Power Plant Operators

51-8013.00

Median wage $102,040/yr29,320 employed (US)Rank #543 of 923 scored · top 59% by substitution

Control, operate, or maintain machinery to generate electric power. Includes auxiliary equipment operators.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution25

Exposure28

Augmentation56

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

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

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

Tasks on the substitution scale

30 rated tasks, binned by substitution score.

Position among all scored occupations

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

Tasks with substitution ≥ 70

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

Why this score

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

Task automatabilityw 35%28

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

Technical feasibility todayw 20%28

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

Cost vs. human wagew 15%27

panel mean rating 2.1/5 → substitution pressure 27/100

Adoption barriersw 20%inverted — strong barriers lower the score19

panel mean rating 4.2/5 (barrier strength) → substitution pressure 19/100

Sector adoption velocityw 10%21

panel mean rating 1.9/5 → substitution pressure 21/100

Task breakdown (30 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 and compile operational data by completing and maintaining forms, logs, or reports.

CI 72–87 · exposure 83 · augmentation 75 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	4/5	Power generation and utilities are information-intensive, already digitized sectors with strong SCADA/IT infrastructure; process automation adoption in utilities is accelerating as companies modernize operations and reduce manual clerical work.
Sector adoption velocity	claude-sonnet-5	3/5	Utilities and energy sector show moderate digitization with historians and automated reporting already common, but broader AI-driven compilation and analysis adoption is still emerging and uneven across plants.
Augmentation potential	claude-haiku-4-5-20251001	4/5	Even where human review remains required, AI-assisted form completion and automated data compilation dramatically accelerate operator productivity by reducing manual keystroke and transcription burden, leaving human focus on judgment and validation.
Augmentation potential	claude-sonnet-5	4/5	AI tools can significantly reduce manual transcription burden and flag anomalies in logs, letting operators focus on verification and judgment calls rather than raw data entry.
Task automatability	claude-haiku-4-5-20251001	5/5	Recording and compiling operational data into forms, logs, or reports is predominantly structured data entry and formatting—core strengths of current AI systems. Off-the-shelf RPA, document automation, and data-processing agents can extract sensor readings, populate standardized forms, and generate reports with high accuracy and >50% time savings.
Task automatability	claude-sonnet-5	4/5	Recording and compiling operational data into logs/forms is largely structured data entry and reporting, which AI-integrated SCADA/DCS systems and automated logging tools can already handle with minimal human input for most of the workflow.
Adoption barriers	claude-haiku-4-5-20251001	2/5	While data validation and quality-assurance sign-off may be required by utility regulations, the recording and compilation itself does not legally mandate human execution—only oversight. Minimal licensing or liability barriers exist for the automation of this routine clerical task.
Adoption barriers	claude-sonnet-5	2/5	Some regulatory recordkeeping requirements (e.g., NRC, EPA compliance logs) may require human attestation or oversight, but the underlying data compilation itself faces few hard legal barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	5/5	Inference and integration costs for automated data logging and report generation are minimal (pennies per record); the human loaded cost for manual data entry is orders of magnitude higher, making AI substitution economically decisive.
Cost vs. human wage	claude-sonnet-5	4/5	Automated data logging via existing SCADA/historian systems is vastly cheaper per data point than manual transcription once implemented, though initial integration costs exist.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	Mature products (RPA platforms, business process automation software, AI data extraction tools) reliably perform structured data logging and report generation in production environments. Some residual friction exists around integration with legacy SCADA systems and validation of critical safety data, but the core capability is widely deployed.
Technical feasibility today	claude-sonnet-5	4/5	Modern plant control systems and industrial data historians already automatically log sensor readings, alarms, and operational parameters into structured reports, though some manual verification and narrative log entries persist.

Prepare and submit compliance, operational, and safety forms or reports.

CI 55–67 · exposure 70 · augmentation 75 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Utilities and power operators are in the middle adoption tier—many have invested in SCADA and operational data systems but lag in full AI-driven compliance automation. Pilots of automated reporting exist, but widespread production deployment remains limited; organizational inertia and regulatory caution slow velocity.
Sector adoption velocity	claude-sonnet-5	2/5	Power generation is a heavily regulated, industrial sector with historically slow IT/AI adoption relative to information or financial services, though some digitization of reporting is underway.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI substantially improves operator productivity by auto-populating fields from real-time plant data, flagging anomalies, and organizing compliance information, allowing operators to focus on verification and sign-off rather than manual data entry. This maintains human oversight while reducing administrative burden.
Augmentation potential	claude-sonnet-5	4/5	AI can significantly speed up drafting, formatting, and cross-checking of reports and forms, letting operators focus on verification and final approval.
Task automatability	claude-haiku-4-5-20251001	4/5	Most compliance, operational, and safety forms are structured, rule-based documents with clear fields and data sources. Current AI systems can extract relevant operational data, populate fields, and generate reports with high accuracy, achieving substantial time savings. However, final human review and sign-off remain necessary due to liability and regulatory requirements, preventing full end-to-end automation.
Task automatability	claude-sonnet-5	4/5	Report drafting from structured data logs (readings, incident details) is highly amenable to LLM-based generation and templating, especially with integration to plant data historians.ed forms can be largely auto-populated and drafted, though final review is needed.
Adoption barriers	claude-haiku-4-5-20251001	3/5	While compliance forms must ultimately be signed by authorized operators (creating human-contact friction), the data-gathering and drafting portions face lower barriers. Regulatory oversight of the reporting content itself is high, but automation of form population is not prohibited; human review and authorization remain required gatekeepers.
Adoption barriers	claude-sonnet-5	4/5	Compliance and safety reports often require certified operator sign-off and are subject to regulatory scrutiny (e.g., NERC, OSHA), creating strong liability and authorization barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	AI-driven form and report generation costs are substantially lower than human labor for data entry and compilation once integrated into plant management systems. Inference and oversight costs are minimal relative to the operator time saved on routine paperwork.
Cost vs. human wage	claude-sonnet-5	4/5	Automating routine data compilation and report drafting is far cheaper than paying an operator's time for manual writing, though data integration and validation add some cost.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	Mature automation products (document generation, form-filling, report generation platforms) are widely deployed in industrial and regulated sectors. Systems can reliably extract plant data, map to form templates, and produce standardized reports at scale with minimal error rates on well-structured inputs.
Technical feasibility today	claude-sonnet-5	3/5	Document automation and compliance-reporting software exist and are used in industrial settings, but full end-to-end submission of regulatory safety forms with AI is not yet standard practice at most plants.

Take regulatory action, based on readings from charts, meters and gauges, at established intervals.

CI 38–55 · exposure 58 · augmentation 75 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power generation is a capital-intensive, risk-averse sector with strong regulatory and labor oversight; adoption of autonomous AI for direct regulatory actions remains slow, with pilots far outpacing production deployment.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are traditionally slow-moving, heavily regulated, and risk-averse with long capital cycles, resulting in gradual rather than rapid AI/automation adoption compared to information-sector industries.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI excels at continuously monitoring readings and alerting operators to threshold breaches or recommended actions, significantly enhancing human operator productivity and situational awareness while preserving human final authority.
Augmentation potential	claude-sonnet-5	4/5	AI-enhanced monitoring, anomaly detection, and predictive analytics substantially help operators interpret readings and anticipate issues, improving decision speed and accuracy while humans retain final authority over regulatory actions.
Task automatability	claude-haiku-4-5-20251001	4/5	Current AI systems can reliably read digital gauges, interpret charts, and apply established procedural rules to determine required actions with high accuracy. However, some residual complexity around edge cases and the need for occasional human judgment in novel situations prevents a full 5 rating.
Task automatability	claude-sonnet-5	3/5	Reading sensor data and taking predefined regulatory actions is well within SCADA/DCS automation capability, but full end-to-end automation requires integration with physical control systems and handling edge cases that current AI alone doesn't fully own without human oversight.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Power plant operations are heavily regulated; federal and state law typically mandate that licensed human operators take responsibility for regulatory compliance and critical control actions, creating a hard legal barrier to full automation.
Adoption barriers	claude-sonnet-5	4/5	Power plant operations are heavily regulated (NRC, state utility commissions) with certification requirements for operators and strict liability for safety-critical actions, creating strong barriers to full autonomous replacement.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	Sensor monitoring, data interpretation, and regulatory rule application via AI cost a fraction of a continuously vigilant human operator's loaded wage, especially over 24/7 operations.
Cost vs. human wage	claude-sonnet-5	3/5	Control automation hardware and software have significant upfront capital and maintenance costs comparable to operator staffing costs when amortized, though at scale continuous automated monitoring is cheaper per reading than continuous human monitoring.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Industrial control systems with AI-assisted monitoring exist in production, but full autonomous regulatory action at power plants remains limited due to stringent safety and liability requirements; most deployed systems flag actions for human approval rather than execute independently.
Technical feasibility today	claude-sonnet-5	3/5	Automated control systems (DCS/SCADA with control logic) already perform much of this monitoring and adjustment in modern plants, but these are traditional control systems rather than general AI, and human operators remain in the loop for regulatory compliance and exception handling.

Regulate equipment operations and conditions, such as water levels, based on instrument data or from computers.

