Non-Destructive Testing Specialists

17-3029.01
Median wage $78,350/yr59,930 employed (US)Rank #536 of 923 scored · top 58% by substitution

Test the safety of structures, vehicles, or vessels using x-ray, ultrasound, fiber optic or related equipment.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution26
Exposure27
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

16 rated tasks, binned by substitution score.

Position among all scored occupations

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

Tasks with substitution ≥ 70

0%

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

Why this score

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

Task automatabilityw 35%28

panel mean rating 2.1/5 → substitution pressure 28/100

Technical feasibility todayw 20%26

panel mean rating 2.0/5 → substitution pressure 26/100

Cost vs. human wagew 15%25

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

Adoption barriersw 20%inverted — strong barriers lower the score23

panel mean rating 4.1/5 (barrier strength) → substitution pressure 23/100

Sector adoption velocityw 10%23

panel mean rating 1.9/5 → substitution pressure 23/100

Task breakdown (16 tasks)

Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.

Prepare reports on non-destructive testing results.

51

CI 43–60 · exposure 58 · augmentation 75 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5NDT is concentrated in highly regulated sectors (aerospace, oil & gas, manufacturing) with slow digitization and strong preference for human expertise and liability protection; adoption of AI-driven reporting remains limited to pilot projects.
Sector adoption velocityclaude-sonnet-52/5NDT is a niche, largely industrial/manufacturing field with slower digital transformation and AI adoption compared to information-sector occupations, though software-assisted reporting is growing in some sub-sectors.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can significantly assist specialists by auto-generating initial report drafts, formatting data, flagging anomalies, and suggesting interpretations, allowing humans to focus on complex analysis and sign-off rather than transcription and routine documentation.
Augmentation potentialclaude-sonnet-54/5AI can effectively assist by auto-populating report templates, flagging anomalies from sensor data, and standardizing language, significantly speeding up the reporting process while the specialist verifies findings.
Task automatabilityclaude-haiku-4-5-202510013/5AI can draft reports from test data, extract findings, and format results with significant time savings, but typically requires human review to validate technical interpretation, ensure regulatory compliance, and sign off on conclusions that may have safety implications.
Task automatabilityclaude-sonnet-54/5Report writing from structured test data (measurements, pass/fail criteria, images) is largely templated and can be generated by AI given standardized inputs, though final review by a certified inspector remains typical.
Adoption barriersclaude-haiku-4-5-202510014/5NDT reports often require professional certification, legal sign-off by licensed specialists, and liability considerations in safety-critical contexts (aerospace, pressure vessels, structural); regulatory frameworks typically mandate human responsibility for report accuracy.
Adoption barriersclaude-sonnet-53/5NDT reports often require certification (e.g., ASNT Level II/III sign-off) for compliance in aerospace, pipeline, and structural industries, creating liability and regulatory friction even though the drafting itself isn't restricted.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI-assisted report generation has comparable all-in cost to human labor when accounting for integration, data preparation, and required human review cycles, though cost advantage grows if validation requirements are minimal.
Cost vs. human wageclaude-sonnet-54/5Automated report generation from digital inspection data is far cheaper than having a skilled technician manually compile reports, though some oversight cost remains for accuracy and certification sign-off.
Technical feasibility todayclaude-haiku-4-5-202510013/5Products exist for automated report generation from sensor data and structured testing output, but they operate in constrained domains and require domain expertise oversight; no mature end-to-end solution reliably produces certifiable NDT reports without human review.
Technical feasibility todayclaude-sonnet-53/5NDT software suites and general-purpose AI drafting tools can generate reports from data logs today, but integration with specific equipment outputs and industry-specific formatting varies, so reliable production use is uneven across the field.

Interpret or evaluate test results in accordance with applicable codes, standards, specifications, or procedures.

44

CI 20–67 · exposure 50 · augmentation 75 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Adoption is underway in aerospace, automotive, and heavy industry but remains inconsistent; many inspection services still employ primarily manual interpretation. Regulatory constraints and certification requirements slow wider deployment compared to less-regulated information tasks.
Sector adoption velocityclaude-sonnet-52/5Manufacturing and inspection sectors are relatively slow to adopt AI-driven interpretation compared to information/finance industries, with pilots more common than production use.
Augmentation potentialclaude-haiku-4-5-202510015/5AI substantially assists specialists by automating data preprocessing, highlighting suspicious regions, suggesting preliminary interpretations, and generating standardized reports, allowing the expert to focus on judgment and sign-off. This is a prime example of assistive deployment.
Augmentation potentialclaude-sonnet-53/5AI-based image enhancement and anomaly detection can meaningfully assist inspectors in spotting potential flaws faster, though the human remains responsible for final interpretation and certification.
Task automatabilityclaude-haiku-4-5-202510014/5AI can reliably interpret standardized test results (ultrasonic, radiographic, eddy current data) against known codes and specifications with high accuracy, achieving significant time savings. However, rare edge cases, ambiguous results, or novel defect patterns may still require human expertise, preventing a perfect 5.
Task automatabilityclaude-sonnet-52/5Interpreting NDT results (radiographs, ultrasonic traces, etc.) against codes requires trained judgment on ambiguous defect signatures, and while AI can flag anomalies, full independent interpretation against standards is not yet reliably automatable end-to-end.
Adoption barriersclaude-haiku-4-5-202510013/5Significant barriers exist: NDT results often require certified specialists to sign off for regulatory/safety reasons (aerospace, nuclear, pressure equipment directives), and liability concerns mean organizations retain human review. This prevents full automation despite technical feasibility.
Adoption barriersclaude-sonnet-55/5NDT interpretation for safety-critical components (aerospace, pipelines, pressure vessels) requires certified personnel (e.g., ASNT Level II/III) to sign off per code, making this a hard licensing/liability barrier.
Cost vs. human wageclaude-haiku-4-5-202510014/5Automated interpretation of NDT results (image analysis, standard-matching, report generation) costs far less per evaluation than a human specialist's loaded wage, particularly at scale. The AI cost (inference + model maintenance) is typically 1-5% of specialist labor cost.
Cost vs. human wageclaude-sonnet-52/5AI image analysis tools have real inference and integration costs plus mandatory human oversight, so total cost is not dramatically below a certified technician's wage for this specific judgment task.
Technical feasibility todayclaude-haiku-4-5-202510014/5Machine learning and rule-based systems are deployed in production for NDT data interpretation (defect detection in ultrasonic/radiographic imagery, automated report generation). Products exist with demonstrated reliability on common scenarios, though some rely on human verification for complex or marginal cases.
Technical feasibility todayclaude-sonnet-52/5Some AI-assisted defect detection tools exist in radiography/ultrasonic image analysis, but they are narrow, require human sign-off, and are not broadly deployed as standalone interpreters across NDT modalities.

