Transit and Railroad Police

33-3052.00
Median wage $90,230/yr4,390 employed (US)Rank #893 of 923 scored · top 97% by substitution

Protect and police railroad and transit property, employees, or passengers.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution9
Exposure8
Augmentation45

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

12 rated tasks, binned by substitution score.

Position among all scored occupations

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

Tasks with substitution ≥ 70

0%

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

Why this score

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

Task automatabilityw 35%6

panel mean rating 1.3/5 → substitution pressure 6/100

Technical feasibility todayw 20%10

panel mean rating 1.4/5 → substitution pressure 10/100

Cost vs. human wagew 15%11

panel mean rating 1.4/5 → substitution pressure 11/100

Adoption barriersw 20%inverted — strong barriers lower the score9

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

Sector adoption velocityw 10%14

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

Task breakdown (12 tasks)

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

Prepare reports documenting investigation activities and results.

31

CI 25–37 · exposure 33 · augmentation 63 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Transit and railroad police are in traditionally conservative, public-sector, safety-critical organizations with strong compliance cultures. Adoption of AI for investigation documentation lags professional services; most departments remain at pilot or early exploration stages.
Sector adoption velocityclaude-sonnet-52/5Public safety and transit policing are traditionally slow adopters of AI tools relative to information/finance sectors, though some AI report-writing pilots are emerging in broader law enforcement.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist officers by auto-populating standard fields, suggesting narrative structure, and organizing evidence summaries, meaningfully reducing drafting time. However, the officer must remain in the loop for judgment calls, legal sufficiency, and final certification.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully speed up drafting, formatting, and summarizing investigation notes into report templates, letting officers focus on verification and follow-up, while human oversight remains essential.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can draft narrative summaries from structured incident data and templates, investigation reports require domain-specific legal language, evidence chain documentation, and evidentiary judgment that demand human review and modification. Partial automation of formatting and data entry is possible, but achieving 50% time savings at equal quality would require significant customization and oversight.
Task automatabilityclaude-sonnet-53/5AI can draft structured incident reports from officer notes, transcripts, or dictation, but factual accuracy, chain-of-custody details, and legal sufficiency require human review, limiting full end-to-end automation.
Adoption barriersclaude-haiku-4-5-202510014/5Investigation reports are often evidence in legal proceedings and require human certification by the investigating officer. Regulatory, liability, and evidentiary standards typically require a licensed officer to author and sign the report, creating a hard barrier to full automation.
Adoption barriersclaude-sonnet-54/5Law enforcement reports are legal documents requiring officer certification, accuracy under oath, and departmental/legal accountability, creating strong barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI tools for report generation have moderate integration and setup costs, while a transit police officer's loaded wage is substantial. After accounting for human review, fact-checking, and legal compliance verification, AI cost advantage is marginal or nonexistent.
Cost vs. human wageclaude-sonnet-53/5AI drafting can cut report-writing time substantially, but required human verification, editing, and system integration keep costs roughly comparable rather than dramatically cheaper.
Technical feasibility todayclaude-haiku-4-5-202510012/5General-purpose document generation tools exist, but no deployed product reliably handles the specialized legal and investigative reporting requirements unique to transit/railroad police investigations. Products struggle with context-dependent detail selection and maintaining admissibility standards.
Technical feasibility todayclaude-sonnet-52/5General-purpose LLMs and some law-enforcement report-writing tools exist and are piloted in policing, but adoption in transit/rail police specifically is limited and error rates in factual specificity remain a concern.

Provide training to the public or law enforcement personnel in railroad safety or security.