CI 34–55 · exposure 50 · augmentation 75 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power generation is a heavily regulated, risk-averse sector with long equipment lifespans and entrenched operational procedures. Although industrial automation has advanced, autonomous AI operation of critical equipment faces slow adoption due to safety culture, regulatory constraints, and the mature operator workforce.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and power generation are traditionally slow-adopting, capital-intensive, and safety-conservative sectors, with automation upgrades happening incrementally rather than through rapid AI-driven transformation.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI excels at real-time monitoring, alerting operators to out-of-spec conditions, and suggesting corrective actions based on instrument data. Operators using AI-assisted dashboards and anomaly detection systems can manage more equipment or respond faster, significantly raising their productivity while maintaining human oversight of safety-critical decisions.
Augmentation potential	claude-sonnet-5	4/5	AI-enhanced monitoring, predictive analytics, and anomaly detection tools significantly aid operators in interpreting instrument data and anticipating equipment issues, improving decision quality while humans remain in control.
Task automatability	claude-haiku-4-5-20251001	4/5	Current AI systems can monitor instrument data and computer readouts to identify when water levels or other parameters deviate from setpoints, and can execute routine regulatory adjustments with high consistency. However, complex anomalies, equipment failures, or interactions requiring real-time judgment of safety-critical conditions still benefit from human oversight, limiting it from a perfect 5.
Task automatability	claude-sonnet-5	2/5	While SCADA and control systems already automate much routine regulation, the task as a broad operator responsibility—monitoring, judgment under abnormal conditions, and integrating multiple instrument feeds—still requires human oversight for safety-critical decisions beyond simple control loop logic.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulation of power plant equipment is subject to strict licensing requirements (operators must hold certifications), regulatory oversight (NERC, NRC, state utility commissions), and liability frameworks that mandate human operator responsibility and sign-off. These legal and safety barriers significantly limit full substitution without explicit regulatory change.
Adoption barriers	claude-sonnet-5	4/5	Power plants are heavily regulated, often requiring licensed operators to be present and accountable for safety-critical equipment adjustments, creating strong liability and regulatory barriers to full removal of human oversight.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	AI systems for equipment monitoring and control have very low marginal inference costs compared to the fully-loaded wages of power plant operators (often $60k–$100k+ annually). Once integrated, automated regulation operates continuously at minimal cost relative to human labor.
Cost vs. human wage	claude-sonnet-5	3/5	Control systems are already embedded in plant capital costs and are cost-effective for routine regulation, but the human operator role persists for oversight, so net cost savings from further AI are moderate rather than transformative.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	SCADA and automation products exist in power plants today and can handle basic parameter regulation, but they typically operate under strict human supervision and require certified operators to validate decisions. Fully autonomous AI regulation of critical equipment remains largely supervised or in pilot phases rather than fully deployed at scale.
Technical feasibility today	claude-sonnet-5	3/5	Automated control systems (DCS/SCADA, PID controllers) reliably regulate many parameters like water levels in production plants today, but full closed-loop autonomous operation without human operators is not standard due to safety and regulatory requirements.

Control generator output to match the phase, frequency, or voltage of electricity supplied to panels.

CI 25–41 · exposure 42 · augmentation 63 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	The power sector has adopted supervisory control and data acquisition (SCADA) for decades, but true end-to-end autonomous control adoption remains limited due to regulatory requirements and risk aversion in critical infrastructure, placing it in the 'pilot and incremental' rather than 'deep production' category.
Sector adoption velocity	claude-sonnet-5	2/5	Power generation is a highly regulated, capital-intensive, physical-infrastructure sector with slow technology adoption cycles and long asset lifespans, unlike fast-moving information-sector industries.
Augmentation potential	claude-haiku-4-5-20251001	4/5	Real-time monitoring dashboards, predictive alerts for frequency drift, and automated load-balancing suggestions substantially enhance operator productivity and situational awareness, enabling faster response to grid disturbances while the operator retains final control authority.
Augmentation potential	claude-sonnet-5	3/5	AI-based predictive analytics and anomaly detection can help operators anticipate synchronization issues and optimize output parameters, improving decision support without removing the operator from direct control.
Task automatability	claude-haiku-4-5-20251001	3/5	Modern power plants use automated control systems that can regulate generator output for phase, frequency, and voltage matching, but human operators currently remain in the loop for decision-making, load changes, and fault response—achieving roughly half the labor content reduction rather than full end-to-end automation at equal quality.
Task automatability	claude-sonnet-5	2/5	This is a real-time control task with safety-critical feedback loops; while automatic synchronization and voltage regulation systems have long existed, full end-to-end AI replacement of operator judgment and oversight is not yet standard for the broader task scope described.svg
Adoption barriers	claude-haiku-4-5-20251001	4/5	Strong regulatory and safety barriers exist: power plant operators must be licensed, grid codes require certified personnel to approve synchronization, and liability for frequency/voltage mismatches falls on human operators, making full autonomous control without human sign-off legally and operationally prohibited.
Adoption barriers	claude-sonnet-5	5/5	Grid interconnection and safety regulations typically require certified, licensed operators to monitor and control generator synchronization and output, given the severe consequences of equipment damage or grid instability from errors.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Automation infrastructure (sensors, controllers, integration) for power plants is capital-intensive and still requires significant operator wages for monitoring and intervention, making the all-in cost per task-equivalent roughly comparable to or slightly higher than human operation alone.
Cost vs. human wage	claude-sonnet-5	2/5	Existing automated control hardware already handles much of this at low marginal cost, but replacing the operator's oversight role entirely with AI would require redundant safety systems, certification, and monitoring that add significant cost relative to current staffing levels.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed SCADA and real-time control systems exist and perform routine synchronization tasks reliably in production plants, but material constraints remain around handling edge cases, emergency transitions, and the requirement for licensed human oversight—limiting deployment to assisted control rather than fully autonomous operation.
Technical feasibility today	claude-sonnet-5	3/5	Automatic synchronizers, AVRs, and SCADA-based control systems are mature and widely deployed in power plants, but they are traditional control engineering rather than AI, and human operators remain in the loop for exception handling and final authority.

Monitor power plant equipment and indicators to detect evidence of operating problems.

CI 25–39 · exposure 42 · augmentation 75 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption of AI-driven monitoring in power plants is slow and cautious due to regulatory mandates, legacy equipment, risk-averse organizational culture, and the critical role of human operators in compliance. Pilots exist but production deployment at scale remains limited.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are traditionally slow adopters of new digital tools due to safety-critical nature, legacy infrastructure, and regulatory caution, though some predictive maintenance pilots exist.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI systems currently assist operators by filtering noise from sensor streams, prioritizing alerts, and suggesting diagnostic hypotheses, significantly raising operator situational awareness and response speed while the operator remains the decision-maker.
Augmentation potential	claude-sonnet-5	4/5	AI-driven anomaly detection, predictive maintenance dashboards, and alarm prioritization significantly help operators catch problems earlier and reduce cognitive load while they remain in control.
Task automatability	claude-haiku-4-5-20251001	3/5	Current AI systems can detect anomalies in sensor data and alert operators to deviations from normal parameters, automating roughly half the monitoring workload. However, contextual judgment about equipment degradation, interactive troubleshooting, and response prioritization still require human expertise, preventing full end-to-end automation at the required quality threshold.
Task automatability	claude-sonnet-5	2/5	Sensor-based monitoring and anomaly detection software can flag many issues, but continuous, holistic situational monitoring with physical presence and judgment across diverse plant systems still requires human operators; only partial automation achievable today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Power plant operation is heavily regulated (NRC, NERC, state energy commissions) with explicit licensing requirements for operators to monitor and respond to equipment issues; automated systems cannot legally sign off on safety-critical detections without licensed human validation.
Adoption barriers	claude-sonnet-5	4/5	Power plants are heavily regulated critical infrastructure requiring licensed/certified operators on-site for safety and regulatory compliance, creating strong barriers to full substitution.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI monitoring infrastructure requires substantial upfront investment in sensor integration, model training, and maintenance, making the per-task cost comparable to or slightly higher than a human operator's loaded wage when amortized across a power plant's actual alert frequency.
Cost vs. human wage	claude-sonnet-5	2/5	Monitoring software and sensors have high upfront integration and maintenance costs, and human oversight is still required for safety-critical decisions, keeping all-in costs comparable to or higher than labor savings suggest.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Monitoring systems with AI-driven anomaly detection exist in production at some utilities, but they typically flag alerts with material false-positive rates and require skilled human validation. No system reliably performs independent diagnosis of complex multi-system failures without human oversight.
Technical feasibility today	claude-sonnet-5	3/5	SCADA/DCS systems with alarm management and predictive analytics are deployed widely in power plants, but they augment rather than replace operator monitoring, and false positives/negatives remain a real issue.

Verify that well field monitoring data conforms to applicable regulations.

CI 25–37 · exposure 33 · augmentation 63 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power generation and well operations are moderately digitized but adoption of autonomous compliance verification remains limited; most utilities and operators are still in early pilot phases with rule-based compliance assists rather than full AI automation.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and power generation are traditionally slower adopters of AI compared to information/finance sectors, with compliance-critical monitoring functions particularly cautious about automation.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-assisted dashboards and automated anomaly detection can help operators spot deviations faster and organize large data streams, improving verification speed, but the operator must still interpret context and take responsibility for the final compliance attestation.
Augmentation potential	claude-sonnet-5	4/5	AI can strongly assist by flagging anomalies, cross-referencing regulatory thresholds, and pre-screening large data sets, significantly speeding up the human verification process.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can parse structured monitoring data and compare it against regulatory thresholds, the task requires contextual judgment about geothermal or hydrocarbon well conditions, interpretation of edge cases, and sign-off accountability that current systems cannot reliably perform end-to-end. Partial automation of threshold-checking is feasible, but full task automation with 50% time savings at equal quality is not demonstrated today.
Task automatability	claude-sonnet-5	3/5	Data conformance checking against regulatory thresholds is a rules-based comparison task that AI/software can largely automate, though interpretation of ambiguous regulatory language and edge cases still needs human review.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory frameworks (EPA, state geothermal/petroleum rules) typically require a licensed operator to verify and certify compliance; liability for false negatives is severe, and autonomous automation without explicit human sign-off faces legal and certification barriers that protect the role.
Adoption barriers	claude-sonnet-5	4/5	Regulatory compliance verification often requires accountable, sometimes certified personnel to attest to conformance, creating liability and legal sign-off barriers that resist full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Current compliance verification tools (SCADA integration, automated reporting) cost significant setup and licensing; the overhead of integration, data validation, and required human oversight often approaches or exceeds the cost of a technician performing manual spot-checks and verification.
Cost vs. human wage	claude-sonnet-5	3/5	Automated data checks can be cheap to run, but building and maintaining a compliant, auditable rules engine plus human oversight for regulatory sign-off keeps overall cost roughly comparable to trained technician review.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No production systems reliably verify well field compliance autonomously; rule-based compliance checkers exist but require human validation of anomalies and regulatory interpretation. Deployed tools assist with data aggregation and alerting, but verification and sign-off remain manual across the industry.
Technical feasibility today	claude-sonnet-5	2/5	Automated data validation and SCADA-integrated monitoring systems exist, but purpose-built products that verify well field data against evolving regulatory compliance rules with full reliability in production are limited and often custom-built.

