Metal-Refining Furnace Operators and Tenders
51-4051.00Median wage $54,430/yr16,780 employed (US)Rank #461 of 923 scored · top 50% by substitution
Operate or tend furnaces, such as gas, oil, coal, electric-arc or electric induction, open-hearth, or oxygen furnaces, to melt and refine metal before casting or to produce specified types of steel.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution28
Exposure25
Augmentation39
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
15 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
7%
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%26
panel mean rating 2.0/5 → substitution pressure 26/100
Technical feasibility todayw 20%23
panel mean rating 1.9/5 → substitution pressure 23/100
Cost vs. human wagew 15%26
panel mean rating 2.0/5 → substitution pressure 26/100
Adoption barriersw 20%inverted — strong barriers lower the score43
panel mean rating 3.3/5 (barrier strength) → substitution pressure 43/100
Sector adoption velocityw 10%21
panel mean rating 1.9/5 → substitution pressure 21/100
Task breakdown (15 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.
Record production data, and maintain production logs.
76CI 72–79 · exposure 75 · augmentation 75 · importance 4.5/5 · click for rater detail
Record production data, and maintain production logs.
76| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Manufacturing and refining sectors are actively adopting MES, IoT, and automated logging systems; adoption is well-established in production facilities and reflects steady, deep penetration in digitized industrial settings. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Heavy industry and metal manufacturing adopt digitization more slowly than information sectors, but automated data logging and industrial IoT are increasingly standard in modernized plants. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI augments operators by auto-populating logs, flagging anomalies, and surfacing insights from production data in real time, significantly raising productivity and data quality even when human operators remain involved in verification or decision-making. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI and automated systems substantially reduce manual burden on operators by auto-capturing and organizing data, letting them focus on monitoring and exception handling. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Recording production data and maintaining logs is largely structured data entry and documentation, which AI systems can automate end-to-end using OCR, sensor data integration, and automated form-filling. Modern production systems already interface with digital logging systems, allowing 60–80% time savings with high accuracy. |
| Task automatability | claude-sonnet-5 | 4/5 | Recording standardized production data and maintaining logs is a structured, repetitive data-entry task easily handled by sensors, SCADA integration, and automated logging systems with human oversight. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Data recording and logging, while sometimes subject to regulatory documentation requirements, do not require a licensed professional to perform or sign off; minor oversight and system validation friction exist but do not prevent substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks automated logging, though some operational and safety documentation may require human sign-off or verification for compliance purposes. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated sensor-to-log pipelines and AI-driven data entry cost a fraction of human operator time once infrastructure is deployed; ongoing inference and integration cost substantially less than loaded wages for routine logging and record-keeping. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated data logging via existing plant sensors and software is far cheaper than paying an operator's time to manually transcribe readings, though initial integration costs exist. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (MES systems with AI integration, automated data capture from sensors and instruments) reliably perform this task in manufacturing environments at scale today. Industrial IoT platforms and ERP integrations demonstrate reliable production in real organizations. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Industrial automation and MES (Manufacturing Execution Systems) already capture furnace parameters and auto-populate production logs in many plants today, though some manual entry and verification persists in smaller operations. |
Regulate supplies of fuel and air, or control flow of electric current and water coolant to heat furnaces and adjust temperatures.
54CI 30–79 · exposure 55 · augmentation 50 · importance 4.6/5 · click for rater detail
Regulate supplies of fuel and air, or control flow of electric current and water coolant to heat furnaces and adjust temperatures.
54| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large integrated steel, aluminum, and chemical plants have largely automated furnace control over the past two decades; smaller foundries and specialty metal shops lag but are steadily adopting modern DCS. Measured displacement is high in mature industrial sectors, with human operators increasingly supervising banks of automatically regulated furnaces rather than manually adjusting each one. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Heavy manufacturing and metal refining are traditionally slow-adopting, capital-intensive sectors with long equipment lifecycles, and process control automation adoption tends to be incremental rather than driven by recent AI advances. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-enabled furnace control systems assist operators by providing real-time diagnostics, predictive alerts for thermal anomalies, and historical trend analysis; operators remain in the loop for judgment calls on emergency shutdowns, maintenance timing, and recipe adjustments, raising their effectiveness in managing multiple units simultaneously. |
| Augmentation potential | claude-sonnet-5 | 3/5 | Modern sensor-based monitoring and predictive analytics can assist operators by flagging anomalies and suggesting adjustments, improving decision-making without replacing the operator's role in physical control and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Modern industrial furnace control systems already rely on automated PID controllers and supervisory control systems that regulate fuel, air, coolant, and temperature with minimal human intervention. Current AI and legacy industrial automation can handle the full feedback-loop task of maintaining setpoint temperatures with <50% of traditional manual operator time; the remaining human role is exception handling and monitoring rather than active regulation. |
| Task automatability | claude-sonnet-5 | 2/5 | This is a physical process-control task requiring real-time sensor monitoring and equipment adjustment in a hazardous industrial environment; while control-loop automation exists, full end-to-end automation replacing the human tender is not achievable with off-the-shelf AI today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While automation is common, regulatory oversight (OSHA, environmental permits, safety interlocks) and liability for temperature excursions or equipment damage create material friction; operators remain on-site for emergency override and fault diagnosis, and some jurisdictions require licensed personnel to certify setpoints or respond to alarms. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Safety regulations, liability for equipment failure/accidents, and the need for human oversight in hazardous high-temperature industrial settings create meaningful friction, though not strict licensing requirements comparable to medical or legal fields. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | A single automated control system serves multiple furnaces and runs continuously at marginal inference cost (sensor reads, PID calculations), compared to one or more human furnace operators earning full loaded wages; the AI cost per task-equivalent is orders of magnitude lower. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Advanced furnace automation requires significant capital investment in sensors, actuators, and control systems that must be integrated with existing plant infrastructure, making the all-in cost comparable to or higher than retaining human operators in many facilities. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Deployed SCADA systems, DCS (distributed control systems), and advanced furnace controllers are in production across steel, aluminum, and chemical industries worldwide, reliably performing real-time regulation of fuel/air/coolant/temperature 24/7 at scale with proven reliability metrics. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Industrial control systems and PLCs have long automated portions of furnace regulation, but these are engineering/control systems rather than general AI products, and human operators remain in the loop for monitoring and adjustment at most facilities. |
Weigh materials to be charged into furnaces, using scales.
49CI 25–72 · exposure 50 · augmentation 50 · importance 4.5/5 · click for rater detail
Weigh materials to be charged into furnaces, using scales.
49| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Metal refining is a mature, physically-grounded industrial sector with lower digital transformation rates compared to information-intensive sectors. While some facilities use automated weighing systems, full task automation adoption remains limited and slow in production environments. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Heavy industry and metal manufacturing have moderate digitization; automated material handling and weighing systems are common in larger plants but adoption is uneven across smaller operations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | Digital scales with integrated data logging, alerts, and batch-tracking systems provide meaningful assistance to operators by reducing manual recording and enabling real-time inventory checks. However, augmentation is confined to measurement and documentation aspects; core safety and judgment remain human-centric. |
| Augmentation potential | claude-sonnet-5 | 3/5 | Digital scales and sensor readouts assist operators by providing precise, real-time weight data, improving accuracy and speed even where full automation of charging isn't implemented. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While weighing materials using scales is mechanically simple and could be partially automated with load cells and data logging systems, the task requires integration with furnace charging workflows and real-time judgment about material placement. Current AI systems lack reliable physical interaction capabilities to consistently perform the full charging workflow end-to-end with 50% time savings. |
| Task automatability | claude-sonnet-5 | 4/5 | Weighing materials is a straightforward physical measurement task easily handled by automated scales, load cells, and PLC-integrated batching systems that already exist in industrial settings.can be fully automated with sensor-based weighing and control systems. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Furnace operations are heavily regulated under occupational safety and environmental standards; operator certification and direct human responsibility for material handling and safety monitoring create strong legal and liability barriers to full automation. OSHA requirements and union agreements in many facilities mandate human operators maintain control. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal requirement mandates human weighing; the main barriers are capital investment and integration with existing furnace charging equipment, which is a moderate but surmountable organizational hurdle. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Deploying automated scales and data systems has upfront capital costs comparable to operator wages, and requires integration and maintenance. The cost advantage over a single operator performing this task is marginal once all infrastructure and oversight are included. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated scale and batching systems have high upfront cost but very low marginal cost per weighing operation compared to continuous human labor, especially at scale in continuous furnace operations. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Industrial weighing systems with automated logging exist and are deployed, but they typically operate as passive measurement devices requiring human interpretation and action. No mature product autonomously performs the complete task of weighing, deciding charge quantity, and physically managing furnace material loading without significant human oversight. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Automated weighing/batching systems are mature, widely deployed technology in metal refining and other process industries, though some plants still rely on manual weighing for smaller or specialty batches. |
Drain, transfer, or remove molten metal from furnaces, and place it into molds, using hoists, pumps, or ladles.
44CI 5–84 · exposure 45 · augmentation 25 · importance 4.4/5 · click for rater detail
Drain, transfer, or remove molten metal from furnaces, and place it into molds, using hoists, pumps, or ladles.
44| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large-scale and mid-tier metalworking operations have been automating casting and pouring for decades; adoption is mature and ongoing in the industrial/manufacturing sector. Smaller foundries and specialized shops show slower adoption, but the overall sector trend is well-established and accelerating. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Metal refining is a heavy industrial, physically intensive sector with low AI/software adoption; this specific task relies on industrial automation/robotics rather than AI systems, and change is slow. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | This task has limited augmentation upside; AI assistance in real-time adjustment of temperature or pouring rate could enhance human control, but the physical danger and the superior speed and consistency of full automation make human-in-the-loop assistance less compelling than outright replacement. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI/sensor systems can assist with monitoring furnace temperature, timing, and predictive maintenance to inform when and how to drain metal, but does not materially change the physical execution of this task. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Pouring molten metal from furnaces into molds involves highly structured, repetitive physical motions that can be fully automated with robotic arms, hoists, and sensor-guided systems. Modern foundries already deploy automated casting systems that achieve >50% time savings and equal or superior quality compared to manual operators. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a physical manipulation task involving molten metal, hoists, pumps, and ladles in a hazardous industrial environment; no AI system can perform the physical transfer of molten material end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While there are occupational safety regulations and equipment standards, no licensing requirement mandates a human pour molten metal, and automation is legal and common. Primary barriers are capital investment and worker retraining friction, not regulatory or liability constraints on automation itself. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Extreme safety, liability, and physical handling requirements around molten metal create strong barriers to any non-human or non-specialized-robotic substitution, though not a licensing requirement per se. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Automated casting equipment has high capital costs but very low per-unit inference and operational costs once installed. For high-volume operations, the cost per task-equivalent is substantially below the loaded wage of a furnace operator; the ratio favors automation at typical throughput. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI-based substitute performing this physical task, so cost comparison to human labor is not applicable; any automation here would be specialized robotics, not general AI, at high capital cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Industrial robotic pouring and casting systems are deployed in production metalworking facilities today, though deployment is not yet universal across all foundry types. Mature solutions exist for high-volume operations; some specialized or smaller foundries still rely on manual labor, indicating material but not universal production maturity. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product performs molten metal transfer autonomously; this remains a manual or robotics/automation-controlled process, not an AI cognitive task. |
Observe air and temperature gauges or metal color and fluidity, and turn fuel valves or adjust controls to maintain required temperatures.
28CI 25–30 · exposure 25 · augmentation 50 · importance 4.5/5 · click for rater detail
Observe air and temperature gauges or metal color and fluidity, and turn fuel valves or adjust controls to maintain required temperatures.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Metal refining is capital-intensive and safety-focused, with slow technology adoption cycles. Most facilities rely on legacy control systems and skilled human operators; autonomous AI control of critical furnace parameters remains rare in production despite decades of industrial automation. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Heavy industry and metal refining are traditionally slow adopters of AI/automation compared to information-sector industries, with capital-intensive legacy equipment slowing modernization. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-assisted monitoring systems (real-time gauge alerts, temperature trend prediction, recommendations) can improve operator situational awareness and reduce manual monitoring burden. However, the human operator typically remains the primary decision-maker for control adjustments in this safety-critical role. |
| Augmentation potential | claude-sonnet-5 | 3/5 | Sensor dashboards, predictive analytics, and automated alerts can meaningfully assist operators in monitoring gauges and predicting temperature deviations, improving decision-making without replacing the operator. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI vision systems can monitor gauges and infer temperature from metal color, the task requires real-time physical control adjustments in a dynamic, safety-critical environment with tight feedback loops. Current AI lacks the integrated sensorimotor capability and reliability to close the loop end-to-end without human oversight, limiting automation to well below 50% time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | While sensor-based monitoring and PID/control-loop automation can handle temperature regulation, visual assessment of metal color/fluidity as a quality cue and integrated judgment in variable furnace conditions still requires human oversight or specialized computer vision not yet standard in most plants.'}, |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Heavy regulatory barriers exist: furnace operations are covered by OSHA and industry safety standards that typically require a licensed, certified human operator to be responsible for temperature control and safety. Liability for metal quality and process failures creates strong legal and contractual requirements for human sign-off. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically, but safety regulations, liability for furnace failures/explosions, and reliance on experienced judgment for quality control create meaningful organizational and safety-driven barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Integrating vision systems, control algorithms, safety certification, and continuous monitoring infrastructure is expensive. The loaded cost of AI infrastructure and oversight labor approaches or exceeds the wage of a skilled furnace operator, especially when factoring in reliability and liability requirements. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Retrofitting furnaces with sensors, cameras, and control automation requires significant capital investment, and human operators remain relatively low-cost compared to specialized industrial automation systems for this niche task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Computer vision for gauge reading and metal color classification exists in research, but deployed industrial systems that independently manage furnace temperature control remain limited. Most production furnaces still rely on human operators or legacy automation with heavy human supervision, not autonomous AI agents. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Advanced steel mills use automated combustion control and some optical pyrometry, but fully autonomous systems replacing operator judgment on color/fluidity across diverse furnace types are not widely deployed in production. |
Observe operations inside furnaces, using television screens, to ensure that problems do not occur.