21

CI 16–25 · exposure 17 · augmentation 63 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5These sectors (transportation, law enforcement) are moderately digitized and cautious about automating high-stakes training. Adoption of AI-delivered training remains limited; human instructors remain the norm in railroad and transit police academies.
Sector adoption velocityclaude-sonnet-52/5Public safety and transit policing sectors are slow to adopt AI-driven training tools compared to fast-moving digital industries, with most adoption limited to pilot e-learning content.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist trainers by generating slides, quizzes, or scenario outlines and automating administrative tasks like scheduling, but the core instructional role—explaining concepts, responding to confusion, assessing readiness—remains human-led. Modest productivity gain overall.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully help trainers by drafting curricula, generating scenario-based exercises, and creating quizzes or multimedia content, enhancing the human trainer's efficiency significantly.
Task automatabilityclaude-haiku-4-5-202510011/5Training delivery inherently requires interactive instruction, real-time feedback, and adaptation to learner questions—tasks that demand human pedagogical judgment and presence. Current AI cannot meaningfully replace the relational and adaptive components that define effective training delivery.
Task automatabilityclaude-sonnet-52/5AI can generate training materials or presentations but delivering interactive, credible safety/security training to law enforcement or the public requires live human instruction, Q&A, and authority that current AI cannot fully replace end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Railroad and law enforcement training typically requires certified instructors who must sign off on competency and are often subject to regulatory or organizational oversight. Liability for inadequate safety or security training creates strong disincentives to substitute human instruction with automated systems.
Adoption barriersclaude-sonnet-54/5Training law enforcement personnel often requires certified instructors and adherence to agency-specific protocols and legal standards, creating institutional and credentialing barriers to full AI substitution.
Cost vs. human wageclaude-haiku-4-5-202510012/5Human trainers still command significant loaded wages; AI could reduce material preparation costs but cannot eliminate the human instructor role entirely, keeping overall cost savings modest relative to the human wage for live instruction.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply produce supporting materials, but the core delivery still requires paid instructors, travel, and credentialed trainers, keeping overall costs comparable to human-led programs.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can generate training materials and scripts, no deployed system reliably delivers interactive, authoritative safety or security training that meets law enforcement or railroad standards without human instruction. AI tutoring products exist but lack the credibility and context-awareness required for this specialized domain.
Technical feasibility todayclaude-sonnet-52/5Products exist for content generation (slides, e-learning modules, chatbots for FAQs) but no deployed AI system independently conducts full railroad safety/security training programs for law enforcement audiences.

Interview neighbors, associates, or former employers of job applicants to verify personal references or to obtain work history data.

20

CI 20–20 · exposure 16 · augmentation 50 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5HR and transit agencies remain relatively low-adoption sectors for agent-based automation; while some large firms experiment with AI-assisted screening, autonomous reference interviewing is not yet standard practice in production environments, particularly in the transit/police domain where human oversight of hiring is customary.
Sector adoption velocityclaude-sonnet-52/5Law enforcement and transit security sectors are slow adopters of AI for sensitive personnel vetting tasks, with most automation limited to administrative record checks.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist a human HR officer by drafting interview templates, flagging inconsistencies in written references, or organizing call schedules, meaningfully raising their efficiency. However, the human must remain the primary interviewer to ensure compliance and credibility assessment.
Augmentation potentialclaude-sonnet-53/5AI can help by generating interview scripts, summarizing notes, transcribing calls, and flagging inconsistencies, providing moderate productivity gains to the investigator.
Task automatabilityclaude-haiku-4-5-202510012/5Conducting reference interviews requires interpersonal judgment, contextual questioning, and the ability to detect deception or inconsistency—capabilities where current AI systems lack reliability. While AI could assist with scheduling or drafting standard questions, the core task of extracting truthful, nuanced information through conversation remains beyond current AI's independent performance threshold.
Task automatabilityclaude-sonnet-52/5While AI can draft interview questions or transcribe/summarize calls, actually conducting nuanced interviews with references and assessing credibility requires human judgment and rapport that current systems cannot reliably replace end-to-end.'
Adoption barriersclaude-haiku-4-5-202510014/5Reference verification touches hiring decisions and legal liability; many organizations face regulatory requirements around data collection and verification accuracy, and employers typically prefer human judgment on reference credibility. The human-contact expectation and liability asymmetry (poor reference calls carry legal/employment risk) create substantial friction.
Adoption barriersclaude-sonnet-54/5Background investigations for police/security-sensitive positions often have legal and procedural requirements mandating trained personnel to conduct and certify interviews, creating strong barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of building and maintaining reliable automated reference-checking systems (including error handling, legal review, and human fallback) approaches or exceeds the wage of a junior HR professional who can perform interviews competently, making the all-in economic case weak.
Cost vs. human wageclaude-sonnet-52/5AI could cheaply assist with scheduling or transcription, but the core interviewing and judgment work still requires paid investigator time, so overall cost savings are limited.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably conducts autonomous reference interviews at scale. While chatbots can simulate Q&A, they cannot establish credibility, navigate sensitive HR conversations, or make hiring-critical judgment calls about reference validity in production systems.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously conducts background reference interviews for law enforcement hiring; existing background-check tools automate database searches, not interpersonal interviews.