Monitor well fields periodically to ensure proper functioning and performance.

CI 25–32 · exposure 25 · augmentation 75 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Power utilities have digitized monitoring extensively (SCADA, remote sensors), but actual displacement of periodic inspection roles has been slow; automation remains in the augmentation phase rather than replacement. Industry conservatism and regulatory constraints limit fast adoption of autonomous decision-making.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and power generation are traditionally slow adopters of AI-driven monitoring compared to information/finance sectors, with pilots more common than widespread production deployment.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI-driven dashboards, predictive maintenance alerts, and anomaly detection substantially assist operators in monitoring efficiency and detecting problems faster. These tools keep humans in the loop while significantly raising their diagnostic speed and coverage per shift.
Augmentation potential	claude-sonnet-5	4/5	AI-driven predictive maintenance and anomaly detection tools can significantly aid operators in identifying issues faster and prioritizing inspection routes, meaningfully boosting productivity while humans remain responsible for verification.
Task automatability	claude-haiku-4-5-20251001	2/5	Routine monitoring of gauges and sensors could be partially automated through telemetry systems, but periodic inspection and troubleshooting require on-site judgment and physical presence. Current AI cannot perform the full end-to-end task including problem diagnosis and corrective actions with 50% time savings at equal quality.
Task automatability	claude-sonnet-5	2/5	Physical inspection and monitoring of dispersed well field equipment requires sensor deployment and physical presence; AI can assist with data analysis but cannot fully replace periodic on-site or remote monitoring end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory frameworks (NRC, FERC) and safety standards require licensed operators to certify plant status and take corrective action. Liability asymmetry is high: sensor misinterpretation could cause catastrophic failure, creating a hard barrier to full automation without human sign-off.
Adoption barriers	claude-sonnet-5	3/5	Safety-critical infrastructure monitoring often requires human oversight and accountability for regulatory compliance, though not always a licensed sign-off specifically for this subtask.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Telemetry and monitoring software are cost-effective, but the loaded cost of a power plant operator wage remains competitive with the infrastructure and oversight required for autonomous monitoring systems that must maintain safety and reliability standards.
Cost vs. human wage	claude-sonnet-5	2/5	Sensor networks, telemetry infrastructure, and analytics software require significant capital and maintenance costs that are not clearly cheaper than having an operator periodically check systems, especially at smaller plants.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	SCADA systems and automated monitoring exist in production, but they typically alert humans rather than autonomously ensure proper functioning. Deployed products handle data collection reliably but not the full interpretation, judgment, and corrective action cycle that 'ensure proper functioning' implies.
Technical feasibility today	claude-sonnet-5	2/5	SCADA and remote monitoring systems with anomaly detection exist in some facilities, but comprehensive autonomous well field monitoring products are narrow and not universally deployed at scale.

Inspect records or log book entries or communicate with plant personnel to assess equipment operating status.

CI 25–30 · exposure 30 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power generation is a capital-intensive, heavily regulated, risk-averse sector with entrenched legacy systems and workforce. Adoption of AI for operational assessment is slow; pilots exist but production deployment remains limited. The sector lags compared to finance or tech.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and industrial control sectors are traditionally slow adopters of AI due to safety-critical infrastructure, legacy systems, and regulatory caution, with pilots more common than deployment.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist operators by highlighting logbook anomalies, summarizing recent entries, or flagging deviations from normal ranges, raising productivity for human review. However, augmentation is confined to information presentation rather than autonomous decision-making, as the human operator retains responsibility for assessment.
Augmentation potential	claude-sonnet-5	3/5	AI-based anomaly detection, log summarization, and predictive maintenance tools can meaningfully assist operators in interpreting data and flagging issues faster, even though humans stay firmly in the loop.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can read and summarize logbook entries or structured operational data, the task requires assessing equipment status by integrating multiple information sources (records, communication with personnel, contextual judgment) and making decisions that typically demand human expertise. Current systems lack the reliability to replace this judgment-dependent assessment end-to-end.
Task automatability	claude-sonnet-5	2/5	Reviewing logs and communicating with staff to assess status involves synthesizing informal, multimodal, real-time information and building situational trust, which current AI can partially assist but not fully replace end-to-end.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power plant operations are heavily regulated (FERC, NRC, state utility rules) and often require licensed operators to sign off on equipment status assessments. Liability asymmetry—a missed equipment failure can cause safety, environmental, or financial harm—creates strong organizational and legal friction against full substitution.
Adoption barriers	claude-sonnet-5	4/5	Power plant operations are heavily regulated and require certified operators to make final safety-critical assessments, creating strong liability and licensing barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	An AI system for logbook parsing and anomaly detection costs less per query than labor once deployed, but integration with plant systems, ongoing oversight, and the need for human verification to validate findings narrow the cost advantage. Not yet an order of magnitude cheaper when accounting for setup and oversight.
Cost vs. human wage	claude-sonnet-5	2/5	Sensor/log analytics can be cheap to run, but the human communication and judgment component still requires operator time, so overall automation cost savings are modest given integration and oversight needs.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Products exist to parse and flag anomalies in logbook data and sensor logs, and chatbots can conduct basic information gathering, but no deployed system reliably assesses the full operational status of complex power plant equipment without human verification. Error rates in edge cases remain material.
Technical feasibility today	claude-sonnet-5	2/5	Some monitoring dashboards and anomaly-detection tools exist in industrial settings, but integrating log review with live personnel communication for status assessment is not a mature deployed product.

Operate landfill gas, methane, or natural gas fueled electrical generation systems.

CI 25–25 · exposure 25 · augmentation 50 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power utilities are moderately digitized and risk-averse; adoption of AI-only operation remains rare in production. While remote monitoring and predictive maintenance pilots are common, actual autonomous operation of generation units has not achieved meaningful deployment velocity.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and industrial energy sectors are conservative adopters of full autonomy in safety-critical physical operations, though control-system automation has existed for decades; genuine AI-driven autonomous operation is still nascent.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI systems today can usefully augment operators by flagging anomalies in sensor data, predicting maintenance needs, and automating routine logging; these assistive functions improve situational awareness without replacing human judgment in critical decisions.
Augmentation potential	claude-sonnet-5	3/5	AI-based monitoring, predictive maintenance, and anomaly detection tools can meaningfully assist operators in real-time decision-making and efficiency, though the operator remains essential for control actions.
Task automatability	claude-haiku-4-5-20251001	2/5	Operating electrical generation systems requires continuous monitoring, real-time decision-making under fault conditions, and physical intervention (valve adjustments, emergency shutdowns) that current AI cannot perform end-to-end. While AI can assist with sensor data interpretation and predictive maintenance, human operators remain essential for safe, reliable operation.
Task automatability	claude-sonnet-5	2/5	Physical monitoring and control of gas-fueled generation equipment requires on-site presence, sensor interpretation, and manual intervention that current AI cannot fully replace end-to-end.can be partially automated via SCADA/control systems, but not eliminated.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory agencies (EPA, FERC, state utility commissions) mandate human operator licensing and on-site presence for electrical generation facilities. Liability and safety standards legally require credentialed human oversight, creating hard adoption barriers.
Adoption barriers	claude-sonnet-5	4/5	Power plant operation is subject to safety regulations, licensing, and liability requirements; many jurisdictions require certified operators or supervisory sign-off for generation systems, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Integration of AI-assisted monitoring into existing control systems requires significant upfront engineering, cybersecurity investment, and ongoing validation. The all-in cost (infrastructure, integration, liability insurance) remains comparable to or exceeds the loaded salary of a single operator.
Cost vs. human wage	claude-sonnet-5	2/5	Control automation software is cheap to run, but the task requires physical presence and safety oversight, meaning human operators still carry substantial cost that AI does not fully displace.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	SCADA systems and industrial automation exist but do not autonomously operate generation systems without human oversight; current deployments require operators to supervise and intervene. No mature product reliably operates these systems fully autonomously in production environments.
Technical feasibility today	claude-sonnet-5	2/5	Automated control systems and SCADA exist and are widely deployed, but full autonomous operation without human operators is not standard in production; humans remain required for safety-critical decisions and physical response.

Diagnose or troubleshoot problems with gas collection systems.

CI 25–25 · exposure 25 · augmentation 50 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power generation is a capital-intensive, heavily regulated, and conservative sector. While utilities pilot predictive maintenance and remote monitoring, adoption of AI-driven autonomous diagnostics for critical safety systems remains limited; most plants still rely on certified human operators.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and power generation are traditionally slower-adopting, capital-intensive, physically-oriented sectors where AI pilots for predictive maintenance exist but production-scale autonomous diagnosis is rare.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI augmentation is plausible: sensor dashboards, anomaly alerts, and failure-mode suggestions could assist an operator in narrowing the diagnosis space and speeding investigation. Such tools are emerging in industrial settings but are not yet transformative for this specific task.
Augmentation potential	claude-sonnet-5	3/5	AI-driven sensor analytics and anomaly detection can meaningfully assist operators by flagging irregularities and suggesting likely fault areas, improving speed and accuracy while the operator still performs physical diagnosis and verification.
Task automatability	claude-haiku-4-5-20251001	2/5	Diagnosing gas collection systems requires real-time sensor interpretation, physical inspection capabilities, and domain-specific reasoning about complex mechanical failures. While AI can analyze sensor data and suggest common causes, end-to-end troubleshooting still demands human expertise to physically inspect equipment, perform tests, and make final judgments—falling short of the 50% time-saving bar for full automation.
Task automatability	claude-sonnet-5	2/5	Diagnosing physical gas collection system faults requires sensor inspection, field observation, and hands-on troubleshooting that current AI cannot fully replicate end-to-end, though anomaly detection on sensor data can assist part of the diagnostic process.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power generation operates under strict federal and state regulations (EPA, OSHA, NRC where applicable); only licensed operators with specific certifications can perform certain diagnostic and corrective actions on critical safety systems. Legal and liability requirements create hard barriers to autonomous substitution.
Adoption barriers	claude-sonnet-5	4/5	Power plants are heavily regulated, safety-critical environments where licensed/certified operators are typically required to sign off on troubleshooting decisions involving hazardous gas systems, creating strong liability and regulatory barriers.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	A power plant operator's loaded wage is substantial (~$80k–$100k annually). Current AI diagnostic tools require significant integration, specialized domain training, and human oversight; their all-in cost per troubleshooting event approaches or exceeds the human operator's marginal labor cost for this task.
Cost vs. human wage	claude-sonnet-5	2/5	Sensor analytics platforms exist but require significant integration, calibration, and human oversight, so all-in costs remain comparable to or higher than relying on experienced operators for full diagnosis.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Some diagnostic support tools exist (sensor monitoring, anomaly detection), but no deployed product reliably performs autonomous troubleshooting of gas collection systems at scale. The task requires integration with specialized industrial equipment and real-time physical feedback that current AI systems cannot independently gather or act upon.
Technical feasibility today	claude-sonnet-5	2/5	Some predictive maintenance and anomaly-detection products exist for industrial gas/process systems, but they are narrow-scope decision-support tools rather than autonomous diagnosticians deployed at scale for this specific task.