Document non-destructive testing methods, processes, or results.

34

CI 25–43 · exposure 38 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5NDT is performed in capital-intensive, regulated sectors (aerospace, oil/gas, manufacturing) that adopt new tools slowly; while some plants use basic automation aids, production-scale AI documentation is rare and pilots are limited to large OEMs.
Sector adoption velocityclaude-sonnet-52/5NDT is a specialized industrial/inspection field with lower digitization and slower AI tool adoption compared to information-sector occupations, with pilots more common than production deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by auto-detecting defects in images, generating draft reports, and flagging anomalies, which helps specialists focus on interpretation and quality control. However, augmentation is partial because the core judgment—what constitutes a reportable flaw and how to document it—remains largely human-driven.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully speed up report drafting, summarization of findings, and formatting compliance documentation, letting the specialist focus on interpretation and sign-off while staying in the loop.
Task automatabilityclaude-haiku-4-5-202510012/5Documenting test results (parameters, findings, images) can be partially automated through data extraction and standardized report generation, but the interpretation of complex ultrasonic, radiographic, or eddy-current results and determining what defects mean requires specialist judgment. AI can capture structured data but cannot reliably replace the human decision on what findings warrant documentation or how to frame them.
Task automatabilityclaude-sonnet-53/5AI can draft structured reports and populate templates from test data and dictated notes, but capturing accurate readings, images, and technical judgments still requires human input and verification, so only part of the documentation workflow can be automated end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5NDT documentation is often governed by industry standards (ASME, AWS, ISO) and may require certified personnel sign-off; liability for missed defects creates high error-cost asymmetry; and regulatory bodies in aerospace, pressure vessels, and pipelines mandate human certification over autonomous systems.
Adoption barriersclaude-sonnet-54/5NDT documentation often must meet regulatory and industry-code requirements (e.g., ASNT, ASME) with certified personnel signing off on results, creating strong liability and certification barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for documentation (vision systems, template-based report generation) reduce some manual work but require integration, training data, and human review. The all-in cost remains comparable to or slightly cheaper than human documentation, not an order of magnitude lower.
Cost vs. human wageclaude-sonnet-53/5AI-assisted drafting and transcription could cut some documentation time cheaply, but integration with specialized NDT data formats, imaging, and compliance standards adds cost, making overall savings moderate rather than dramatic.
Technical feasibility todayclaude-haiku-4-5-202510013/5Products exist for automated report generation and image annotation in NDT (e.g., defect detection in radiographs via computer vision), but they operate within narrow scopes and require human verification; no mature end-to-end system documents all NDT methods reliably without significant oversight.
Technical feasibility todayclaude-sonnet-52/5Generic document generation and transcription tools exist, but there are no widely deployed NDT-specific documentation products handling this reliably in production at scale across the industry.

Make radiographic images to detect flaws in objects while leaving objects intact.

28

CI 25–30 · exposure 30 · augmentation 50 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption is slow and remains concentrated in large aerospace/automotive manufacturing; most facilities still rely on traditional human-led radiography with limited AI integration, reflecting sector conservatism and regulatory friction.
Sector adoption velocityclaude-sonnet-52/5Manufacturing and industrial inspection sectors adopt AI more slowly than digital/professional services; automation here is mostly limited to pilot programs for image analysis, not widespread deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by flagging potential defects in batches of radiographic images and reducing manual review time, but the human specialist remains essential for interpretation, decision-making on image quality, and regulatory sign-off.
Augmentation potentialclaude-sonnet-53/5AI-based image analysis tools can help flag potential flaws and speed up interpretation, providing moderate productivity gains for technicians who still perform and verify the physical testing.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can analyze radiographic images for flaw detection with reasonable accuracy, the full task requires physical positioning of equipment, safety decisions, and real-time judgment about image quality and repeat shots—operations that demand in-person presence and cannot achieve 50% time savings end-to-end today.
Task automatabilityclaude-sonnet-52/5The physical act of positioning equipment, handling radiographic sources, and imaging real objects requires manual/physical operation that current AI cannot perform end-to-end; AI can assist in defect analysis but not the physical imaging process itself.
Adoption barriersclaude-haiku-4-5-202510014/5Regulatory frameworks (ASME, ISO standards) typically require certified human NDT technicians to perform or validate radiographic testing and sign off on results; liability and safety certification create hard legal barriers to full automation.
Adoption barriersclaude-sonnet-54/5NDT radiography is often regulated (e.g., ASNT certification, radiation safety rules) requiring certified human technicians to perform and sign off on inspections, creating strong regulatory and liability barriers.
Cost vs. human wageclaude-haiku-4-5-202510012/5The labor cost of qualified NDT specialists (training, certification, hourly rates) remains high, while AI analysis systems still require integration, oversight, and the full equipment/technician cost for image acquisition, making the all-in cost comparable to or higher than human-only workflow today.
Cost vs. human wageclaude-sonnet-52/5Radiographic equipment, robotic positioning, and certified oversight remain costly; AI-assisted analysis reduces some labor but the overall imaging process still requires expensive hardware and human certification, keeping costs comparable to human-only workflows.
Technical feasibility todayclaude-haiku-4-5-202510013/5AI-based image analysis systems exist for radiographic defect detection (e.g., in manufacturing QA), but deployment remains narrow, often requiring human oversight and producing material false-positive/negative rates; the image capture itself remains manual and context-dependent.
Technical feasibility todayclaude-sonnet-52/5Automated defect-detection software exists and is used to analyze radiographs, but fully autonomous systems that both capture and interpret radiographic images in production settings are narrow and not widespread.