Plan or implement special safety or preventive programs, such as fire or accident prevention.

14

CI 11–16 · exposure 5 · augmentation 50 · importance 3.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Transit and railroad sectors are moderately digitized, but safety program planning remains primarily human-driven. While data analytics and monitoring tools are increasingly adopted, the strategic design and implementation of safety programs themselves have seen slow AI adoption, typical of heavily regulated safety-critical domains.
Sector adoption velocityclaude-sonnet-52/5Law enforcement and transit safety sectors are slow adopters of AI for planning-level functions, with pilots limited mostly to surveillance/detection tools rather than program design.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by analyzing incident data, identifying trends, generating program templates, and supporting compliance documentation. However, the human safety professional remains central to contextual judgment, stakeholder engagement, and final program design—making this a genuine assistive role rather than transformative productivity gain.
Augmentation potentialclaude-sonnet-53/5AI can assist by analyzing incident data, drafting policy documents, and identifying risk patterns, aiding but not replacing the human planning process.
Task automatabilityclaude-haiku-4-5-202510011/5Planning and implementing safety programs requires strategic judgment, stakeholder collaboration, regulatory knowledge, and organizational context that current AI systems cannot perform end-to-end. While AI might assist in data analysis or documentation, the core task of designing and executing safety programs demands human expertise and decision-making.
Task automatabilityclaude-sonnet-51/5Designing and implementing safety programs requires site-specific judgment, stakeholder coordination, and physical inspection that current AI cannot perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Safety programs in transit and rail are heavily regulated and often require certification, audit trails, and documented human accountability. Liability exposure and regulatory requirements for safety compliance create strong barriers; a human safety officer or supervisor must sign off on and oversee program implementation for legal and operational reasons.
Adoption barriersclaude-sonnet-54/5Safety program design in rail/transit is subject to regulatory oversight (e.g., FRA, transit authorities) and liability concerns, requiring accountable human officers to plan and sign off.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for safety analysis are relatively affordable, but full program implementation—including training, coordination, and compliance verification—still requires human specialists. The all-in cost of AI-assisted safety programs remains comparable to or higher than human-led approaches for this supervisory-level task.
Cost vs. human wageclaude-sonnet-52/5AI can cut some analysis/drafting time but the bulk of the task (stakeholder engagement, physical assessment, decision-making) still requires paid human labor, so overall cost savings are limited.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed products reliably perform comprehensive safety program planning and implementation independently. Some AI tools exist for hazard detection or incident analysis, but actual organizational safety program design and rollout remains human-led and is not demonstrably automated in production transit or railroad operations.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product plans or implements transit safety/prevention programs; this remains a human planning and organizational function.

Monitor transit areas and conduct security checks to protect railroad properties, patrons, and employees.