Operate, control, or monitor equipment, such as acid or gas carbon dioxide removal units, carbon dioxide compressors, or pipelines, to capture, store, or transport carbon dioxide exhaust.

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

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power plant operations remain highly regulated, capital-intensive, and risk-averse sectors with slow technology adoption cycles. While remote monitoring and decision-support tools are gradually adopted, displacement of human operators through AI remains minimal and restricted to non-safety-critical observation tasks.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and heavy industry are traditionally slow adopters of AI for physical control systems, with pilots for predictive maintenance more common than full autonomous control deployment.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist operators through real-time monitoring dashboards, predictive maintenance alerts, and data analysis of equipment performance trends. Such augmentation helps operators make faster, more informed decisions about equipment control and can improve efficiency, though operators remain firmly in the decision-making loop.
Augmentation potential	claude-sonnet-5	3/5	AI-based predictive analytics and anomaly detection can meaningfully assist operators in monitoring equipment health and optimizing carbon capture processes, though the operator remains essential for control and response.
Task automatability	claude-haiku-4-5-20251001	2/5	While monitoring and data collection could be partially automated, the task requires real-time decision-making, safety oversight, and response to equipment anomalies in complex industrial systems. Current AI cannot reliably handle the full scope of equipment control, fault diagnosis, and safety-critical interventions that human operators must perform.
Task automatability	claude-sonnet-5	2/5	While sensor monitoring and anomaly detection can be assisted by AI, the physical operation and control of carbon capture equipment involves real-time safety-critical decisions and physical interventions that current AI cannot fully replace end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power plant operations are heavily regulated by environmental and safety standards (EPA, OSHA); operators must be licensed or certified; equipment failures carry high consequence costs and liability; and regulatory frameworks explicitly require human accountability for hazardous industrial processes. These create strong legal and compliance barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Power plant operations are heavily regulated with safety certifications, licensed operator requirements, and liability concerns around emissions and equipment failure, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Integration of AI oversight systems would require significant capital investment in sensors, connectivity, and safety-critical validation. The cost of these systems plus necessary human supervision and liability coverage would likely exceed the loaded cost of experienced power plant operators.
Cost vs. human wage	claude-sonnet-5	2/5	AI monitoring software is cheap to run, but the required sensor infrastructure, integration with legacy plant control systems, and mandatory human oversight keep total costs comparable to or only modestly below human operator costs.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Specialized SCADA and industrial control systems exist for power plant operations, but they are purpose-built for specific plants and require extensive customization. General-purpose AI lacks the domain specificity and safety certifications needed for autonomous operation of carbon dioxide capture and transport equipment in production environments.
Technical feasibility today	claude-sonnet-5	2/5	Some SCADA/AI-assisted monitoring systems exist in industrial plants, but fully autonomous operation of carbon capture and compression equipment is not yet deployed reliably in production at scale.

Operate, control, or monitor integrated gasification combined cycle (IGCC) or related equipment, such as air separation units, to generate electricity from coal.

CI 25–25 · exposure 25 · augmentation 50 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power generation is capital-intensive, risk-averse, and heavily regulated; plants operate 20–40 years and upgrade cautiously. AI agent deployment in production is rare; most sites use legacy SCADA systems and human operators, with only incremental sensor analytics adoption.
Sector adoption velocity	claude-sonnet-5	2/5	Energy generation, especially coal-based IGCC, is a slow-adopting, capital-intensive, physically-oriented sector with limited digitization compared to information-based industries.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist with anomaly detection, predictive maintenance alerts, and trend analysis, improving operator situational awareness and reducing cognitive load. However, augmentation is limited to advisory functions; the operator retains full control and responsibility.
Augmentation potential	claude-sonnet-5	3/5	AI-based predictive maintenance, anomaly detection, and decision-support tools can meaningfully assist operators in monitoring plant performance and detecting faults earlier.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can monitor sensor data and flag anomalies, the task requires real-time control decisions, failure response, and safety-critical adjustments that demand domain expertise and accountability. Current systems can assist but cannot reliably operate the full end-to-end process with 50% time savings at equal quality.
Task automatability	claude-sonnet-5	2/5	Monitoring integrated gasification and air separation equipment involves complex physical process control, anomaly response, and safety-critical judgment that current AI cannot fully replace end-to-end.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Nuclear Regulatory Commission and FERC oversight apply to many plants; operators hold licensed certifications and must legally sign off on operational decisions. Liability for equipment damage, safety failures, and grid reliability creates strong legal and contractual requirements for human accountability.
Adoption barriers	claude-sonnet-5	4/5	Power plant operation is heavily regulated, requires licensed/certified operators, and carries high liability for errors involving safety and environmental compliance, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	A power plant operator earns $70–$90k+ annually with full benefits. AI monitoring and partial automation systems cost millions to integrate per site and require ongoing maintenance, tuning, and human oversight, making the all-in cost comparable or higher than human operation.
Cost vs. human wage	claude-sonnet-5	2/5	Specialized industrial control systems and sensors require significant capital and integration costs, and human operators are still needed for oversight, making cost savings modest at best.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Industrial control systems exist but are purpose-built for specific plants, not general-purpose AI solutions. No off-the-shelf AI product reliably operates IGCC equipment in production; deployed automation is narrow, rule-based, and requires human operators in the loop for critical decisions.
Technical feasibility today	claude-sonnet-5	2/5	Advanced process control and predictive analytics software exist in industrial settings, but full autonomous operation of IGCC plants by AI is not deployed; humans remain the primary operators.

Receive outage calls and request necessary personnel during power outages or emergencies.

CI 25–25 · exposure 25 · augmentation 63 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Utility operations are traditionally risk-averse and heavily regulated; digital transformation in power grids is slow and cautious. Most utilities still rely on human dispatchers for emergency calls, with AI adoption limited to non-critical back-office functions; few production deployments exist in this specific emergency-response context.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are a traditionally slow-adopting, highly regulated, safety-critical sector where AI deployment in core operational decision-making is still nascent.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist by auto-logging call details, suggesting personnel based on skill/availability, and flagging critical system states, raising dispatcher efficiency. However, the human must remain in the loop for actual emergency decisions, personnel deployment authorization, and real-time communication during outages.
Augmentation potential	claude-sonnet-5	4/5	AI-based outage management systems, automated call routing, and predictive alerts can substantially speed up identification of outage scope and personnel needs while humans retain final control.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI could handle basic call intake and routing, the task requires real-time judgment about emergency severity, personnel availability, and dynamic decision-making during high-stakes outages. Current systems lack the reliability needed to make autonomous dispatching decisions without substantial human oversight, meaning time savings fall short of the 50% threshold.
Task automatability	claude-sonnet-5	2/5	AI can help triage and log outage calls and trigger notification workflows, but judgment about severity, resource allocation, and personnel dispatch during emergencies still requires human decision-making and accountability.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power utilities face regulatory oversight (NERC standards, state PUC regulations) and implicit legal liability for outage response decisions; there is no explicit licensing requirement on dispatchers themselves, but organizational and operational risk aversion is high. Safety-critical nature and potential liability for incorrect personnel dispatch create strong adoption friction.
Adoption barriers	claude-sonnet-5	4/5	Power grid emergency response is heavily regulated with certified operator requirements and strict liability for outage handling, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI systems for emergency call handling and dispatch routing require significant custom integration, real-time data linkage to personnel systems, and continuous human oversight. Setup and operational costs are comparable to or exceed the loaded wage of a single dispatcher, especially when factoring in error consequences.
Cost vs. human wage	claude-sonnet-5	2/5	Emergency response systems still require human operators on standby for safety-critical judgment, so AI mainly supplements rather than replaces the labor cost, keeping ratios comparable rather than dramatically cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably handles emergency call receipt and personnel dispatch autonomously in production power systems. Call centers use basic IVR and ticketing, but actual emergency triage and personnel requisitioning remain human-driven due to safety-critical requirements and the need to interpret context and urgency.
Technical feasibility today	claude-sonnet-5	2/5	Outage management systems with automated call intake and alerting exist, but full autonomous handling of emergency personnel requests is not standard in deployed control room systems.

Open and close valves and switches in sequence to start or shut down auxiliary units.