Visually examine materials, structures, or components for signs of corrosion, metal fatigue, cracks, or other flaws, using tools and equipment such as endoscopes, closed-circuit television systems, and fiber optics.

28

CI 25–30 · exposure 30 · augmentation 50 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5NDT occurs across highly regulated sectors (aerospace, nuclear, pipelines) where adoption is cautious and often pilot-stage; while manufacturing and infrastructure digitization is advancing, deep AI automation in NDT remains slow due to liability and certification requirements.
Sector adoption velocityclaude-sonnet-52/5Industrial inspection sectors (oil/gas, aerospace, manufacturing) are physically-oriented and have historically slower digitization and AI adoption compared to information-based industries, though pilots of AI-assisted defect detection are emerging.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-assisted image analysis can help specialists prioritize areas to examine closely and flag potential defects for human review, moderately improving inspection throughput; however, the specialist must validate every AI-flagged anomaly, limiting overall productivity gains.
Augmentation potentialclaude-sonnet-53/5AI-based image analysis tools can flag potential flaws in captured footage, helping inspectors prioritize areas of concern and improve consistency, though the human must still operate equipment and validate findings.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can analyze visual data from endoscopes and CCTV systems to detect obvious defects, the task requires expert judgment to distinguish critical flaws from benign surface variations, assess severity, and make qualification decisions that demand domain knowledge and contextual reasoning AI cannot reliably replicate at production quality.
Task automatabilityclaude-sonnet-52/5Physical inspection with hands-on equipment positioning (endoscopes, borescopes, fiber optics) requires manual manipulation in complex physical environments that current AI cannot autonomously perform end-to-end, though image analysis of captured footage can be partially automated.
Adoption barriersclaude-haiku-4-5-202510014/5NDT work is heavily regulated by industry standards (ASME, ISO, AWS) and often requires human inspectors to be certified and personally liable for inspection results; many contracts and safety-critical applications legally mandate human sign-off, creating hard adoption friction.
Adoption barriersclaude-sonnet-54/5NDT often requires certified technicians (e.g., ASNT certification) for regulatory/safety-critical applications like aerospace and pipelines, creating liability and licensing barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI-powered visual inspection systems require significant capital investment in hardware (specialized cameras, endoscope interfaces) and software licensing, plus ongoing human oversight to validate findings; total cost per inspection often remains comparable to or exceeds direct human labor.
Cost vs. human wageclaude-sonnet-52/5Physical equipment deployment, sensor positioning, and certified oversight remain costly; AI reduces analysis time on captured images but does not eliminate the human labor and equipment costs central to the task.
Technical feasibility todayclaude-haiku-4-5-202510013/5Computer vision systems exist for defect detection in industrial settings, but current products show material false-positive and false-negative rates on complex materials and fatigue patterns; they function best as preliminary screening tools rather than autonomous end-to-end inspectors.
Technical feasibility todayclaude-sonnet-52/5Some AI-assisted defect detection products exist (automated crack/corrosion detection in imagery) but they are narrow, require human verification, and are not yet reliable standalone replacements for certified inspectors in production NDT workflows.

Interpret the results of all methods of non-destructive testing (NDT), such as acoustic emission, electromagnetic, leak, liquid penetrant, magnetic particle, neutron radiographic, radiographic, thermal or infrared, ultrasonic, vibration analysis, and visual testing.

25

CI 25–25 · exposure 25 · augmentation 63 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption is slow and cautious; pilots of AI-assisted analysis exist in large aerospace and energy firms, but production deployment remains limited because of liability concerns, inspector certification requirements, and conservative industry risk tolerance.
Sector adoption velocityclaude-sonnet-52/5Manufacturing, oil & gas, and industrial inspection sectors are traditionally slow adopters of AI, with automation limited to pilot programs and assistive tools rather than widespread production deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI tools can assist by highlighting anomalies in radiographic or ultrasonic data, flagging regions of concern, and reducing signal noise, raising specialist productivity on data review; however, augmentation is partial and depends on which NDT method is in use.
Augmentation potentialclaude-sonnet-54/5AI-based image analysis and anomaly detection tools meaningfully speed up flaw identification and triage across several NDT methods, letting human specialists focus on borderline cases and final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist in analyzing some NDT data streams (radiographic images, ultrasonic signals, thermal imagery), interpretation requires contextual judgment about material properties, safety criticality, and decision-making under uncertainty that remains largely human-dependent. End-to-end automation with ≥50% time savings and equal quality is not demonstrated across the full range of NDT methods.
Task automatabilityclaude-sonnet-52/5Interpreting diverse NDT modalities requires integrating physical sensor data, material science knowledge, and contextual judgment about defect significance; while AI can assist with pattern detection in specific modalities (e.g., ultrasonic or radiographic images), full cross-method interpretation with safety-critical sign-off is not yet automatable end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5NDT interpretation is often embedded in regulated inspection workflows (aerospace, nuclear, pressure vessels) where liability, certification of the inspector, and regulatory sign-off by a qualified human are hard requirements; automation of the final interpretation decision faces legal and compliance barriers.
Adoption barriersclaude-sonnet-54/5NDT results often support safety-critical decisions (aerospace, pipelines, pressure vessels) requiring certified NDT Level II/III inspectors to sign off per industry standards (ASNT, ASME), creating strong regulatory and liability barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI systems (image analysis, signal processing) still require significant integration, validation, and expert human review, keeping total cost per interpretation close to or exceeding the cost of a specialist performing the task directly.
Cost vs. human wageclaude-sonnet-52/5Specialized AI vision models for defect detection can be cheaper per inspection than a certified human, but integration, calibration, and mandatory human verification for safety-critical findings keep overall costs comparable to human labor in most deployments.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed products exist for narrow subsets (e.g., radiographic defect detection), but they require expert oversight, struggle with edge cases, and have not achieved reliable independent deployment at production scale across the diversity of NDT methods and contexts described.
Technical feasibility todayclaude-sonnet-52/5Deployed products exist for narrow sub-tasks like automated defect detection in radiographic or ultrasonic images, but no product reliably interprets the full range of NDT methods listed with production-grade accuracy across industries.