9

CI 7–11 · exposure 9 · augmentation 50 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Transit agencies have deployed camera systems and monitoring aids, but actual displacement of security officers is minimal. Human officers remain the primary security presence in most transit systems; adoption of full automation is limited by legal, safety, and organizational barriers.
Sector adoption velocityclaude-sonnet-52/5Transit security is adopting AI-assisted camera analytics and anomaly detection but sector-wide deployment of autonomous security agents remains nascent and pilot-stage.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-powered video analysis, threat flagging, and real-time alerts can assist transit police in identifying anomalies and prioritizing patrols, improving situational awareness. However, augmentation is limited to information support rather than transformative productivity gains.
Augmentation potentialclaude-sonnet-53/5AI-powered video analytics, anomaly detection, and dispatch support can help officers prioritize patrols and flag suspicious activity, improving efficiency while humans retain enforcement duties.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time physical presence, dynamic threat assessment, judgment about suspicious behavior, and immediate intervention—capabilities far beyond current AI systems. Monitoring can be partially automated with cameras and sensors, but the integrated security decision-making and physical response components cannot be fully automated.
Task automatabilityclaude-sonnet-51/5This task requires physical presence, situational judgment, and use-of-force authority in dynamic environments that current AI cannot perform end-to-end.'},
Adoption barriersclaude-haiku-4-5-202510015/5Transit and railroad security is heavily regulated; only licensed, trained, and authorized personnel can conduct official security checks, carry enforcement authority, and be held liable for breaches. Legal and regulatory requirements mandate human officers.
Adoption barriersclaude-sonnet-55/5Sworn police authority, arrest powers, and legal accountability require a certified officer; this is a licensed law-enforcement function with strict regulatory and liability barriers.
Cost vs. human wageclaude-haiku-4-5-202510011/5Comprehensive security monitoring with current AI vision systems, integration, and required human oversight is expensive; total cost per effective security check remains higher than hiring transit police, particularly given liability exposure from false negatives.
Cost vs. human wageclaude-sonnet-52/5Surveillance software is cheap to run but does not replace the human presence, authority, and response capability needed, so full substitution costs remain high when factoring necessary human backup.
Technical feasibility todayclaude-haiku-4-5-202510012/5Video surveillance systems exist and can flag some anomalies, but current AI cannot reliably conduct comprehensive security assessments, de-escalate situations, or make nuanced judgments about threats. Deployed systems assist but fall far short of replacing the core task end-to-end.
Technical feasibility todayclaude-sonnet-52/5AI-enabled camera analytics and sensor systems exist for surveillance flagging, but no product independently conducts security checks or physically patrols and protects transit properties.

Direct or coordinate the daily activities or training of security staff.

6

CI 0–11 · exposure 0 · augmentation 38 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Public safety and transit agencies lag significantly in AI adoption due to regulatory requirements, unionization, civil service rules, and the human-critical nature of security management; adoption remains minimal and primarily limited to scheduling aids rather than coordination.
Sector adoption velocityclaude-sonnet-52/5Security and transit policing sectors are slow to adopt AI for managerial functions, with most AI use limited to surveillance or dispatch support rather than personnel leadership.
Augmentation potentialclaude-haiku-4-5-202510012/5AI can offer limited assistance with scheduling, reporting, or training content recommendations, but the core task of directing and coordinating staff requires human judgment and presence, so augmentation is minimal and peripheral to the primary function.
Augmentation potentialclaude-sonnet-53/5AI can help by generating training materials, analyzing staff performance data, or optimizing shift schedules, providing moderate assistance to human supervisors.
Task automatabilityclaude-haiku-4-5-202510011/5Directing and coordinating daily activities and training of security staff requires dynamic human judgment, interpersonal interaction, and contextual decision-making that current AI systems cannot perform end-to-end. AI cannot reliably manage personnel, resolve conflicts, or adapt training to individual staff needs without continuous human oversight.
Task automatabilityclaude-sonnet-51/5Directing and training security staff requires real-time leadership, judgment, and interpersonal supervision that current AI cannot perform end-to-end.'},'feasibility':{'rating':1,'rationale':'No deployed AI product manages or trains security personnel autonomously; this remains a human management function.
Adoption barriersclaude-haiku-4-5-202510015/5Personnel management and training direction are legally and organizationally bound to human supervisors; liability, accountability, and duty-of-care requirements mandate a licensed or authorized human remain responsible for directing staff and ensuring training compliance.
Adoption barriersclaude-sonnet-54/5Security and law enforcement supervision typically requires accountable human authority, chain-of-command structures, and legal responsibility for personnel actions, creating strong institutional barriers.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of implementing AI systems for personnel coordination, combined with the specialized domain knowledge and human oversight required, substantially exceeds the loaded wage of a manager performing these duties today.
Cost vs. human wageclaude-sonnet-52/5AI tools can lower administrative overhead (scheduling, training content generation) but cannot replace the supervisory labor itself, so cost savings are partial.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed products reliably perform personnel management, staff coordination, and training direction as a standalone function. While AI can assist with scheduling or content generation, no production system autonomously directs and coordinates security staff activities.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product autonomously directs or trains security staff; management software only assists with scheduling or record-keeping.