CI 18–29 · exposure 25 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation is a heavily regulated, conservative sector with multi-year validation timelines; adoption of autonomous AI for critical operational sequences is extremely slow due to safety culture, regulatory oversight, and reluctance to remove human operators from control.
Sector adoption velocity	claude-sonnet-5	2/5	Power generation is a highly regulated, capital-intensive, physical-infrastructure sector with slow technology adoption cycles compared to information-based industries.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-assisted dashboards and automated alerts can help operators manage complex sequences more efficiently, and SCADA systems already provide significant augmentation; however, the core task of intentional sequencing remains human-directed with digital support rather than AI-driven transformation.
Augmentation potential	claude-sonnet-5	3/5	AI-enabled monitoring and predictive systems can assist operators by flagging optimal sequencing timing or anomalies, improving situational awareness without replacing the person executing critical switch actions.
Task automatability	claude-haiku-4-5-20251001	2/5	While individual valve and switch operations can be automated, the sequencing requirement and need to respond to real-time plant conditions and safety interlocks demand human oversight; current AI cannot reliably handle the full closed-loop control and exception handling required in a live power plant.
Task automatability	claude-sonnet-5	2/5	This is a physical sequencing/control action tied to real equipment; while control logic could be scripted, current general AI systems cannot physically or reliably autonomously execute plant valve/switch operations end-to-end without dedicated industrial control integration.rating
Adoption barriers	claude-haiku-4-5-20251001	5/5	Power plant operations are heavily regulated (NRC, FERC, NERC) with strict licensing requirements for operators; federal law and industry standards mandate that licensed operators must directly supervise or perform critical auxiliary unit control, creating hard legal barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Safety regulations, licensing requirements for plant operators, and liability for equipment damage or safety incidents strongly favor human oversight and sign-off on critical sequencing operations.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Integration of AI-driven automation into a power plant's control infrastructure requires significant engineering, validation, and redundancy costs that may exceed the wage savings from removing one operator's sequencing task.
Cost vs. human wage	claude-sonnet-5	3/5	Where DCS automation already exists, marginal cost of automated sequencing is very low, but retrofitting older plants with AI-driven control requires significant capital and engineering investment, keeping overall cost ratio moderate rather than clearly cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Some industrial automation systems can execute pre-programmed valve sequences, but deploying autonomous AI for critical auxiliary unit startup/shutdown in regulated power plants is not standard practice; most real-world operations still rely on manual operator control with SCADA assistance.
Technical feasibility today	claude-sonnet-5	2/5	Distributed control systems (DCS) and SCADA already automate many startup/shutdown sequences in modern plants, but this is decades-old industrial automation, not generally-available AI products, and many plants still require operator execution or oversight for these sequences.

Analyze the layout, instrumentation, or function of electrical generation or transmission facilities.

CI 20–25 · exposure 25 · augmentation 50 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power plant operations are laggard sectors in AI adoption due to safety-critical nature, regulatory constraints, and organizational conservatism. Pilots exist but production deployment of autonomous analysis remains rare across the industry.
Sector adoption velocity	claude-sonnet-5	2/5	Energy sector adoption of AI for engineering analysis is slow due to safety-critical nature, legacy infrastructure, and regulatory caution, with pilots more common than production deployment.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist operators by rapidly parsing schematics, flagging instrumentation anomalies, or summarizing facility documentation, raising their efficiency in locating and understanding complex data. However, the human operator remains the decision-maker and bottleneck for critical judgments.
Augmentation potential	claude-sonnet-5	3/5	AI can help by analyzing sensor data, flagging anomalies, or assisting in interpreting technical documentation, providing moderate productivity gains while humans retain final analytical judgment.
Task automatability	claude-haiku-4-5-20251001	2/5	Current AI can extract and summarize facility schematics and basic instrumentation data, but analyzing complex interactions between electrical systems, detecting anomalies, and making decisions about facility function requires domain expertise and real-time judgment that AI cannot reliably provide end-to-end with 50% time savings at equal quality today.
Task automatability	claude-sonnet-5	2/5	AI can assist in analyzing schematics or data patterns, but comprehensive analysis of plant layout, instrumentation, and function requires physical inspection, tacit knowledge, and integration with proprietary control systems that current AI cannot fully replicate end-to-end.》Automatability limited. reionalnnotused
Adoption barriers	claude-haiku-4-5-20251001	5/5	Power generation is heavily regulated (NERC, FERC, NRC standards); operators must be licensed, and critical infrastructure security and safety requirements mandate human verification and sign-off on facility analysis. Automation of this task faces hard regulatory and liability barriers.
Adoption barriers	claude-sonnet-5	4/5	Power plant operations are heavily regulated with licensing and safety-critical sign-off requirements, meaning human engineers/operators must verify and authorize such analyses.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Integration of AI tools into critical power infrastructure requires substantial overhead for validation, testing, and cybersecurity. The loaded cost of specialized AI systems plus required human oversight and verification remains comparable to or higher than a trained operator's time spent on the same analysis.
Cost vs. human wage	claude-sonnet-5	2/5	Given the need for specialized engineering oversight, safety certification, and integration costs, AI tools are not clearly cheaper than skilled human operators/engineers for this task today.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Some diagnostic and visualization tools exist in research and narrow commercial contexts (diagram parsing, anomaly detection), but no deployed product reliably performs comprehensive layout analysis and instrumentation interpretation at production scale in actual power plant operations.
Technical feasibility today	claude-sonnet-5	2/5	Some AI-assisted diagnostic and monitoring tools exist in power plants, but full autonomous analysis of layout and instrumentation function is not deployed at scale in production.

Control or maintain auxiliary equipment, such as pumps, fans, compressors, condensers, feedwater heaters, filters, or chlorinators, to supply water, fuel, lubricants, air, or auxiliary power.

CI 20–25 · exposure 25 · augmentation 63 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power generation is a capital-intensive, heavily regulated sector with slow technology adoption cycles and strong labor unions; pilot projects exist but few if any production deployments have replaced operator control over auxiliary equipment at scale. The sector prioritizes operational continuity and regulatory compliance over rapid automation.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are traditionally slow-moving, capital-intensive, and risk-averse industries with long equipment lifecycles, resulting in gradual rather than rapid AI adoption for operational control tasks.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-powered monitoring dashboards, predictive maintenance alerts, and anomaly detection can assist operators in identifying equipment issues and optimizing performance, meaningfully raising situational awareness and response time on this task while the operator retains control.
Augmentation potential	claude-sonnet-5	4/5	AI-driven monitoring, anomaly detection, and predictive maintenance tools significantly help operators anticipate equipment issues and optimize auxiliary system performance while humans remain in control.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI could monitor sensor data and flag anomalies, the task requires real-time physical control decisions, equipment calibration, and response to dynamic system conditions that current AI systems cannot reliably execute end-to-end without substantial human oversight. Autonomous control of critical plant infrastructure meets neither the reliability nor the time-saving threshold.
Task automatability	claude-sonnet-5	2/5	While SCADA/DCS systems automate routine setpoint control, this task involves physical monitoring, manual valve/switch operation, and judgment-based intervention on auxiliary equipment that current AI cannot fully replace end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Regulatory agencies (NRC, EPA, state PSCs) mandate that licensed operators maintain direct control and accountability over critical auxiliary systems; liability for equipment failure or safety incidents falls on certified staff, creating hard legal barriers to full automation. Human sign-off and continuous presence are not preferences but legal requirements.
Adoption barriers	claude-sonnet-5	4/5	Power plants are heavily regulated with mandatory certified/licensed operators, strict safety protocols, and liability concerns that require human oversight and legal accountability for control actions.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The cost of integrating AI control systems, including safety certification, redundancy, liability coverage, and continuous human oversight, currently exceeds the wage cost of a power plant operator performing these duties. Industrial automation requires high upfront capital and ongoing maintenance that does not yet justify the labor replacement economics.
Cost vs. human wage	claude-sonnet-5	2/5	Automation control systems have high upfront capital and integration costs, sensors, and redundancy requirements, while human operators are a comparatively modest ongoing cost, making AI not clearly cheaper for this specific task.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Deployed SCADA and PLC systems can automate routine auxiliary equipment operation in specific, well-defined scenarios, but these are narrow domain-specific tools, not generalizable AI. Current general AI lacks the domain expertise, safety certification, and real-time control capabilities to reliably manage this task independently in production power plants.
Technical feasibility today	claude-sonnet-5	2/5	Industrial control systems and predictive maintenance software exist and are deployed, but full autonomous control of auxiliary equipment without human operators is not standard practice in production plants today.

Start or stop generators, auxiliary pumping equipment, turbines, or other power plant equipment as necessary.

CI 18–25 · exposure 30 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation remains a heavily regulated, traditionally slow-adopting sector. Operator staffing and control authority are mandated by law, and organizational reluctance to cede safety-critical decisions to automation is profound and persistent.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are a traditionally slow-adopting, highly regulated, capital-intensive sector where control automation evolves incrementally rather than through rapid AI-driven transformation.
Augmentation potential	claude-haiku-4-5-20251001	4/5	SCADA systems, predictive maintenance alerts, and AI-driven anomaly detection already significantly assist operators by filtering data, highlighting actionable conditions, and automating routine monitoring, allowing operators to focus on complex decision-making and rare events.
Augmentation potential	claude-sonnet-5	3/5	AI-based predictive analytics and automated control recommendations can assist operators in timing and sequencing equipment starts/stops, improving efficiency while humans retain final control.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI monitoring systems can track operational parameters, starting/stopping critical power plant equipment requires real-time decision-making under variable conditions, safety interlocks, and coordination with grid operators. Current AI cannot reliably handle the full complexity of state assessment and safe equipment sequencing end-to-end, and regulatory frameworks mandate human operator authority.
Task automatability	claude-sonnet-5	2/5	Modern SCADA/DCS systems can automate start/stop sequences under normal conditions, but safe judgment during abnormal states, equipment faults, or emergencies still requires human oversight, limiting full end-to-end automation.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Power plant operators are licensed professionals; federal regulations (FERC, NRC for nuclear) and industry standards (NERC) mandate that qualified human operators maintain direct control and situational awareness. Liability for equipment damage or grid failure creates hard legal and insurance barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Power plant operations are heavily regulated, often require licensed operators, and carry high liability for equipment damage, safety incidents, or grid instability, mandating human presence and sign-off.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI monitoring and alerting systems reduce operational overhead, but the safety-critical nature of equipment control means that supervisory infrastructure and human oversight must remain in place, keeping total cost-to-value ratio unfavorable compared to experienced human operator wages.
Cost vs. human wage	claude-sonnet-5	2/5	Automation control systems are already integrated into plants, but they don't eliminate the need for licensed operators on-site, so incremental AI cost savings are modest relative to retained staffing costs.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Supervisory control systems (SCADA) exist but are decision-support tools; humans retain final authority over equipment commands. No deployed product autonomously initiates major equipment state changes in real power plants without explicit human authorization and override capability.
Technical feasibility today	claude-sonnet-5	3/5	Automated startup/shutdown sequencers and control logic are deployed in many plants, but full autonomous operation without a human operator present is not standard practice due to safety and regulatory requirements.

Operate, control, or monitor gasifiers or related equipment, such as coolers, water quenches, water gas shifts reactors, or sulfur recovery units, to produce syngas or electricity from coal.