Select, calibrate, or operate equipment used in the non-destructive testing of products or materials.

25

CI 25–25 · exposure 25 · augmentation 50 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of full automation in NDT remains limited; while larger aerospace and energy firms are piloting AI-assisted inspection analysis, the sector remains conservative, with most organizations still relying on certified human technicians to operate equipment and validate findings.
Sector adoption velocityclaude-sonnet-52/5Manufacturing and industrial inspection sectors adopt AI slowly due to physical infrastructure, safety certification requirements, and the specialized nature of NDT equipment, with pilots more common than production deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI shows useful potential in assisting technicians by automating defect detection in stored imagery, recommending inspection procedures, and flagging anomalies for review, though the human technician remains responsible for equipment operation and final interpretation.
Augmentation potentialclaude-sonnet-53/5AI-assisted defect recognition and signal analysis software can help technicians interpret NDT data faster, though core equipment selection, calibration, and physical operation remain manual tasks.
Task automatabilityclaude-haiku-4-5-202510012/5While AI could assist with calibration scheduling and procedure selection, the physical operation of NDT equipment (ultrasonic probes, radiography apparatus, eddy current tools) and real-time interpretation of sensor outputs requires hands-on manipulation and contextual judgment that current AI systems cannot reliably perform end-to-end without human oversight.
Task automatabilityclaude-sonnet-52/5This task requires physical handling and calibration of NDT equipment (ultrasonic, radiographic, magnetic particle devices) directly on materials, which current AI cannot perform end-to-end; only data analysis portions are automatable.4o rather than physical setup remains manual.4o
Adoption barriersclaude-haiku-4-5-202510014/5NDT work is heavily regulated by industry standards (ASTM, API, ASME) and often requires certified technicians; liability and safety considerations mean that regulatory bodies and clients mandate human certification for equipment operation and results sign-off.
Adoption barriersclaude-sonnet-54/5NDT technicians often require certification (e.g., ASNT levels) and liability for missed defects in safety-critical industries (aerospace, pipelines) creates strong oversight and sign-off requirements.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI solutions for NDT support (analysis tools, defect detection) represent significant upfront licensing costs and still require human expertise for operation and validation, making them roughly comparable to or more expensive than the technician's labor in many small-to-medium operations.
Cost vs. human wageclaude-sonnet-52/5Physical equipment operation still requires skilled technicians on-site; AI cannot replace the hands-on calibration and setup, so cost savings are limited to partial data-analysis augmentation rather than full task substitution.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some AI applications exist for post-inspection image analysis and anomaly detection, but no deployed product reliably handles equipment selection, calibration, and operation autonomously; most production systems still require skilled technicians to conduct the testing procedure itself.
Technical feasibility todayclaude-sonnet-52/5Deployed products can assist with defect detection in captured NDT images/signals, but selecting and calibrating physical equipment on-site is not something any commercial product performs autonomously today.

Map the presence of imperfections within objects, using sonic measurements.

25

CI 25–25 · exposure 25 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI-assisted tools is emerging in high-volume manufacturing and aerospace but remains slow in field NDT services and smaller operations. Pilots are common in large enterprises, but production-scale autonomous deployment without human review is rare due to regulatory and liability constraints.
Sector adoption velocityclaude-sonnet-52/5Manufacturing and inspection sectors are traditionally slower adopters of AI compared to information/finance sectors, with pilots emerging but production-scale autonomous NDT still limited.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by automating signal preprocessing, highlighting anomaly regions, and suggesting defect classifications, allowing the specialist to focus on validation and judgment. This raises inspection throughput and consistency but keeps the human in the loop for critical safety decisions.
Augmentation potentialclaude-sonnet-53/5AI-based signal processing and pattern recognition tools can help technicians flag anomalies faster and improve consistency, providing useful augmentation during analysis of sonic data.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can process ultrasonic or sonic data and identify signal anomalies computationally, the task requires skilled interpretation of complex waveforms in varied materials and geometries, selection of appropriate sonic frequency/method for specific objects, and judgment calls on defect severity and safety significance. Current AI can assist with pattern detection but cannot reliably replace the specialist's contextual knowledge and real-time decision-making.
Task automatabilityclaude-sonnet-52/5Interpreting sonic/ultrasonic flaw data involves physical probe manipulation and calibrated judgment that current AI cannot fully replace end-to-end; some signal analysis can be automated but not the full task with equal quality.
Adoption barriersclaude-haiku-4-5-202510014/5NDT is heavily regulated in aerospace, pressure vessel, and safety-critical industries; certifications (ASNT, ISO) mandate that a qualified human specialist perform and sign off on inspections. Legal liability for missed defects creates strong organizational and regulatory friction against full automation without human approval.
Adoption barriersclaude-sonnet-54/5Many NDT applications (aerospace, pressure vessels, pipelines) require certified inspectors (e.g., ASNT certification) and regulatory sign-off, creating strong legal/liability barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI-assisted systems require expensive sensor hardware, specialized software licenses, and mandatory human technician oversight for verification and sign-off. The total cost per inspection remains comparable to or higher than direct specialist labor, especially when factoring in liability and re-inspection rates.
Cost vs. human wageclaude-sonnet-52/5Specialized ultrasonic equipment, calibration, and certified oversight remain costly; AI software adds analysis value but doesn't eliminate need for skilled technicians and equipment costs, so savings are modest.
Technical feasibility todayclaude-haiku-4-5-202510012/5Automated defect-detection systems exist in controlled lab and manufacturing contexts, but real-world NDT involves variable materials, access constraints, and liability-critical judgments. Deployed AI tools are narrow (e.g., trained on specific material types) and typically require specialist review; no general-purpose product reliably performs this end-to-end without human expertise.
Technical feasibility todayclaude-sonnet-52/5AI-assisted defect detection algorithms exist in NDT software (e.g., ultrasonic flaw classification), but reliable autonomous end-to-end mapping in production across diverse materials/geometries is not yet standard.