Patrol railroad yards, cars, stations, or other facilities to protect company property or shipments and to maintain order.

5

CI 3–7 · exposure 5 · augmentation 50 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI-assisted surveillance tools is gradual in rail and transit; most yards still rely on traditional patrols. Sector digitization is slower than information/finance, and conservative risk tolerance around security delays deeper automation.
Sector adoption velocityclaude-sonnet-52/5Rail and transit security is a physically-grounded, low-digitization sector where AI adoption is limited to surveillance-assist tools, with actual patrol duties still performed by human officers.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-powered cameras, motion detection, and alert systems can assist officers by flagging anomalies and reducing blind spots, but the human officer remains the primary decision-maker and responder in this inherently human-presence-dependent role.
Augmentation potentialclaude-sonnet-53/5AI-enabled cameras, sensors, and analytics can flag anomalies or intrusions to help officers prioritize patrol routes, but the core patrol and enforcement task remains human-driven.
Task automatabilityclaude-haiku-4-5-202510011/5Patrolling physical spaces to maintain order and protect property requires real-time situational awareness, discretionary judgment in security incidents, and dynamic human interaction. Current AI cannot autonomously perform end-to-end physical security patrols that meet legal and operational standards.
Task automatabilityclaude-sonnet-51/5Physical patrolling, situational judgment, and enforcement authority in transit/rail environments cannot be performed end-to-end by current AI systems; this requires embodied presence and legal authority to act.
Adoption barriersclaude-haiku-4-5-202510015/5Railroad and property security involves legal liability for incidents, regulatory compliance for transportation security, and implicit requirement for a uniformed, accountable human to respond to and document incidents. Liability law and transportation regulations create hard barriers to full automation.
Adoption barriersclaude-sonnet-55/5This is sworn law enforcement work requiring police authority, arrest powers, and legal accountability that only a certified officer can exercise, making substitution essentially impossible without human presence.
Cost vs. human wageclaude-haiku-4-5-202510011/5Security patrols require continuous human presence, situational judgment, and legal accountability; even with supporting AI, the human labor cost dominates. Autonomous systems would face prohibitive liability and integration costs relative to human officers.
Cost vs. human wageclaude-sonnet-51/5Physical security patrol requires human presence, vehicles, and enforcement capability; AI camera monitoring adds cost on top of human officers rather than replacing the all-in labor cost.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI-powered surveillance cameras and monitoring systems exist in production, they assist human officers rather than replacing patrol duties. No deployed system independently performs the full patrol, threat assessment, and response functions this role requires.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously patrols rail yards and enforces order; existing camera/sensor systems only support human officers rather than replace them.

Examine credentials of unauthorized persons attempting to enter secured areas.