28CI 25–30 · exposure 25 · augmentation 50 · importance 4.3/5 · click for rater detail
Observe operations inside furnaces, using television screens, to ensure that problems do not occur.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Metal refining is a capital-intensive, safety-critical, and traditionally labor-managed industry; adoption of autonomous monitoring remains slow and experimental in most operations, with most sites still relying on human observers despite digitization of other processes. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Heavy industry and metal refining are traditionally slow adopters of AI compared to information/professional sectors, with automation focused more on process control than perceptual monitoring. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI could assist operators by flagging potential anomalies or highlighting patterns in furnace-camera footage, reducing fatigue and improving detection of subtle problems, but the human operator would remain the primary decision-maker and actor—a moderate augmentation rather than transformation of productivity. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-based visual analytics can augment operators by highlighting anomalies or trends on screens, improving attentiveness and response time without removing the human from the loop. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Visual monitoring of furnace operations via camera feeds could theoretically be partially automated with computer vision, but the task requires real-time anomaly detection, contextual understanding of equipment states, and immediate human judgment to intervene—capabilities current AI systems perform unreliably at production scale. Meaningful automation would require extensive integration, custom training on furnace-specific anomalies, and humans remaining in the loop for most decisions, falling well short of 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Computer vision could flag anomalies in furnace video feeds, but full end-to-end monitoring with equal-quality judgment about complex thermal/metallurgical states still requires human interpretation and response, limiting time savings.dictionary |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong safety and liability barriers exist: furnace operation failures can cause equipment damage, worker injury, or environmental harm, creating legal and insurance requirements that a trained, licensed operator typically must bear responsibility for monitoring and oversight. Regulatory frameworks in metallurgy often mandate human accountability for continuous observation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically for this monitoring subtask, but safety-critical nature of furnace operations creates strong organizational caution and liability concerns around relying solely on automated visual monitoring. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI computer-vision systems, when customized for furnace monitoring with adequate uptime and integration support, carry substantial costs in setup, training, and required human oversight that approach or exceed the loaded wage of an operator performing the same surveillance task. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Installing and maintaining specialized vision/AI monitoring systems with sensor integration and calibration is costly relative to an operator's marginal task cost, though it may lower risk over time. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While computer vision systems can detect basic visual changes in footage, no mature, deployed product reliably performs autonomous furnace health monitoring with the accuracy and low false-positive rates required in high-temperature industrial operations. Prototypes exist but lack the robustness and domain specificity needed for production deployment in critical metallurgical settings. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some industrial vision-based anomaly detection systems exist in steel/metal plants, but they are narrow, often supplementary alarms rather than full replacement of visual furnace observation by operators. |
Inspect furnaces and equipment to locate defects and wear.
26CI 21–30 · exposure 17 · augmentation 50 · importance 4.4/5 · click for rater detail
Inspect furnaces and equipment to locate defects and wear.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Metal refining is a capital-intensive, traditional sector with slower digital transformation; most furnace inspection remains manual. While some large operations pilot thermal imaging, widespread deployment of autonomous AI inspection agents is still limited, reflecting both technical and organizational conservatism. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Heavy manufacturing and metal-refining are traditionally slow-adopting sectors for AI, with physical inspection tasks lagging behind office-based automation trends. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-powered thermal imaging and defect-detection tools can meaningfully assist human inspectors by highlighting anomalies and automating routine scanning, reducing inspection time and improving consistency. However, the human inspector's judgment about severity and maintenance scheduling remains essential, limiting augmentation to moderate levels. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-enabled sensors, thermal imaging, and predictive analytics can flag anomalies and prioritize inspection points, meaningfully assisting operators even though physical inspection remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Visual inspection of furnaces for defects and wear can be partially automated using computer vision and thermal imaging systems, but current AI struggles with the complex 3D assessment of wear patterns in harsh, high-temperature environments and requires significant human validation. This falls short of the 50% time-saving threshold because human expertise in interpreting subtle signs of equipment degradation remains critical. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires physical inspection of furnace interiors and equipment surfaces in harsh industrial environments, which current AI systems cannot perform end-to-end without robotic embodiment and sensor infrastructure that is not standard. Off-the-shelf AI cannot replace the physical inspection process today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Safety-critical furnace inspection creates moderate barriers: regulatory requirements often mandate documented human sign-off, and liability for missed defects that cause equipment failure or worker injury falls on operators. However, no explicit licensing requirement prevents AI augmentation of human inspectors. |
| Adoption barriers | claude-sonnet-5 | 3/5 | While not always legally mandated to be human-performed, safety regulations in industrial settings and liability for undetected furnace failures create strong organizational incentives to retain human inspection and sign-off. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Initial AI inspection system costs (hardware, software, integration, maintenance) are comparable to or exceed the cost of periodic human inspector labor, especially when factoring in the need for human oversight and validation of AI findings. The ROI improves only at scale across many furnaces. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Sensor arrays, thermal cameras, and monitoring software require significant capital investment and integration costs that often exceed the marginal cost of a skilled operator performing visual/physical checks, especially at smaller facilities. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While thermal imaging and computer vision products exist for industrial inspection, they are primarily narrow-scoped tools requiring extensive setup, calibration, and human interpretation rather than end-to-end autonomous inspection systems operating reliably in production metalworks. Deployed solutions have material limitations in variable industrial settings with extreme heat and complex geometries. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some industrial predictive-maintenance and thermal-imaging/computer-vision systems exist for equipment monitoring, but full defect and wear inspection of furnaces via deployed AI products remains narrow, sensor-dependent, and not a substitute for hands-on physical inspection. |
Operate controls to move or discharge metal workpieces from furnaces.