Identify defects in concrete or other building materials, using thermal or infrared testing.

25

CI 25–25 · exposure 25 · augmentation 63 · importance 3.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Construction and infrastructure sectors show slower AI adoption than information or finance; while thermal imaging is established, autonomous defect identification via AI remains in pilot and early deployment phases with limited production displacement.
Sector adoption velocityclaude-sonnet-52/5Construction and structural inspection sectors are traditionally slow adopters of AI, with physical, safety-critical inspection tasks lagging behind sectors like finance or IT.
Augmentation potentialclaude-haiku-4-5-202510014/5AI-assisted thermal image analysis—flagging suspicious regions, suggesting defect types, and enhancing image clarity—can significantly boost a specialist's throughput and consistency while they retain final judgment and sign-off authority.
Augmentation potentialclaude-sonnet-53/5AI-based image processing and pattern recognition can help flag potential defects in thermal images, improving technician efficiency, though human judgment remains essential for confirmation and certification.
Task automatabilityclaude-haiku-4-5-202510012/5While thermal and infrared imaging can be automated to capture and process images, the task requires skilled interpretation of subtle thermal signatures in context of material properties, environmental conditions, and defect severity—judgments that current AI systems cannot reliably make end-to-end without significant human oversight and domain expertise.
Task automatabilityclaude-sonnet-52/5Physical probe/sensor operation and equipment positioning are still manual; AI can assist with image interpretation but cannot end-to-end perform the physical inspection and data capture.
Adoption barriersclaude-haiku-4-5-202510014/5NDT specialists often require certification (ASNT, ACCP) or regulatory authorization under building codes and standards (ASTM, ISO); liability and safety-critical consequences of missed defects create strong legal and professional barriers to full automation without licensed sign-off.
Adoption barriersclaude-sonnet-54/5NDT often requires certified technicians (e.g., ASNT certification) and sign-off for structural safety assessments, creating strong regulatory and liability barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5The hardware (thermal cameras) and AI software costs, combined with required human expert review and calibration, remain comparable to or exceed the loaded wage of a skilled NDT specialist for typical inspection volumes.
Cost vs. human wageclaude-sonnet-52/5Specialized thermal cameras, on-site labor, and calibration remain costly; AI reduces some analysis time but does not eliminate need for skilled technicians and equipment costs, keeping overall cost comparable to human-only workflows.
Technical feasibility todayclaude-haiku-4-5-202510012/5Thermal imaging software exists for defect detection, but deployed systems remain narrow and require substantial human validation; error rates are material, especially for distinguishing genuine defects from thermal artifacts or benign variations, limiting production-scale autonomous deployment.
Technical feasibility todayclaude-sonnet-52/5AI-assisted thermal/infrared image analysis exists in research and niche products (e.g., defect detection in thermography), but deployed, reliable production systems performing full inspections are rare and mostly human-supervised.

Identify defects in solid materials, using ultrasonic testing techniques.

24

CI 23–25 · exposure 25 · augmentation 50 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5NDT specialists work in highly regulated, safety-critical sectors (aerospace, manufacturing, utilities) with strong quality and certification cultures; adoption of fully autonomous AI systems remains minimal; most use cases involve operator assistance rather than replacement in production.
Sector adoption velocityclaude-sonnet-52/5Manufacturing and inspection sectors are relatively slow adopters of AI-driven physical inspection tools compared to information-based industries.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by auto-classifying signal patterns, flagging anomalies for review, and reducing manual scrolling through data, raising technician throughput; however, the assistance is limited to signal interpretation—final judgment and certification remain human responsibilities.
Augmentation potentialclaude-sonnet-53/5AI-based signal analysis and pattern recognition can help technicians flag potential defects faster and reduce interpretation errors, though the human remains essential for probe handling and final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5Ultrasonic testing data interpretation requires pattern recognition that AI can assist with, but end-to-end automation faces challenges: automated scanning can capture ultrasonic signals, but judgment about defect severity, material context, and actionability often requires human expertise and physical intuition that current AI systems struggle to replicate reliably without significant human oversight.
Task automatabilityclaude-sonnet-52/5Ultrasonic defect identification requires physical probe manipulation, calibration, and interpretation of complex waveforms in varied material geometries, which current AI cannot perform end-to-end without robotic hardware and human setup.'
Adoption barriersclaude-haiku-4-5-202510014/5Strong regulatory and liability barriers exist: aerospace, pressure vessels, medical devices, and nuclear contexts require licensed or certified personnel to perform and certify defect assessments; legal responsibility for missed defects typically cannot be transferred to an automated system, creating hard adoption friction.
Adoption barriersclaude-sonnet-54/5NDT technicians typically require certification (e.g., ASNT levels) and sign-off responsibility for safety-critical inspections, creating strong regulatory and liability barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI inference for signal classification is cheap, but integration with ultrasonic hardware, calibration, validation, and mandatory human oversight add substantial costs; the total cost per task remains comparable to or above a skilled technician's loaded wage when accounting for liability and quality assurance.
Cost vs. human wageclaude-sonnet-52/5Equipment, sensor deployment, and certified human oversight keep costs comparable to or higher than human inspectors despite some software efficiency gains.
Technical feasibility todayclaude-haiku-4-5-202510012/5While some AI-based defect detection systems exist in research and limited production (e.g., signal classification models), they typically supplement rather than replace human operators, and deployed products show material false-positive/negative rates in complex real-world materials; no mature, production-scale system handles the full task autonomously.
Technical feasibility todayclaude-sonnet-52/5Some AI-assisted signal interpretation and defect classification tools exist in NDT software, but they are narrow-scope aids rather than autonomous inspection systems deployed at scale.

