI substitute for the interpersonal negotiation and representation involved, so cost comparison favors the human entirely.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed product can independently confer with schools, state authorities, or community groups to develop standards or programs; this requires human authority, accountability, and the ability to represent an organization in real negotiations.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product conducts multi-party stakeholder conferences with schools or government bodies to co-develop health standards; this remains firmly in the human domain.

Evaluate situations or make determinations when a worker has refused to work on the grounds that danger or potential harm exists.

CI 0–6 · exposure 0 · augmentation 38 · 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	Safety-critical determinations in regulated industries are among the slowest-adopting domains for full automation. Labor unions, regulators, and liability concerns mean organizations retain human technicians for this task; no evidence of material AI adoption in production systems.
Sector adoption velocity	claude-sonnet-5	2/5	Occupational health and safety functions in industrial/field settings adopt AI slowly, mostly for documentation and monitoring rather than judgment-based determinations.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI could assist by gathering hazard data, organizing documentation, or suggesting relevant regulations, but the core determination—whether a worker's safety concern is justified and what action to take—remains a high-stakes human judgment where augmentation value is limited without the human retaining full decision authority.
Augmentation potential	claude-sonnet-5	3/5	AI can help technicians access safety regulations, past incident data, and hazard checklists to inform their determination, but the core evaluative judgment remains human-driven.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires nuanced judgment about worker safety claims, assessment of legitimate vs. illegitimate refusal grounds, and potentially adversarial decision-making with legal implications. Current AI systems cannot reliably evaluate the complex contextual, safety, and legal factors needed to make determinations that affect worker protections and employer liability.
Task automatability	claude-sonnet-5	1/5	This requires in-person judgment, physical inspection of hazardous conditions, and authoritative determination with legal consequences; current AI cannot perform on-site evaluation or make binding safety determinations.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Strong legal and regulatory barriers exist: OSHA and labor law frameworks mandate that trained human safety professionals assess work refusals; liability for incorrect determinations (exposing workers to harm or wrongly penalizing them) creates hard organizational friction; and the determination often requires sign-off by a licensed or certified occupational safety professional.
Adoption barriers	claude-sonnet-5	5/5	Occupational safety law typically requires a qualified/authorized person to investigate and adjudicate work-refusal claims, with significant liability exposure, making this a hard human-authority requirement.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Even if automation were feasible, the cost of AI infrastructure, integration, compliance oversight, and human review would likely approach or exceed the cost of a technician performing the task, given the low-volume, context-specific nature of refusal determinations.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical inspection and authoritative sign-off required, so there is no viable AI cost comparison for full task completion.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI product reliably performs end-to-end safety determinations on work refusals in production settings. The task requires real-time site assessment, worker interviews, hazard analysis, and legal judgment that exceed current capability and would expose organizations to liability if delegated to AI alone.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product independently evaluates work-refusal safety disputes or makes such determinations; this remains a human judgment call requiring site presence and authority.

Help direct rescue or firefighting operations in the event of a fire or an explosion.

CI 0–0 · exposure 0 · augmentation 38 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Emergency services and rescue operations are laggard sectors with respect to autonomous AI deployment due to the life-safety stakes, regulatory requirements, and deeply entrenched command structures. Adoption of AI for core decision-making in active operations remains minimal.
Sector adoption velocity	claude-sonnet-5	1/5	Emergency response and physical safety operations are among the least digitized, slowest-adopting domains for AI due to their physical, high-stakes, real-time nature.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI could assist with supporting tasks like resource tracking, communication routing, or hazard detection alerts in real time, but the core directing function—commanding responders in immediate danger—must remain under human control; augmentation is marginal here.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with situational awareness (e.g., sensor data, predictive fire behavior modeling, communication routing) to support human decision-makers, though the core directing function remains human-led.
Task automatability	claude-haiku-4-5-20251001	1/5	Directing rescue or firefighting operations requires real-time situational assessment, dynamic decision-making under uncertainty, coordinating multiple human responders, and adapting to rapidly changing conditions. Current AI systems cannot autonomously command on-scene emergency operations at the speed and judgment required.
Task automatability	claude-sonnet-5	1/5	Directing emergency rescue/firefighting operations requires real-time physical presence, situational judgment, and human coordination in hazardous, unpredictable environments; no AI system can perform this end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Directing rescue and firefighting operations is strictly a licensed, human-directed function governed by emergency services law, OSHA regulations, and chain-of-command protocols. Legal liability, worker safety statutes, and organizational structure make autonomous AI substitution infeasible.
Adoption barriers	claude-sonnet-5	5/5	Incident command in fire/explosion emergencies is legally and organizationally required to be led by qualified, certified personnel due to life-safety and liability concerns, creating hard regulatory and professional barriers.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The cost of a trained occupational health and safety technician directing operations is far lower than developing, deploying, and maintaining AI systems capable of real-time emergency command, plus the liability and redundancy overhead required for life-safety systems.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI substitute performing this function, so cost comparison favors humans entirely; any AI system used would be a costly ancillary tool, not a replacement.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI system reliably directs rescue or firefighting operations end-to-end in production. While AI can assist with routing or resource optimization in controlled contexts, autonomous command of emergency responders in active fire/explosion scenarios does not exist in operational systems.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product directs live firefighting or rescue operations; at most AI provides sensor data or dispatch support, but human incident commanders remain essential in practice.

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.