25CI 25–25 · exposure 25 · augmentation 50 · importance 4.5/5 · click for rater detail
Operate controls to move or discharge metal workpieces from furnaces.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Metal refining remains a capital-intensive, geographically concentrated sector with slow digitization in many facilities. While large integrated mills have adopted some automation, the broader sector (smaller foundries, regional refineries) lags significantly; adoption is measured in years, not months. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Heavy industry and metal refining are traditionally slow adopters of AI-driven automation compared to information/professional services, though some large plants use automation controls. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-assisted monitoring systems (thermal imaging, predictive maintenance alerts, workpiece tracking) can meaningfully assist operators in managing furnace conditions and scheduling discharge cycles, improving efficiency and safety decision-making while the operator remains in control. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-based monitoring, predictive maintenance, and control optimization can assist operators in timing and safety decisions, improving efficiency while humans remain in control of physical operations. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Moving and discharging metal workpieces from furnaces involves physical manipulation in a high-temperature, hazardous environment with precise spatial coordination. Current AI lacks integrated robotic systems deployed at scale in furnaces; while some industrial robots exist, they require extensive custom integration and cannot autonomously handle the variability in workpiece geometry, furnace conditions, and safety constraints. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires physical control of heavy machinery and real-time sensory judgment (visual/thermal cues, timing) in a hazardous environment; current AI cannot end-to-end replace the physical operation of moving or discharging molten/hot workpieces.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong regulatory and safety barriers exist: operators must comply with OSHA requirements, thermal safety protocols, and worksite hazard management. The physical environment demands human judgment on real-time safety and equipment integrity; liability for automated failures in high-temperature operations creates organizational friction. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Safety regulations, liability for industrial accidents, and the need for human oversight in hazardous high-temperature environments create strong barriers to full unsupervised automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Custom robotic furnace systems are capital-intensive, requiring significant upfront investment, integration costs, and maintenance. For most small-to-mid-size operations, the total cost of ownership remains higher than employing a skilled furnace operator, though large-scale mills show better economics. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Retrofitting furnaces with sensors, robotics, and control automation is capital-intensive; AI software alone cannot perform the physical task, so all-in automation cost is often comparable to or higher than a trained operator, at least initially. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Specialized robotic systems for furnace operations exist in limited production deployments, but most modern furnaces still rely on human operators due to the complexity of thermal management, workpiece positioning, and real-time safety decisions. No general-purpose AI system reliably performs this task end-to-end without human oversight. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some industrial furnaces use automated/robotic control systems, but these are engineering/automation solutions rather than general AI products, and full autonomous discharge control is not widely deployed reliably across the industry. |
Draw smelted metal samples from furnaces or kettles for analysis, and calculate types and amounts of materials needed to ensure that materials meet specifications.
24CI 18–30 · exposure 20 · augmentation 38 · importance 4.6/5 · click for rater detail
Draw smelted metal samples from furnaces or kettles for analysis, and calculate types and amounts of materials needed to ensure that materials meet specifications.
24| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Metal refining is a capital-intensive, safety-regulated, and historically lower-digitization sector with limited venture capital interest in furnace automation. Adoption has remained incremental and concentrated in large integrated mills, not widespread across the industry. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Heavy industry/metals processing is a low-digitization, physically intensive sector with historically slow AI/robotics adoption compared to information-sector benchmarks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI can assist with specification lookups and material calculation suggestions given input data, but the hazardous sampling task itself offers little room for human-AI collaboration at present. The augmentation is limited to the calculation portion and does not materially transform operator productivity on the full task. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-assisted calculation tools and predictive models for alloy composition can meaningfully help operators determine material types/amounts, even though the physical sampling remains manual. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can theoretically assist with calculation of material types and amounts given specification data, the physical task of drawing metal samples from high-temperature furnaces requires real-time sensorimotor control in hazardous conditions that current robotics cannot reliably perform. The analytical component alone is incomplete without the sampling step, preventing end-to-end automation. |
| Task automatability | claude-sonnet-5 | 2/5 | Physical sampling from molten metal requires manual/robotic intervention that current general AI cannot perform; only the calculation portion (material specs, ratios) is automatable, leaving the bulk of the task unaddressed by off-the-shelf AI. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: occupational safety regulations (OSHA) mandate human oversight and authorization for hazardous furnace operations; liability for failed sample extraction or miscalculation is high; and direct human control of dangerous equipment is often legally required. Metallurgical expertise and on-site troubleshooting discretion also create organizational friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement per se, but safety regulations around molten metal handling, quality/safety liability for material specification errors, and physical plant environment constraints create meaningful friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Specialized robotics capable of operating in furnace environments with integrated analytical feedback systems would be extremely capital-intensive, likely far exceeding the loaded wage of a furnace operator over several years. Integration and maintenance costs would further compound the expense relative to human labor. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Robotic sampling systems and lab automation require significant capital investment and specialized integration, so near-term all-in AI costs are not clearly cheaper than skilled operator labor for this hybrid physical-cognitive task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed commercial system reliably performs both the hazardous sample extraction from active furnaces and integrated quality-specification calculations in production smelting environments. Robotic sample extraction in such conditions remains developmental, and fully integrated solutions do not exist at scale in real refineries. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some plants use software for metallurgical calculations and sensor-based composition analysis, but the physical sample-drawing step and integrated end-to-end automation are not widely deployed in production. |
Prepare material to load into furnaces, including cleaning, crushing, or applying chemicals, by using crushing machines, shovels, rakes, or sprayers.