Examine structures or vehicles such as aircraft, trains, nuclear reactors, bridges, dams, and pipelines, using non-destructive testing techniques.

23

CI 20–25 · exposure 25 · augmentation 63 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI in NDT remains limited to assistive tools in high-value sectors (aerospace, nuclear) where pilots exist; most NDT work is performed by certified technicians using conventional methods. Full automation adoption is slow due to regulatory constraints, the cost of integrated robotic systems, and the safety-critical nature of the domain.
Sector adoption velocityclaude-sonnet-52/5This is a highly regulated, physical, safety-critical field with slow technology adoption cycles due to certification requirements and conservative industry practices in aerospace, energy, and civil infrastructure sectors.
Augmentation potentialclaude-haiku-4-5-202510013/5AI augmentation is already in use: image analysis tools help technicians identify and classify defects faster, flagging anomalies for review and reducing false negatives. This improves technician productivity on data interpretation but does not replace the physical inspection work or the need for human judgment on measurements and remediation.
Augmentation potentialclaude-sonnet-54/5AI-based image analysis and pattern recognition tools meaningfully assist inspectors by flagging potential defects and anomalies in scan data, improving speed and consistency while the certified human remains responsible for final assessment.
Task automatabilityclaude-haiku-4-5-202510012/5Non-destructive testing fundamentally requires physical inspection equipment (ultrasonic, radiographic, eddy current probes) positioned on real structures in situ. While AI can analyze resulting data (images, signals), the hands-on examination, sensor positioning, and decision to retest or adjust methodology based on physical feedback remain heavily manual, preventing 50% time savings end-to-end today.
Task automatabilityclaude-sonnet-52/5While AI-enabled sensors and image analysis can assist in interpreting NDT data (e.g., ultrasound, radiography), the physical inspection process—positioning equipment, accessing structures, and adapting to on-site conditions—remains largely manual and cannot yet be fully automated end-to-end.'
Adoption barriersclaude-haiku-4-5-202510014/5Regulatory frameworks (ASME, EN, FAA standards) explicitly require certified and licensed NDT technicians to perform and certify examinations on safety-critical structures; equipment use and findings often require professional stamp or sign-off. Liability and legal liability for missed defects creates strong barriers to unsupervised automation.
Adoption barriersclaude-sonnet-55/5NDT inspections of critical infrastructure like aircraft, nuclear reactors, and dams are subject to strict regulatory certification and licensing requirements mandating qualified human inspectors to sign off on safety-critical findings.
Cost vs. human wageclaude-haiku-4-5-202510012/5NDT equipment (ultrasonic devices, X-ray machines, automated scanning rigs) and the AI inference for defect detection remain capital-intensive and require integration with specialized hardware. Human technician labor per inspection is moderate, and the all-in cost of automation (hardware + software + calibration) is typically higher than hourly technician rates for routine inspections.
Cost vs. human wageclaude-sonnet-52/5AI image analysis software adds value but does not replace the cost of specialized equipment, field technicians, and certified inspectors, so overall cost savings are modest rather than order-of-magnitude.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI products can assist with post-hoc defect detection in NDT imagery (e.g., crack classification in radiographs), but no deployed system reliably performs the full task—equipment setup, positioning, measurement interpretation, and safety sign-off—without skilled technician oversight. Current applications are narrow aids rather than end-to-end solutions.
Technical feasibility todayclaude-sonnet-52/5Some deployed products use AI for defect detection in radiographic or ultrasonic images, but these are narrow-scope tools requiring human setup, calibration, and validation rather than autonomous field inspection systems.

Evaluate material properties, using radio astronomy, voltage and amperage measurement, or rheometric flow measurement.

21

CI 16–25 · exposure 20 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5The non-destructive testing sector consists largely of specialized, small-to-medium firms and in-field operations with modest digitization. Adoption of AI-assisted tools is emerging in data analysis but remains slow; production-scale replacement of core testing tasks is rare.
Sector adoption velocityclaude-sonnet-52/5Manufacturing and industrial inspection sectors have historically been slow to adopt AI for physical measurement tasks, with automation focused on data logging and analytics rather than full task replacement.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by automating data analysis, pattern recognition in measurement results, report generation, and decision support on material property trends. These augmentations improve productivity and reduce manual analysis time, though the human specialist remains responsible for measurement execution and judgment.
Augmentation potentialclaude-sonnet-53/5AI can assist by analyzing sensor data patterns, flagging anomalies, and supporting interpretation of rheometric or electrical measurements, improving technician efficiency without replacing hands-on measurement.
Task automatabilityclaude-haiku-4-5-202510012/5Current AI can assist with data interpretation from measurement instruments and generate preliminary reports, but cannot autonomously perform the hands-on measurement setup, calibration, equipment operation, or real-time decision-making required to evaluate material properties across the diverse methods (radio astronomy, electrical measurement, rheometric flow). The physical manipulation and contextual judgment remain human-dependent.
Task automatabilityclaude-sonnet-52/5This requires physical instrumentation, sensor operation, and interpretation of specialized readings from equipment that AI cannot physically operate or interface with end-to-end; only data analysis portions could plausibly be assisted. ,
Adoption barriersclaude-haiku-4-5-202510014/5Non-destructive testing is subject to industry standards (ASME, ISO) and often requires certified personnel to conduct and certify results, particularly in regulated industries (aerospace, pressure vessels, nuclear). Legal and liability requirements for sign-off on material integrity testing create significant regulatory barriers to full automation.
Adoption barriersclaude-sonnet-54/5Many NDT roles require certification (e.g., ASNT levels) and sign-off for safety-critical inspections in aerospace, energy, and manufacturing, creating strong regulatory and liability barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for data processing and interpretation are relatively low-cost, but the specialized measurement equipment, calibration, and human expertise required to execute this task remain substantial. The all-in cost of automating the full workflow does not yet undercut the loaded wage of a trained non-destructive testing specialist.
Cost vs. human wageclaude-sonnet-51/5AI cannot replace the physical measurement equipment and skilled technician labor required; cost savings from AI are minimal since the bottleneck is physical instrumentation and calibration, not data processing.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can analyze post-measurement data and support interpretation, no deployed product reliably performs end-to-end material property evaluation autonomously. Some software assists with data processing and standardized report generation, but the core measurement task—selecting appropriate methods, operating specialized equipment, and validating results—remains outside production-grade AI automation.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs autonomous physical NDT measurement using these specialized techniques; this remains a manual, instrument-driven task performed by trained technicians.