4

CI 0–7 · exposure 5 · augmentation 38 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Transit police and railroad security remain human-intensive, regulated sectors with slow technology adoption. Security decisions involve legal and safety responsibility that organizations retain with trained human officers rather than deploying autonomous systems.
Sector adoption velocityclaude-sonnet-52/5Physical security and law enforcement sectors adopt AI tools (e.g., facial recognition, credential scanners) slowly and cautiously due to legal, privacy, and liability constraints, with humans retaining decision authority.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could assist marginally by pre-screening documents or flagging anomalies for officer review, but the task inherently requires human judgment, legal authority, and real-time social assessment that limits meaningful productivity augmentation.
Augmentation potentialclaude-sonnet-53/5AI-assisted ID verification, badge scanning, and facial recognition systems can help flag anomalies or verify credentials faster, augmenting the officer's assessment without replacing their judgment or authority.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time human judgment to assess physical credentials, detect forgeries, recognize subtle inconsistencies, and make security decisions under uncertain conditions. Current AI cannot reliably perform end-to-end credential verification and access-control decisions in dynamic, high-stakes environments at 50%+ time savings.
Task automatabilityclaude-sonnet-51/5This requires physical presence, judgment about deception, authority to detain, and real-time interaction with potentially uncooperative individuals—far beyond current AI capability to execute end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Transit and railroad security is heavily regulated (TSA, railroad security protocols, local law enforcement standards), and only authorized personnel are legally permitted to make access decisions at secured areas. Liability for wrongful entry or denial creates a hard barrier to full automation.
Adoption barriersclaude-sonnet-55/5This is a law enforcement function requiring sworn authority, legal power to detain/arrest, and accountability under use-of-force and civil rights law—AI cannot legally perform this role.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of AI systems (hardware, software, integration, and required human oversight) combined with liability exposure for false admissions or denials exceeds the loaded wage of a transit police officer performing this task.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this full task, so cost comparison favors the human officer who provides necessary judgment, authority, and physical response capability.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can assist with document classification and OCR on static images, no deployed system reliably authenticates credentials, detects sophisticated forgeries, or independently makes access-control decisions in real operational security contexts without human oversight.
Technical feasibility todayclaude-sonnet-51/5No deployed product independently examines credentials and makes access-control decisions with enforcement authority in secured transit areas; this remains a human security/law enforcement function.

Apprehend or remove trespassers or thieves from railroad property or coordinate with law enforcement agencies in apprehensions and removals.

0

CI 0–0 · exposure 0 · augmentation 38 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Transit and railroad police are part of public safety infrastructure with minimal automation adoption; the field remains heavily reliant on human officers and is not undergoing AI-driven displacement in production systems.
Sector adoption velocityclaude-sonnet-51/5Physical security and policing sectors show minimal AI adoption for hands-on enforcement actions; surveillance tools are used but not physical intervention.
Augmentation potentialclaude-haiku-4-5-202510012/5AI can assist modestly by detecting trespassers via video surveillance or alerting officers to location and activity, but the core apprehension task remains entirely human-driven, limiting the scope of augmentation.
Augmentation potentialclaude-sonnet-53/5AI-powered surveillance, sensor detection, and predictive analytics can help identify trespassers and coordinate dispatch, aiding officers' situational awareness even though the physical act remains human-performed.
Task automatabilityclaude-haiku-4-5-202510011/5Apprehending or removing trespassers and thieves requires physical presence, situational judgment, de-escalation, and potential use of force—tasks that current AI systems cannot perform. No meaningful portion of this safety-critical, physically-grounded task can be automated end-to-end today.
Task automatabilityclaude-sonnet-51/5Physical apprehension and removal of trespassers requires embodied action, use of force authority, and real-time judgment that no current AI system can perform.
Adoption barriersclaude-haiku-4-5-202510015/5Legal and liability barriers are absolute: only sworn law enforcement or authorized security personnel can lawfully apprehend suspects or remove trespassers, and these actions carry strict liability frameworks requiring human judgment and accountability. Automation is legally prohibited.
Adoption barriersclaude-sonnet-55/5Use of force and arrest authority is legally restricted to sworn, licensed law enforcement officers, creating a hard legal barrier to automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI has no cost advantage because the task requires human presence and physical action. Deploying AI-based detection systems would supplement, not replace, the trained officer performing the apprehension, adding cost rather than reducing it.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute for physical apprehension, so the comparison is moot—human officers remain the only viable option, making AI effectively infinitely costlier for this specific action.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product can apprehend, remove, or physically interact with persons. While computer vision can detect trespassers, the core task of apprehension and removal remains entirely dependent on human law enforcement personnel.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product physically apprehends or removes people; this remains entirely a human law enforcement function.