CI 18–25 · exposure 25 · augmentation 63 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation and coal gasification sectors are mature, capital-intensive, and highly regulated. Adoption of autonomous control is negligible; pilots may exist but production deployment is rare due to safety mandates, union protections, and the high cost of failure in critical infrastructure.
Sector adoption velocity	claude-sonnet-5	2/5	Power generation is a heavy industrial, highly regulated sector with legacy infrastructure; AI adoption for control functions is mostly in pilot/advisory stages rather than replacing operators in production.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist operators through real-time sensor monitoring, predictive maintenance alerts, and anomaly flagging that help them optimize process parameters and catch issues early. However, the operator remains the decision-maker, and augmentation is limited by legacy control interfaces and the need for human validation before critical actions.
Augmentation potential	claude-sonnet-5	4/5	AI-based predictive analytics, anomaly detection, and digital twin simulations are increasingly used to help operators anticipate equipment issues and optimize gasifier performance, meaningfully boosting operator productivity while humans remain in control.
Task automatability	claude-haiku-4-5-20251001	2/5	While monitoring and data logging could be partially automated, the task requires real-time decision-making in response to equipment anomalies, safety risks, and process upsets that demand human judgment and responsiveness. Current AI systems lack the integrated physical-world reasoning and safety-critical control needed for end-to-end autonomous operation of complex industrial systems.
Task automatability	claude-sonnet-5	2/5	While sensor data monitoring can be partially automated with control systems and AI-assisted anomaly detection, the physical operation, real-time judgment calls, and safety-critical interventions on gasifiers and related equipment require human presence and cannot yet meet the 50% time-saving bar end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory requirements (OSHA, EPA, NRC in some contexts) mandate operator oversight and certification; liability and error-cost asymmetry are severe (equipment damage, environmental discharge, safety incidents). Many jurisdictions legally require a licensed operator to monitor and sign off on critical operations, creating hard adoption barriers.
Adoption barriers	claude-sonnet-5	5/5	Power plant operations are subject to strict regulatory oversight (EPA, OSHA, NERC), require licensed/certified operators, and carry severe liability and safety risks from equipment failure, explosions, or toxic gas release, making full automation a hard barrier.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Implementing AI monitoring and control requires significant upfront capital (sensor networks, integration with SCADA, validation), ongoing maintenance, and human oversight. The loaded cost of a skilled power plant operator is relatively low compared to the cost of AI deployment, integration, and the liability of failures in this safety-critical domain.
Cost vs. human wage	claude-sonnet-5	2/5	Industrial control systems and AI monitoring tools have high integration and validation costs, and the need for continuous human oversight for safety compliance keeps AI from being dramatically cheaper than skilled plant operators.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Deployed systems exist for narrow monitoring and alerting (e.g., sensor data analysis and anomaly detection), but no production AI system reliably operates or controls multi-unit gasifier systems independently. Reliable operation requires integration with legacy control systems, handling of edge cases, and safety certification that remains research-stage or pilot-level.
Technical feasibility today	claude-sonnet-5	2/5	Advanced process control and predictive maintenance software exist in industrial plants, but full autonomous operation of gasification equipment including sulfur recovery and water gas shift reactors is not deployed at scale; humans remain in the control room for critical decisions.

Adjust controls to generate specified electrical power or to regulate the flow of power between generating stations and substations.

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

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power utilities are risk-averse and operate under strict regulatory constraints; adoption of AI for autonomous control remains minimal despite digitization of monitoring systems. Most investment is in human-in-the-loop optimization rather than replacement.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are conservative, highly regulated, and slow to adopt full autonomy in control rooms; automation exists but is incremental and paired with human operators rather than replacing them.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-powered monitoring, forecasting, and decision-support tools can enhance operator efficiency by highlighting anomalies or optimal setpoints, but the human operator remains responsible for executing commands. Assistive potential is moderate given the complexity and stakes.
Augmentation potential	claude-sonnet-5	4/5	Automated control systems, predictive analytics, and decision-support tools significantly aid operators in monitoring and adjusting power flow, improving efficiency and response time while humans remain in charge.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI could theoretically monitor and suggest control adjustments, the task requires real-time decision-making under variable grid conditions with safety-critical consequences. Current AI lacks the integrated control authority and fail-safe mechanisms to autonomously manage power generation/distribution at scale without human oversight.
Task automatability	claude-sonnet-5	2/5	While control adjustments can be partially automated via SCADA/DCS and automatic generation control, safe real-time regulation of grid power involves handling anomalies, equipment quirks, and safety-critical judgment that current AI cannot fully replace end-to-end.'
Adoption barriers	claude-haiku-4-5-20251001	5/5	Electrical grid operation is heavily regulated by NERC, FERC, and other bodies that mandate human operator presence and authority. Operators must be licensed, and liability for grid failures creates strong legal and organizational barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Power grid operation is heavily regulated (NERC/FERC in the US), requires certified operators, and carries severe liability/safety consequences for errors, mandating human oversight and authorization.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The costs of AI systems capable of managing power plant controls (including redundancy, validation, and integration with legacy infrastructure) remain high relative to experienced operator wages, and regulatory compliance adds overhead.
Cost vs. human wage	claude-sonnet-5	2/5	Control automation systems require significant capital investment, specialized integration, and continuous human oversight, so cost savings versus a trained operator are modest rather than order-of-magnitude.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI product reliably operates power plant controls end-to-end in production environments. Utilities use SCADA systems and optimization software that assist operators but retain human control; full autonomous operation remains rare and experimental.
Technical feasibility today	claude-sonnet-5	2/5	Automated generation control and control-room automation exist in production, but full autonomous adjustment without human operators overseeing and intervening is not deployed at scale.

Operate or maintain distributed power generation equipment, including fuel cells or microturbines, to produce energy on-site for manufacturing or other commercial purposes.

CI 7–25 · exposure 13 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Distributed power generation is still a niche application in most sectors, with slow penetration and high capital barriers. Organizations using such systems tend to be risk-averse and highly regulated, leading to cautious adoption of automation.
Sector adoption velocity	claude-sonnet-5	2/5	Power generation and industrial plant operations are a physically-intensive, moderately digitized sector with slow AI adoption for hands-on tasks.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist operators through predictive maintenance alerts, real-time performance monitoring dashboards, and anomaly detection on equipment health. Such augmentation usefully supports human operators but does not transform the core task of making safety-critical operational decisions.
Augmentation potential	claude-sonnet-5	3/5	AI-based monitoring, predictive maintenance, and diagnostic tools can meaningfully assist operators in tracking equipment performance and scheduling maintenance.
Task automatability	claude-haiku-4-5-20251001	2/5	Distributed power generation equipment requires real-time physical monitoring, safety-critical decision-making, and rapid response to equipment faults or load changes. While AI could assist in predictive maintenance and monitoring, current systems cannot reliably perform the full operational loop (load balancing, failover, fuel management, emergency shutdown) end-to-end at the quality and safety level required.
Task automatability	claude-sonnet-5	1/5	Operating and maintaining physical distributed generation equipment requires hands-on manipulation, inspection, and physical intervention that current AI cannot perform end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Heavy regulatory requirements, licensing of power generation operators, liability for safety and grid stability, and insurance requirements all mandate human operator presence and sign-off. Legal standards for distributed generation equipment operation impose hard barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Power generation equipment operation typically requires certified operators, safety compliance, and regulatory oversight, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Full automation would require expensive sensor networks, redundant systems, and continuous AI oversight infrastructure. The operational and integration costs remain high relative to the salary of power plant operators, especially given the small installed base of distributed generation in most industries.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot replace the physical labor and equipment handling involved, so a human operator remains necessary and cost-comparable AI substitution does not exist.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI-based supervisory and monitoring systems exist in research and some pilot deployments, but no production system reliably operates distributed power equipment autonomously without human operators present. Reliability and liability concerns prevent autonomous operation in real industrial settings.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product physically operates or maintains fuel cells or microturbines autonomously; automation here remains at monitoring/advisory research stage.

Examine and test electrical power distribution machinery and equipment, using testing devices.

CI 7–25 · exposure 13 · augmentation 50 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Power utilities are traditionally conservative sectors with slow adoption cycles. While remote monitoring and sensor systems are gradually deployed, the sector remains heavily dependent on human operators for testing and certification due to regulatory requirements and safety criticality.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities and power generation are traditionally slow-adopting sectors for physical automation, though sensor/IoT monitoring adoption is growing steadily but not yet replacing hands-on testing.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-assisted diagnostics through sensor data interpretation and anomaly detection can help operators prioritize maintenance and interpret complex test results, improving their efficiency on the diagnostic aspects of the task. However, the core testing activity remains human-dependent.
Augmentation potential	claude-sonnet-5	3/5	AI-enabled predictive maintenance and diagnostic software can help operators prioritize what to test and interpret sensor data, meaningfully assisting but not replacing the physical testing task.
Task automatability	claude-haiku-4-5-20251001	2/5	While visual inspection and routine equipment testing could be partially automated with sensors and computer vision, the task requires complex judgment about equipment condition, safety diagnostics, and contextual decision-making that current AI systems cannot reliably perform end-to-end. Manual intervention and human oversight remain essential for the majority of testing scenarios.
Task automatability	claude-sonnet-5	1/5	This is a hands-on physical inspection and testing task requiring manual use of testing devices on physical equipment, which current AI cannot perform end-to-end without robotic embodiment.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power generation and distribution is heavily regulated by FERC, NERC, and state utilities commissions, requiring licensed operators to perform and certify equipment testing. Safety-critical infrastructure constraints and liability for equipment failures create strong legal and procedural barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Power plant equipment testing typically requires certified/licensed operators and strict safety protocols due to high liability and risk of catastrophic failure, creating strong regulatory and safety barriers.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Automated monitoring systems require substantial capital investment in specialized sensors, integration with legacy power systems, and ongoing maintenance. The loaded cost of these systems remains comparable to or higher than skilled operator wages when all infrastructure and oversight costs are included.
Cost vs. human wage	claude-sonnet-5	1/5	No viable AI substitute exists for the physical testing act, so cost comparison favors humans by default since AI cannot perform the core physical task.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Some diagnostic sensors and automated monitoring systems exist in production, but they typically support human operators rather than replace them. No deployed AI system reliably performs comprehensive electrical equipment testing and diagnosis without significant human supervision and interpretation.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product physically examines and tests power distribution equipment autonomously; sensor-based monitoring exists but not autonomous hands-on testing with devices.