18CI 5–30 · exposure 8 · augmentation 25 · importance 4.5/5 · click for rater detail
Prepare material to load into furnaces, including cleaning, crushing, or applying chemicals, by using crushing machines, shovels, rakes, or sprayers.
18| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Metal refining is a capital-intensive, slow-to-modernize sector with dispersed facilities and bespoke processes. Adoption of advanced automation is pilot-stage at major integrated mills; small to mid-sized foundries and refineries remain labor-dependent due to low digitization and high switching costs. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Heavy industrial metal refining is a low-digitization, physical-labor sector with minimal AI agent adoption for manual material handling tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | Current AI offers minimal augmentation for this task. Sensors and analytics can monitor material properties and suggest batch parameters, but the physical labor of cleaning, crushing, and spraying still depends on the human operator's strength, precision, and real-time judgment rather than AI assistance. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI could assist with monitoring material composition or optimizing chemical application ratios via sensors and analytics, but doesn't meaningfully change the physical labor itself. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Material preparation involves physical handling (cleaning, crushing, chemical application) in a dynamic furnace environment where real-time adjustments and safety monitoring are critical. While individual components like crushing could be partially automated, the end-to-end task requires situational awareness, physical dexterity, and safety judgment that current AI systems cannot reliably replicate without substantial infrastructure investment. |
| Task automatability | claude-sonnet-5 | 1/5 | This is physical manual labor involving handling raw materials with hand tools and machinery in an industrial setting; no current AI system can perform physical manipulation of materials.confidence |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Metal-refining furnace operations are heavily regulated for worker safety, material handling, and environmental compliance (OSHA, EPA). A human operator is legally required to monitor furnace conditions, material quality, and emergency shutdown; automation of prep work alone does not eliminate the human's liability and sign-off role, creating a structural barrier to full substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for this specific task, but physical/industrial environment and safety protocols create moderate operational friction against fast substitution by AI/software alone. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Deploying robotic systems for material prep (conveyor automation, crushing machinery, spraying rigs) carries high capital and integration costs relative to a furnace operator's loaded wage. Ongoing maintenance, customization per material type, and safety compliance add overhead that does not yet achieve parity, let alone cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI software has no direct cost equivalence here since the task is physical; any automation would require specialized industrial robotics with high capital costs exceeding human labor in most contexts. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed production systems reliably automate the full sequence of heterogeneous material preparation tasks (cleaning, crushing, chemical spraying) in active furnace operations. Isolated automation of crushing exists, but integrating it into a coherent furnace-feed workflow with quality and safety verification remains research-stage. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product performs physical material preparation like crushing, cleaning, or chemical application; this requires robotics/physical actuation not general AI systems. |
Scrape accumulations of metal oxides from floors, molds, and crucibles, and sift and store them for reclamation.
14CI 5–24 · exposure 8 · augmentation 13 · importance 3.7/5 · click for rater detail
Scrape accumulations of metal oxides from floors, molds, and crucibles, and sift and store them for reclamation.
14| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Metal refineries are capital-intensive, regulated industrial settings with slow technology adoption cycles. This task is performed by a small, geographically dispersed workforce in legacy facilities where process changes are infrequent and expensive. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Metal refining and heavy industrial manufacturing are low-digitization, physically intensive sectors with minimal AI/robotic adoption for manual floor-level tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist marginally by identifying optimal oxide recovery locations or predicting reclamation value, but the core physical scraping, sifting, and storage work offers little room for AI-based productivity enhancement while an operator remains present. |
| Augmentation potential | claude-sonnet-5 | 1/5 | AI offers essentially no assistance to a human physically scraping oxide residue and sifting material; this is not an information or decision task AI can support. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI-guided robotics could theoretically scrape oxides, the task requires navigating variable physical environments (floors, molds, crucibles of different geometries), detecting oxide accumulation, and handling fragile equipment. Current deployed systems lack the dexterity, environmental adaptation, and real-time error correction to achieve 50% time savings at equal quality end-to-end. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a manual, physical scraping and sifting task in a harsh industrial environment requiring dexterity and mobility; no current AI system can perform this physical labor. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Safety regulations, furnace environment hazards, worker-protection requirements, and the need for human judgment about oxide quality and reclamation viability create substantial adoption friction. The hot, confined, and variable nature of furnace work also creates liability concerns for full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks automation, but physical environment hazards, heat, and material handling create practical organizational and safety barriers to any automation attempt. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The capital cost of a robotic system capable of navigating a furnace floor, identifying oxides, scraping safely without damaging equipment, and sifting/storing material would far exceed the loaded wage of the operator performing this task, which is a routine manual operation. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI/robotic substitute in production, so the comparison defaults to AI being effectively unavailable and thus not cost-competitive. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed commercial products reliably perform this specific task in production furnace environments. Robotic applications in metal refineries are extremely limited and specialized; this combination of scraping, sifting, and storage remains a manual operation in practice. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs this physical scraping/sifting/storage task; advanced robotics for this specific dirty, variable-geometry work is not commercially deployed. |
Remove impurities from the surface of molten metal, using strainers.
12CI 5–19 · exposure 8 · augmentation 25 · importance 4.3/5 · click for rater detail
Remove impurities from the surface of molten metal, using strainers.
12| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Metal refining is a traditional, capital-intensive sector with legacy equipment; automation adoption is slow and concentrated in large mills, and most sites still rely on skilled human operators for safety and quality control. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Heavy industry and metal refining are low-digitization, physical-labor-intensive sectors with minimal AI/robotic adoption for this specific molten-metal handling task. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI-assisted visual monitoring or temperature sensors could provide decision support, but the core task of physically removing impurities with precision and safety remains human-dependent; augmentation potential is limited. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI-based sensors or vision systems could potentially help monitor impurity levels or timing, offering minor process guidance, but they do not currently assist with the physical skimming action itself. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI-controlled robotic systems could theoretically manipulate strainers, the task requires real-time visual assessment of molten metal surface conditions, precise timing, and adaptation to equipment state—capabilities that exist only in limited experimental settings, not production automation. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a manual, physical task requiring direct handling of molten metal with hand tools; no current AI system can perform the physical skimming action itself, only robotics with sensing might approach it, but that's not deployed today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Occupational safety regulations, worker licensing/certification, equipment liability, and the hazardous environment create strong friction; human oversight and signoff are typically required for furnace operations in most jurisdictions. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically bars automation, but extreme heat, safety hazards, and the need for physical precision create strong practical and engineering barriers to substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Integrating and maintaining specialized robotic systems for furnace automation is significantly more expensive than the loaded wage of a furnace operator, especially given low-volume production sites and safety redundancy requirements. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute for this physical action, so any hypothetical automation (specialized robotics) would require heavy capital investment far exceeding the cost of a human operator for this narrow task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed commercial product reliably performs this task end-to-end; it requires integration of computer vision, robotic control, thermal sensing, and safety systems that are not mature in production furnace environments. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI or robotic product performs molten metal dross/impurity skimming in production; this remains a manual operation in foundries and refineries. |
Kindle fires, and shovel fuel and other materials into furnaces or onto conveyors by hand, with hoists, or by directing crane operators.
7CI 5–10 · exposure 0 · augmentation 25 · importance 4.3/5 · click for rater detail
Kindle fires, and shovel fuel and other materials into furnaces or onto conveyors by hand, with hoists, or by directing crane operators.
7| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Metal refining and foundry work remains heavily manual and low-digitization; adoption of advanced automation in this sector is slow, with most operations still reliant on human operators in controlled environments. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Metal refining and heavy industrial furnace operations are a laggard sector for AI adoption, being highly physical, hazardous, and poorly digitized compared to information-based work. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | Limited augmentation is possible—sensors could alert operators to temperature or material levels, but the core tasks of kindling and shoveling require human physical judgment and presence in an inherently hazardous setting. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI could assist with monitoring furnace conditions or optimizing fuel timing via sensors, but offers little direct help with the physical acts of kindling and shoveling. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires physical manipulation in a hazardous industrial environment—kindling fires, shoveling, and coordinating with crane operators. Current AI systems cannot perform these physical actions end-to-end today. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a physical manual labor task involving fire kindling, shoveling, and directing heavy equipment in a furnace environment—current AI systems cannot perform this physical manipulation end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | OSHA regulations, safety protocols, and liability for equipment operation and material handling create strong legal and organizational barriers. Human operators must be licensed/certified for hazardous materials and equipment operation, and human oversight is typically mandated. |
| Adoption barriers | claude-sonnet-5 | 3/5 | While no licensing requirement exists specifically for this task, workplace safety regulations, liability for furnace accidents, and physical environment complexity create meaningful barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Industrial robots capable of handling high-temperature furnace work, vision systems, and coordination infrastructure would cost far more than the loaded wage of a furnace operator, especially accounting for safety redundancy and maintenance. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI/robotic substitute performing this physical task, so any hypothetical automation would require expensive custom robotics far exceeding human labor costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed autonomous systems reliably perform furnace fueling and material handling at industrial scale. Robotics in foundries exist but are narrow and require extensive customization; this is not a mature off-the-shelf capability. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs manual fuel-shoveling and furnace-tending tasks; robotics for this specific unstructured industrial physical work remains research-stage at best. |
Sprinkle chemicals over molten metal to bring impurities to the surface.