Direct security activities at derailments, fires, floods, or strikes involving railroad property.

0

CI 0–0 · exposure 0 · augmentation 25 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5This task sits at the intersection of law enforcement, emergency response, and railroad regulation—sectors with strong incumbent human-centric structures and minimal displacement by AI agents. Adoption of autonomous security direction is negligible.
Sector adoption velocityclaude-sonnet-51/5Physical security and law enforcement in transportation infrastructure show minimal AI-driven displacement; this is a low-digitization, high-physical-presence sector.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could provide limited assistance such as real-time data feeds from cameras, hazard detection alerts, or communication coordination support. However, the core task of directing security operations and making tactical decisions remains firmly human-led, limiting meaningful augmentation.
Augmentation potentialclaude-sonnet-52/5AI could assist with situational awareness via sensors, mapping, or communications logging, but offers little help with the core task of on-scene command and coordination during emergencies.
Task automatabilityclaude-haiku-4-5-202510011/5Directing security at emergency incidents requires real-time situational assessment, dynamic decision-making under uncertainty, coordination with multiple stakeholders, and physical presence at sites with evolving threats. Current AI cannot operate autonomously in this safety-critical, unstructured environment at scale.
Task automatabilityclaude-sonnet-51/5This requires physical presence, real-time on-scene command authority, and adaptive judgment during dangerous emergencies—no AI system can direct security operations at a physical incident scene.
Adoption barriersclaude-haiku-4-5-202510015/5Transit and railroad police are licensed law enforcement with statutory authority to direct security. Only a qualified human officer can legally command security operations, issue orders, and bear liability for decisions at emergencies involving railroad property, derailments, and public safety incidents.
Adoption barriersclaude-sonnet-55/5Directing security at emergencies involving railroad property requires sworn law enforcement authority, legal jurisdiction, and physical presence—hard legal and physical barriers prevent AI substitution.
Cost vs. human wageclaude-haiku-4-5-202510011/5Human transit police officers are specialized, trained personnel whose presence and accountability are legally required. Even if partial aspects (monitoring feeds, alerting) were AI-assisted, the core task of directing security operations cannot be delegated to AI, making the human cost irreplaceable.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this function, so any comparison is moot—human labor is the only viable option and thus effectively cheaper than a nonexistent alternative.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI system can independently direct security operations at emergency scenes. While AI can assist with monitoring or data analysis, assuming command authority and making tactical security decisions in real crises requires human judgment that production AI does not demonstrate.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs on-scene incident command for railroad emergencies; this remains firmly in the domain of trained human officers.

Investigate or direct investigations of freight theft, suspicious damage or loss of passengers' valuables, or other crimes on railroad property.