Control power generating equipment, including boilers, turbines, generators, or reactors, using control boards or semi-automatic equipment.

CI 9–20 · exposure 17 · augmentation 63 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation is a heavily regulated, risk-averse sector with long asset lifespans and entrenched operational practices. Adoption of autonomous AI control of critical equipment is negligible; pilot projects remain rare and deployment is driven by regulatory requirements, not innovation velocity.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are conservative adopters of new control technology due to safety, regulatory approval cycles, and legacy infrastructure, resulting in slow uptake of advanced automation for core control functions.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can usefully assist operators through predictive maintenance alerts, anomaly detection in sensor data, and decision-support dashboards that improve situational awareness. However, augmentation is limited to information enhancement rather than transformative productivity gains, as the core control task remains operator-driven.
Augmentation potential	claude-sonnet-5	4/5	AI-enhanced monitoring, predictive maintenance, and decision-support tools significantly aid operators in anomaly detection and optimization, even though humans remain firmly in control of the equipment.
Task automatability	claude-haiku-4-5-20251001	2/5	Power plant control involves real-time monitoring and quick response to dynamic conditions requiring nuanced judgment about equipment state and safety margins. While AI could assist with routine monitoring and alerts, human operators must retain control authority over critical safety decisions, and the legal/regulatory environment mandates human sign-off, preventing full end-to-end automation with ≥50% time savings.
Task automatability	claude-sonnet-5	1/5	Real-time physical control of high-stakes generating equipment requires embodied monitoring, rapid response to physical anomalies, and safety-critical judgment that current AI cannot perform end-to-end without human presence.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Hard regulatory and legal barriers exist: licensed power plant operators are required by law (NRC, NERC, state PUC rules) to oversee critical control functions, and liability for equipment failure or safety incidents falls on certified human operators. Automation of core control functions faces binding regulatory barriers.
Adoption barriers	claude-sonnet-5	5/5	Power plant operation is heavily regulated, often requiring licensed operators (e.g., NRC certification for nuclear plants), with strict liability and safety oversight mandating human control and sign-off.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Integration costs for AI-assisted power plant control (safety certification, redundancy, oversight infrastructure, liability insurance) would likely approach or exceed the cost of retaining skilled human operators, especially given regulatory compliance overhead.
Cost vs. human wage	claude-sonnet-5	2/5	While automation software exists, the need for redundant safety systems, human oversight, and specialized integration keeps costs comparable to or higher than staffing trained operators for critical infrastructure.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI system today autonomously controls boilers, turbines, generators, or reactors in production power plants. Simulations and narrow control loops exist in research, but real-world deployment would face regulatory barriers and operators retain manual control; feasibility remains below production-ready systems.
Technical feasibility today	claude-sonnet-5	2/5	Advanced control systems and automation already exist in plants, but full autonomous control without human operators overseeing control boards is not deployed at scale due to safety and regulatory requirements.

Communicate with systems operators to regulate and coordinate line voltages and transmission loads and frequencies.

CI 3–25 · exposure 13 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Energy utilities are digitizing operations gradually through smart grids and automated dispatch, but human operators remain embedded in critical decision loops due to safety and regulatory mandates. Adoption of AI coordination is slow relative to information-sector adoption patterns.
Sector adoption velocity	claude-sonnet-5	2/5	Utilities are conservative adopters of AI in core grid operations due to safety-critical nature and legacy infrastructure, though AI is used for forecasting and anomaly detection in adjacent areas.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist operators by providing real-time forecasting of load demand, automated anomaly detection, and suggested parameter adjustments, enhancing situational awareness and decision speed. However, the human operator retains primary responsibility for communication and final coordination decisions.
Augmentation potential	claude-sonnet-5	3/5	AI-based decision support, load forecasting, and anomaly detection tools can assist operators in monitoring and predicting grid conditions, improving their situational awareness during coordination tasks.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can monitor and analyze grid data in real time, the task requires nuanced judgment about load distribution and coordination between multiple human operators across interconnected systems. Current systems can support decision-making but cannot reliably execute end-to-end coordination without human oversight, falling well short of 50% time savings at equal quality.
Task automatability	claude-sonnet-5	1/5	This requires real-time coordination between humans on safety-critical infrastructure decisions, involving judgment calls and accountability that current AI cannot autonomously perform end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Grid operations are heavily regulated by FERC, NERC, and ISO/RTO authorities, which mandate human operator responsibility and accountability for line frequency, voltage, and load balance. Liability for blackouts or system instability creates strong legal and organizational barriers to full automation without human sign-off.
Adoption barriers	claude-sonnet-5	5/5	Power grid operation is heavily regulated (NERC/FERC reliability standards) requiring certified operators, with severe liability for blackouts or equipment damage, making human sign-off legally mandated.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Implementing AI-driven grid coordination requires substantial infrastructure, continuous validation, and human oversight to prevent cascading failures. The cost of integration, monitoring, and fallback systems remains comparable to or exceeds the loaded wage of a power plant operator.
Cost vs. human wage	claude-sonnet-5	1/5	Given the criticality and liability of grid stability, AI cannot yet substitute the human decision-maker, so all-in cost of an AI system replacing this coordination role (with required redundancy and oversight) exceeds current human labor cost.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Deployed SCADA and energy management systems provide decision support and automated controls for narrow, pre-defined scenarios, but no production system reliably handles the full scope of inter-operator communication, negotiation, and dynamic regulation that this task demands. Pilot AI systems exist but lack the real-world reliability needed for independent operation.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously communicates with grid systems operators to regulate voltages and transmission in production; SCADA/automation exists but the interpersonal communication and coordination role remains human-performed.

Collect oil, water, or electrolyte samples for laboratory analysis.

CI 5–23 · exposure 13 · augmentation 25 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation is a capital-intensive, risk-averse sector with slow digital transformation. Adoption of AI-driven or robotic sample collection in production settings is minimal, with most plants continuing traditional human-operator practices due to regulatory conservatism and operational inertia.
Sector adoption velocity	claude-sonnet-5	1/5	Power generation is a heavy industrial, physically-oriented sector with low AI adoption for physical tasks like sample collection; robotics deployment for this specific task is essentially nonexistent.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI can assist by optimizing sampling schedules, flagging which samples to prioritize, or automating data logging and analysis of results. However, the core physical task of collection offers limited augmentation potential for a human already trained in the procedure.
Augmentation potential	claude-sonnet-5	2/5	AI can assist with scheduling sample collection, tracking lab results, or flagging anomalies in returned data, but it does not meaningfully assist with the physical act of sample collection itself.
Task automatability	claude-haiku-4-5-20251001	2/5	Sample collection requires physical manipulation in hazardous environments and precise spatial navigation around industrial equipment. While AI can potentially schedule sampling and log results, the hands-on collection itself—accessing specific points, handling containers safely, and ensuring sample integrity—remains difficult for current systems without significant robotics infrastructure.
Task automatability	claude-sonnet-5	1/5	Physically collecting fluid samples from plant equipment requires manual manipulation, valve operation, and physical presence at equipment sites, which current AI cannot perform.this is a physical, hands-on task not addressable by software AI.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power plant operations are heavily regulated by safety standards, environmental regulations, and industry protocols (e.g., OSHA, EPA, ASME) that typically require a qualified human operator to collect and certify samples for compliance and liability purposes. Chain-of-custody and certification requirements create hard barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Handling hazardous fluids like electrolytes or oils in power plants typically requires trained, often certified personnel following safety protocols, creating substantial procedural and safety barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Specialized robotic systems capable of safe sample collection in power plant environments are expensive to develop, integrate, and maintain. The loaded cost of such automation significantly exceeds the wage of a human technician performing the task, making economic substitution unfavorable today.
Cost vs. human wage	claude-sonnet-5	1/5	AI has no mechanism to physically collect samples, so there is no viable AI cost comparison; a human plus possibly robotic equipment would be required, which is far costlier than existing manual practice.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Robotic sample collection exists in research and limited industrial settings, but no mature, widely-deployed product reliably performs this task at scale in power plants today. Most plants still rely on human operators for the physical act of collection due to the need for judgment and real-time adaptation.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product performs physical sample collection; this remains a manual task requiring a human operator on-site with appropriate tools and safety gear.

Trace electrical circuitry to ensure compliance of electrical systems with applicable codes or laws.

CI 0–23 · exposure 13 · augmentation 38 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power plant operation is a laggard sector for AI adoption due to its high regulatory burden, physical infrastructure constraints, and entrenched workforce with specialized certifications. Automation in this domain proceeds slowly and typically as narrow augmentation rather than replacement.
Sector adoption velocity	claude-sonnet-5	1/5	Power plant operations are a highly physical, safety-critical, low-digitization environment with slow AI adoption for hands-on inspection tasks.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist by summarizing applicable codes, highlighting circuit diagram sections, and flagging potential compliance issues for human review, improving speed and consistency of manual checking. However, the human operator must remain in the loop to make final compliance judgments.
Augmentation potential	claude-sonnet-5	2/5	AI can assist with reference lookup of electrical codes or documentation support, but offers little help with the core physical tracing and verification work.
Task automatability	claude-haiku-4-5-20251001	2/5	Tracing electrical circuitry is a complex spatial and technical task requiring deep domain knowledge, real-time physical inspection, and judgment about code compliance. While AI can assist with circuit diagram analysis or code lookups, end-to-end automation meeting the 50% time-saving bar is not feasible with current systems—human operators remain essential for safety-critical verification.
Task automatability	claude-sonnet-5	1/5	This requires physical tracing of wiring and circuits in plant environments, on-site inspection, and judgment about code compliance tied to specific physical infrastructure—far beyond current AI's capability to execute end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	This task is heavily regulated under electrical codes (NEC, NFPA 70, etc.) and power plant safety standards, with legal liability attached to compliance failures. Licensed electricians and operators are often required to sign off on such work, creating hard regulatory and licensing barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Electrical code compliance in power generation facilities typically requires qualified/licensed personnel and carries significant safety and regulatory liability, creating strong barriers to non-human execution.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Current AI tools (vision systems, code databases) are still relatively expensive to integrate and maintain in a safety-critical power plant setting, while specialist power plant operators have decades of experience. The all-in cost of AI systems does not yet undercut the human labor significantly.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot perform the physical inspection component, so there is no viable AI-only cost comparison; a human must still do the on-site work.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product reliably performs independent electrical code compliance tracing today. Vision systems can assist with diagram recognition and code databases can be queried, but actual tracing on live systems with compliance judgment requires human expertise and has not reached production automation in power plants.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs physical circuit tracing and compliance verification in power plants; this remains a manual, hands-on task performed by trained operators/electricians.