Develop or use new non-destructive testing methods, such as acoustic emission testing, leak testing, and thermal or infrared testing.

21

CI 16–25 · exposure 20 · augmentation 63 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5NDT remains embedded in capital-intensive, regulated sectors (oil & gas, aerospace, manufacturing) with slow digital transformation and strong preference for certified human specialists. Adoption of AI-assisted tools is emerging but remains pilot-stage; production replacement is rare.
Sector adoption velocityclaude-sonnet-52/5NDT occurs in manufacturing, energy, and industrial sectors with historically slower digitization and AI adoption compared to information/professional services sectors.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist NDT specialists by automating data analysis, highlighting anomalies in thermal or acoustic data, and accelerating interpretation of large datasets, allowing specialists to focus on judgment and risk assessment. This augmentation is already demonstrable in tools that flag defects in imagery or signal processing.
Augmentation potentialclaude-sonnet-53/5AI and machine learning can assist in analyzing sensor data patterns (e.g., acoustic emission signals, thermal imaging) to help specialists identify anomalies faster, though the core method development and physical testing remain human-led.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist in analyzing acoustic emission and thermal imaging data post-collection, the physical setup, calibration, and real-time judgment required to deploy NDT equipment and interpret results in context remain heavily manual. AI cannot yet autonomously position sensors, adapt protocols to site conditions, or make safety-critical decisions at the point of testing without substantial human oversight.
Task automatabilityclaude-sonnet-52/5Developing new NDT methods and applying them to physical equipment requires hands-on experimentation, sensor deployment, and physical judgment that current AI cannot perform end-to-end; AI can assist with data analysis but not the physical testing or method development itself.
Adoption barriersclaude-haiku-4-5-202510014/5NDT specialists operate under significant regulatory and liability constraints: codes like ASME, API, and ISO require qualified, certified personnel to perform and sign off on testing. Liability for missed defects or misinterpretation creates strong error-cost asymmetry, and customer/regulatory requirements for a licensed human signature present hard adoption barriers.
Adoption barriersclaude-sonnet-54/5NDT often requires certified specialists (e.g., ASNT certification) for safety-critical inspections in aerospace, energy, and infrastructure, creating strong licensing and liability barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of deploying AI for NDT analysis (software, model training, integration, validation infrastructure) combined with the high labor cost of specialized NDT engineers creates a cost ratio roughly comparable to or exceeding the human specialist's loaded wage, particularly given the need for continuous oversight.
Cost vs. human wageclaude-sonnet-51/5AI cannot replace the physical equipment, sensors, and human expertise required, so there is no meaningful AI cost substitute for this task at present.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI products exist for analyzing NDT data (e.g., thermal image classification, acoustic signal processing), but deployed systems are narrow in scope and require extensive domain expertise to validate. No mature, general-purpose product reliably performs the full development or deployment cycle of new NDT methods in production environments without expert human review.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product autonomously develops or conducts new NDT methods like acoustic emission or infrared testing; this remains a specialized engineering research and field activity performed by humans with instruments.

Produce images of objects on film, using radiographic techniques.

15

CI 9–21 · exposure 20 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5NDT is a specialized, physically-grounded field with slower digital transformation; while AI assists in image analysis, the production of radiographic images themselves remains dependent on licensed human operators in traditional workflows with limited automation adoption.
Sector adoption velocityclaude-sonnet-51/5Industrial inspection and manufacturing sectors using radiographic NDT are physical, safety-regulated, and slow to digitize the physical imaging step itself.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist radiographers by automating image analysis, defect detection, and reporting on radiographs already produced, improving productivity in the interpretation phase; however, it provides limited assistance with the actual image acquisition process itself.
Augmentation potentialclaude-sonnet-53/5AI can assist with image analysis, defect detection, and interpretation of radiographs after they are produced, but does not meaningfully augment the physical image-production step itself.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can process and analyze radiographic images post-production, the physical act of operating radiographic equipment, positioning objects, managing exposure settings, and ensuring safety compliance requires human judgment and hands-on technical control that current systems cannot perform end-to-end.
Task automatabilityclaude-sonnet-52/5Producing radiographic images requires physical setup, equipment positioning, radiation safety protocols, and film handling that current AI cannot perform end-to-end; AI cannot physically operate X-ray or gamma-ray equipment.hey key manual/physical steps remain human-performed.
Adoption barriersclaude-haiku-4-5-202510015/5Radiographic work is heavily regulated by nuclear safety authorities and occupational health standards; only licensed radiographers are legally permitted to operate radiographic equipment and produce images, creating a hard regulatory barrier to automation.
Adoption barriersclaude-sonnet-54/5Radiographic NDT often requires certification (e.g., ASNT levels), radiation safety licensing, and regulatory compliance given the hazards of ionizing radiation, creating strong legal/safety barriers to non-human execution.
Cost vs. human wageclaude-haiku-4-5-202510011/5Radiographic equipment is expensive and requires human operators; AI image analysis tools do not reduce the fundamental need for qualified technicians to perform the actual image acquisition, making AI more expensive than traditional human-led workflows when all costs are considered.
Cost vs. human wageclaude-sonnet-51/5AI has no viable substitute for the physical act of positioning objects, operating radiographic sources, and developing film, so cost comparison favors human technicians who already do this at established rates.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed AI systems can autonomously operate radiographic equipment or produce radiographic images from scratch; existing systems only analyze or interpret pre-existing images. The specialized equipment and safety-critical nature of radiation exposure remain human-operated in all production settings.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product autonomously produces radiographic film images of physical objects; this remains a manual/technician-operated process with specialized equipment.