0

CI 0–0 · exposure 0 · augmentation 38 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Transit and railroad police are government agencies with traditional structures and conservative adoption patterns. Criminal investigation itself is not automating in these sectors; digital tools assist humans rather than replace investigative functions.
Sector adoption velocityclaude-sonnet-51/5Transit/railroad policing is a physical, safety-critical, low-digitization sector with minimal AI agent deployment in investigative work.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could modestly assist by organizing evidence databases, flagging suspicious patterns in historical data, or transcribing interviews, but the core investigative work—interviewing, judgment, direction—remains human-dependent. Assistance is marginal compared to the human investigator's core role.
Augmentation potentialclaude-sonnet-53/5AI tools can assist with reviewing camera footage, flagging anomalies, cross-referencing records, and drafting reports, meaningfully aiding investigators without replacing their judgment or authority.
Task automatabilityclaude-haiku-4-5-202510011/5Investigating crimes requires complex human judgment, interviewing witnesses, evaluating evidence credibility, understanding context and intent, and making determinations about criminal responsibility—tasks that current AI cannot perform end-to-end with sufficient reliability or legal standing. No current AI system can independently conduct investigations meeting investigative standards.
Task automatabilityclaude-sonnet-51/5Investigating theft or crimes requires physical presence, evidence collection, witness interviews, and legal authority to direct investigations—none of which current AI can perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Hard regulatory and legal barriers exist: law enforcement investigation authority is vested in licensed officers, evidence handling has strict legal requirements, and investigative findings must be defensible in court with human accountability. No AI can substitute for the licensed investigator's legal authority and liability responsibility.
Adoption barriersclaude-sonnet-55/5This is inherently a law enforcement function requiring sworn, licensed police authority, chain-of-custody procedures, and legal accountability—hard institutional and legal barriers prevent automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI would require significant human oversight and cannot reduce the core labor cost of investigating crimes, since human investigators must conduct interviews, examine scenes, and make authoritative determinations. The loaded cost of AI plus mandatory human investigation would exceed human-only approaches.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the sworn officer's role, so there is no viable cost comparison for full task replacement.
Technical feasibility todayclaude-haiku-4-5-202510011/5While AI can assist with data analysis or evidence cataloging, no deployed product reliably performs criminal investigation or directs investigations as the task requires. Criminal investigation demands human authority, legal accountability, and judgment that current AI systems lack in production settings.
Technical feasibility todayclaude-sonnet-51/5No deployed product conducts or directs criminal investigations autonomously; AI is at most used for surveillance analytics, not investigation direction.

Enforce traffic laws regarding the transit system and reprimand individuals who violate them.

0

CI 0–0 · exposure 0 · augmentation 38 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Transit agencies have not meaningfully adopted AI systems to enforce traffic laws or reprimand violators in production. Pilot projects for violation detection exist but enforcement authority remains with human officers, indicating laggard adoption in this high-liability domain.
Sector adoption velocityclaude-sonnet-51/5Law enforcement and physical security sectors show minimal AI-driven displacement of frontline enforcement duties; adoption is limited to surveillance/detection support tools, not enforcement itself.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could potentially assist by flagging suspected violations for officer review (e.g., automated camera detection of fare evasion or dangerous behavior), but current systems are too unreliable and the interpersonal reprimand component remains entirely human-driven, limiting augmentation value.
Augmentation potentialclaude-sonnet-53/5AI-powered surveillance cameras, license plate readers, and fare-evasion detection can help flag violations for officers to act on, offering moderate assistance without replacing the enforcement action.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time detection of traffic law violations in a transit system, judgment about appropriate enforcement action, and interpersonal interaction with violators. Current AI systems cannot reliably identify all violation types, lack legal authority to enforce, and cannot manage the complex human interaction required to reprimand individuals.
Task automatabilityclaude-sonnet-51/5This requires physical presence, real-time judgment, authority to detain/cite, and interpersonal confrontation—none of which current AI can perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Enforcement of traffic laws and authority to reprimand individuals requires legal licensing and delegated law enforcement authority. A human police officer must perform or sign off on enforcement actions; this is a hard legal barrier that prevents full substitution.
Adoption barriersclaude-sonnet-55/5Enforcement and reprimand require sworn police authority, legal power to detain/cite, and accountability structures that only licensed officers can hold.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of AI infrastructure (vision systems, integration into transit operations, oversight, legal review) combined with low per-interaction savings does not achieve favorable cost ratios compared to a transit police officer's hourly rate for enforcement work.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this function, so AI cost is not comparable; a sworn officer's cost is the only real input currently.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product performs enforcement and reprimand of transit violations end-to-end. While object detection can identify some violations, no system is operationally authorized or capable of issuing citations or conducting the interpersonal reprimand interactions that define this task.
Technical feasibility todayclaude-sonnet-51/5No deployed product enforces traffic laws or reprimands individuals in transit systems; this remains firmly a human law-enforcement function.

Related occupations — Protective Service

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.