Place standby emergency electrical generators on line in emergencies and monitor the temperature, output, and lubrication of the system.

CI 0–20 · exposure 13 · augmentation 50 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation is a heavily regulated, risk-averse sector with strong unions and strict certification requirements. Adoption of autonomous generator placement is not occurring in production systems; operators remain fully in control and legally accountable.
Sector adoption velocity	claude-sonnet-5	2/5	Power generation is a heavily regulated, safety-critical, physically-oriented sector with slow AI adoption for control-room emergency actions, though sensor-based monitoring is increasingly digitized.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-driven monitoring dashboards and alert systems can assist operators by aggregating sensor data and flagging anomalies in temperature, output, or lubrication, reducing information processing load during an emergency without removing human authority over the critical decision to bring generators online.
Augmentation potential	claude-sonnet-5	3/5	AI-based monitoring systems can track temperature, output, and lubrication data continuously and flag anomalies, helping operators respond faster, though the critical switching action remains human-performed.
Task automatability	claude-haiku-4-5-20251001	2/5	While monitoring temperature, output, and lubrication can be partially automated with sensors and alerts, the decision to place generators on line during emergencies requires real-time situational judgment, safety assessment, and coordination with grid operators—tasks that current AI cannot reliably perform end-to-end. The safety-critical nature and low frequency of true emergencies prevent the 50% time-saving threshold from being met.
Task automatability	claude-sonnet-5	1/5	This requires physical presence, real-time judgment under emergency conditions, and manual/hands-on operation of critical safety equipment; no AI system can perform the physical switchover and monitoring end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Critical infrastructure regulations (NERC, NRC, state public utility commissions) legally require licensed human operators to authorize and oversee generator startup and grid synchronization. Liability for blackout or equipment damage is severe, creating hard regulatory barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Critical infrastructure safety regulations, licensing requirements for power plant operators, and liability for emergency response ensure a human must be present and responsible.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Sensor systems and monitoring software are relatively inexpensive, but the AI system would still require human oversight, validation, and emergency decision-making, making the total cost approach or exceed that of a single operator's labor for the intermittent emergency events.
Cost vs. human wage	claude-sonnet-5	1/5	A human operator with fail-safe judgment and physical response capability is required; AI cannot substitute for the physical action, so cost comparison favors the human as the necessary actor.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Deployed SCADA and monitoring systems can track generator parameters, but no production AI system can autonomously decide to bring a generator online during an emergency or handle the complex handoff with grid operators. Academic and vendor demos exist but lack the reliability and integration maturity needed for unsupervised operation in critical infrastructure.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously places emergency generators online and performs physical monitoring; SCADA/sensor systems assist but do not perform the task itself.

Clean, lubricate, or maintain equipment, such as generators, turbines, pumps, or compressors, to prevent failure or deterioration.

CI 5–14 · exposure 8 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation is a capital-intensive, highly regulated, conservative sector with long equipment lifecycles and established union labor agreements. Adoption of maintenance automation remains minimal and limited to monitoring rather than hands-on execution.
Sector adoption velocity	claude-sonnet-5	1/5	Power plant physical maintenance is a highly non-digitized, physical-labor sector with minimal AI/robotic adoption for hands-on equipment servicing.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI-assisted monitoring systems, predictive maintenance analytics, and decision-support tools can help technicians prioritize work and detect anomalies earlier, raising planning efficiency without replacing hands-on maintenance work.
Augmentation potential	claude-sonnet-5	3/5	AI-driven predictive maintenance and sensor analytics can inform when and what to lubricate/maintain, helping schedule and prioritize physical tasks even though it doesn't perform them.
Task automatability	claude-haiku-4-5-20251001	2/5	Physical maintenance tasks like cleaning, lubricating, and inspecting equipment require robot manipulation in dynamic industrial environments—a capability that exists only in very limited pilot form. While AI can assist in scheduling and monitoring, the hands-on preventive maintenance cannot be meaningfully automated end-to-end today.
Task automatability	claude-sonnet-5	1/5	This is hands-on physical maintenance work—cleaning, lubricating, and servicing heavy industrial equipment—requiring manual dexterity and physical presence that no current AI system can perform.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power plants operate under strict regulatory oversight and safety codes that mandate documented, qualified personnel sign off on maintenance. Equipment-specific expertise and liability for failures create high legal and organizational barriers to full automation.
Adoption barriers	claude-sonnet-5	4/5	Safety-critical maintenance on power generation equipment typically requires certified technicians, lockout-tagout procedures, and regulatory compliance, creating strong barriers to any automated substitution.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The capital cost of specialized maintenance robotics, integration, and per-task oversight far exceeds the loaded wage of a trained power plant operator performing routine preventive maintenance.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical labor at all, so the comparison favors the human worker entirely; any robotic solution would be far more costly than current labor.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed autonomous system reliably performs equipment cleaning, lubrication, and preventive maintenance in power plants at scale. Robotics for these tasks remain largely research-stage or highly specialized; most power plants still depend on human technicians.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product performs physical lubrication or cleaning of turbines/generators; this remains purely a human/robotic mechanical task outside AI's scope.

Make adjustments or minor repairs, such as tightening leaking gland or pipe joints.

CI 5–14 · exposure 8 · augmentation 25 · importance 3.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power generation remains a physically-grounded, highly-regulated sector with long equipment lifecycles; adoption of autonomous repair robots is negligible, with most plants still relying on certified human technicians for safety-critical maintenance.
Sector adoption velocity	claude-sonnet-5	1/5	Power generation is a heavy industrial, physically intensive sector with low digitization of hands-on maintenance tasks and minimal robotic automation deployment for this specific task.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI-powered diagnostics or visual inspection tools could assist operators in identifying leaks, but the core skill—physically executing safe, precise repairs under pressure and regulatory constraint—offers limited augmentation scope today.
Augmentation potential	claude-sonnet-5	2/5	AI can support diagnostics (e.g., predictive maintenance alerts on leaks) but offers little direct assistance for the physical act of tightening joints or making minor mechanical repairs.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI vision systems can identify leaking joints, the task requires precise physical manipulation (tightening gland or pipe) in real-world plant conditions—something no current AI or robotic system deployed in power plants performs reliably or at significant time savings today.
Task automatability	claude-sonnet-5	1/5	This requires physical manipulation of hardware (tightening joints, gland packing) in a physical plant environment, which current AI systems cannot perform without robotic embodiment far beyond deployed capability.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Power plant operations are heavily regulated (NRC, OSHA); operators must be licensed and trained, and any automation of structural/safety repairs faces strict liability and certification barriers that prevent substitution without explicit regulatory approval.
Adoption barriers	claude-sonnet-5	4/5	Physical plant maintenance often requires trained/certified operators for safety and regulatory compliance, and equipment failure has high liability and safety consequences, creating strong barriers to non-human performance.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	Specialized robotic arms with force-feedback capable of precision repairs in hazardous industrial environments cost significantly more than the loaded wage of an experienced power plant operator when accounting for integration, maintenance, and oversight.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI-driven robotic solution for this physical task at any comparable cost; human labor remains the only practical option.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed product performs autonomous physical repairs like tightening leaking joints in operational power plants; this remains a skilled human task requiring dexterity, real-time environmental adaptation, and safety certification that no production system delivers.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs physical minor repairs like tightening leaking pipe or gland joints in power plants; this remains firmly in the human manual maintenance domain.

Repair or replace gas piping.

CI 0–0 · exposure 0 · augmentation 25 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Power plant operations remain highly conservative sectors with entrenched regulatory requirements for human certification and minimal digitization of field maintenance tasks. Adoption of automation in this domain is extremely slow.
Sector adoption velocity	claude-sonnet-5	1/5	Power generation and industrial maintenance trades show minimal AI adoption for physical repair tasks; this is a low-digitization, hands-on trade with no meaningful automation trend.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI could marginally assist through diagnostics or planning (identifying which pipes need replacement), but the core manual repair work and safety-critical decisions remain human-dependent. Augmentation potential is limited to planning phases.
Augmentation potential	claude-sonnet-5	2/5	AI can assist with diagnostics, documentation, or scheduling of repairs, but offers negligible direct assistance during the physical repair/replacement work itself.
Task automatability	claude-haiku-4-5-20251001	1/5	Repairing or replacing gas piping requires physical manipulation in hazardous environments, real-time problem diagnosis of complex systems, and safety-critical decision-making that current AI systems cannot perform. No end-to-end automation exists for this inherently physical task.
Task automatability	claude-sonnet-5	1/5	Physical repair and replacement of gas piping requires manual dexterity, welding/pipefitting skill, and on-site judgment that current AI systems cannot perform; no robotic system can autonomously execute this task in a power plant setting.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Gas piping repair in power plants is heavily regulated by safety codes (ASME, EPA, OSHA) and requires licensed, certified technicians to perform and sign off on work. Legal liability for safety failures creates hard barriers to automation.
Adoption barriers	claude-sonnet-5	5/5	Gas piping work is heavily regulated, requires licensed pipefitters/welders and code compliance inspections, and carries high liability (explosion/leak risk), creating hard legal and safety barriers to any automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	Specialized robotic systems capable of hazardous gas piping work would require extremely expensive custom engineering and safety certification, far exceeding the cost of trained human technicians who can already perform this work.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI substitute performing this physical task, so any AI cost is irrelevant—human labor remains the only option, making AI effectively infinitely more 'expensive' since it cannot do the job at all.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed autonomous robotic systems reliably perform gas piping repair or replacement in power plant settings today. The task requires dexterous manipulation, real-time sensing, and adaptation to site-specific conditions beyond current production-grade systems.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs gas pipe repair/replacement autonomously; this remains firmly in the domain of skilled tradespeople with physical tools.

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.