Conduct liquid penetrant tests to locate surface cracks by coating objects with fluorescent dyes, cleaning excess penetrant, and applying developer.

15

CI 5–25 · exposure 13 · augmentation 38 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI-driven automation in NDT remains slow; most facilities still rely on human inspectors despite pilot projects. Sectors like aerospace and automotive use some automated imaging but do not yet replace the human decision-maker, indicating lagging production adoption.
Sector adoption velocityclaude-sonnet-51/5Manufacturing and inspection sectors using NDT are physical, safety-critical, and slow to adopt automation for hands-on chemical testing procedures.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-assisted crack detection (automated image annotation, anomaly highlighting under UV) can help human inspectors prioritize regions and reduce fatigue, but the task still fundamentally requires human judgment to validate findings and certify results.
Augmentation potentialclaude-sonnet-52/5AI could assist with documentation, defect image analysis after visual capture, or scheduling, but offers minimal help with the core physical application and inspection steps described.
Task automatabilityclaude-haiku-4-5-202510012/5While some steps (dye application, developer application) could be partially automated with robotic systems, the core task requires precise visual inspection under UV light to identify and interpret cracks, plus judgment about contamination and surface preparation—capabilities current AI lacks at production quality. End-to-end automation at 50% time saving is not demonstrated today.
Task automatabilityclaude-sonnet-51/5This is a hands-on physical inspection task requiring manual coating, cleaning, and developer application on physical objects, which current AI systems cannot perform without robotic embodiment far beyond off-the-shelf capability.
Adoption barriersclaude-haiku-4-5-202510014/5NDT testing is heavily regulated by industry standards (ASTM, ISO) and often requires certified inspectors; many safety-critical applications (aerospace, nuclear) mandate human qualification and sign-off, creating legal and liability barriers to full automation.
Adoption barriersclaude-sonnet-54/5NDT liquid penetrant testing typically requires certified technicians (e.g., ASNT certification) and adherence to safety/quality standards, creating strong licensing and liability barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5The specialized equipment (UV inspection systems, robotic applicators, image analysis software) combined with integration and human oversight costs substantially exceed the loaded wage of a skilled NDT technician for most part sizes and geometries, making cost-effectiveness poor today.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing the physical chemical application and inspection steps, so AI cost is effectively inapplicable/infinite relative to human labor for this task.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs full liquid penetrant testing autonomously. Robotic coating systems exist in specialized industrial contexts, but autonomous crack detection via fluorescent dye imaging and developer application remain research-stage or prototype-only, with material gaps in real-world variability.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs the physical liquid penetrant testing process end-to-end; this remains a manual technician task requiring physical dexterity and material handling.

Supervise or direct the work of non-destructive testing trainees or staff.

4

CI 0–7 · exposure 0 · augmentation 38 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Even in highly digitized sectors, actual supervisory authority and staff direction remain almost entirely human-managed; no measurable displacement of supervisors by AI systems is occurring in manufacturing or testing environments.
Sector adoption velocityclaude-sonnet-52/5NDT and industrial inspection sectors are physical, safety-critical, and have low digitization of management functions, resulting in slow AI adoption for supervisory roles.
Augmentation potentialclaude-haiku-4-5-202510012/5AI might assist with scheduling, performance dashboards, or anomaly flagging to help a human supervisor, but the core acts of direction, mentoring, and staff management remain fundamentally human, limiting meaningful augmentation.
Augmentation potentialclaude-sonnet-53/5AI can help with scheduling, tracking certifications, analyzing performance data, or generating training materials, offering moderate assistance to a supervisor without replacing judgment-based oversight.
Task automatabilityclaude-haiku-4-5-202510011/5Supervising and directing human staff requires real-time judgment about individual performance, adaptive coaching, conflict resolution, and strategic task allocation—activities that demand contextual understanding of personnel capabilities and organizational goals that current AI systems cannot execute end-to-end.
Task automatabilityclaude-sonnet-51/5Supervising and directing human staff involves interpersonal leadership, mentoring, performance evaluation, and on-the-spot judgment that current AI cannot perform end-to-end.'
Adoption barriersclaude-haiku-4-5-202510015/5Personnel supervision carries fiduciary, legal, and liability responsibilities that by organizational law and convention require a qualified human manager; liability for worker safety, performance evaluation, and disciplinary decisions cannot be delegated to AI.
Adoption barriersclaude-sonnet-54/5NDT often requires certified personnel levels (e.g., ASNT Level III) to supervise and sign off on inspections, creating strong licensing and liability barriers to AI substitution.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI supervisory tools, where they exist at all, serve only as dashboards or alerting aids; the cost of an AI monitoring/directing system would far exceed the loaded wage of an experienced supervisor who provides judgment and accountability.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI product performing this supervisory role, so no meaningful cost comparison favors AI; a human supervisor is required.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product reliably supervises or directs human workers in real organizations; this remains entirely within human organizational management and lacks the human accountability required for personnel oversight.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product manages or supervises NDT trainees/staff; this remains a human management function with no production substitute.

Related occupations — Architecture & Engineering

How to read this

A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.

What would change this score

New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.