7CI 5–10 · exposure 0 · augmentation 25 · importance 4.2/5 · click for rater detail
Sprinkle chemicals over molten metal to bring impurities to the surface.
7| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Metal refining remains a capital-intensive, low-digitization heavy industry with long equipment lifecycles and strong union labor presence; automation adoption is measured and conservative, not rapid. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Heavy metal refining is a low-digitization, physical industrial sector with minimal AI/robotic deployment for hands-on furnace tending tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with monitoring impurity readings or recommending chemical ratios via sensors and analytics, but the core task of precision chemical application during live furnace operation offers limited room for meaningful human-AI collaboration. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI-driven sensors and process control systems can help monitor furnace conditions and optimize chemical timing/dosage, offering modest assistance, but the physical application remains manual. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task involves real-time physical interaction with extreme-temperature molten metal requiring precise spatial judgment and immediate adaptation. Current AI cannot safely deploy robotic systems in unstructured furnace environments or predict dynamic impurity behavior at scale. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a physical manipulation task requiring dexterous handling of chemicals near molten metal; no current AI system can perform this manual, hazardous physical action end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: safety regulations require licensed operators to oversee molten-metal processes, liability risk is asymmetric (chemical misapplication could destroy product or injure workers), and insurance typically mandates human supervision of furnace operations. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically bars automation, but extreme heat, safety hazards, and precision timing create strong practical and safety-driven barriers to substitution by generic AI. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Robotic systems capable of handling furnace environments with thermal protection and chemical application are extremely expensive to install, maintain, and integrate compared to a furnace operator's loaded wage. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this physical task, so any AI cost comparison is moot; specialized robotics for this would be far more expensive than current human labor. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed commercial product reliably performs chemical sprinkling on active molten metal. This remains exclusively a human operator task in production; no automation is in standard industrial use. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI/robotic product performs this specific molten metal fluxing/skimming operation in production; this remains a manual or specialized-machinery task. |
Direct work crews in the cleaning and repair of furnace walls and flooring.
3CI 0–5 · exposure 0 · augmentation 25 · importance 4.0/5 · click for rater detail
Direct work crews in the cleaning and repair of furnace walls and flooring.
3| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Metal refining is a capital-intensive, regulated heavy industry with deeply embedded manual crew supervision practices and minimal documented AI adoption for operational crew direction tasks. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Heavy industrial metal refining is a low-digitization, physically intensive sector with minimal AI/agent adoption for on-site crew supervision and manual repair tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with scheduling optimization or safety compliance logging, but supervisory direction of live crews performing hazardous work remains fundamentally human-centered; limited augmentation opportunity. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI could assist with scheduling, documentation, or predictive maintenance planning, but offers little direct help for the physical crew-directing and repair task itself. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires real-time direction of physical crews performing hands-on maintenance work in hazardous furnace environments. Current AI cannot manage dynamic crew coordination, safety oversight, or adapt to unpredictable field conditions on site. |
| Task automatability | claude-sonnet-5 | 1/5 | This is physical crew supervision combined with hands-on furnace repair work in a hazardous industrial environment; current AI cannot direct people or perform physical labor.4Its non-cognitive, physical-supervisory nature places it far outside current AI capability. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Furnace operations are heavily regulated for worker safety and environmental compliance; a licensed, responsible human must legally oversee crew work in hazardous conditions and bear accountability for safety incidents. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Safety regulations, physical presence requirements, and the need for experienced human judgment in hazardous furnace environments create strong practical barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | An experienced furnace operator directing crews serves a mission-critical supervisory and safety function; the cost of AI systems capable of real-time crew management with liability coverage would exceed the loaded wage of the human operator. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this physical, on-site supervisory and manual labor task, so AI cost comparison is not applicable and the human remains the only viable option. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system can autonomously direct physical work crews in real-world furnace maintenance operations. This requires embodied presence, authority, and dynamic decision-making that existing products do not reliably perform. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product supervises work crews doing furnace maintenance or performs physical inspection/repair of refractory linings and flooring. |
Related occupations — Production
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.