Customs and Border Protection Officers
33-3051.04Median wage $76,210/yr670,520 employed (US)Rank #706 of 923 scored · top 76% by substitution
Investigate and inspect persons, common carriers, goods, and merchandise, arriving in or departing from the United States or between states to detect violations of immigration and customs laws and regulations.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution19
Exposure21
Augmentation54
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
11 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%22
panel mean rating 1.9/5 → substitution pressure 22/100
Technical feasibility todayw 20%20
panel mean rating 1.8/5 → substitution pressure 20/100
Cost vs. human wagew 15%23
panel mean rating 1.9/5 → substitution pressure 23/100
Adoption barriersw 20%inverted — strong barriers lower the score9
panel mean rating 4.7/5 (barrier strength) → substitution pressure 9/100
Sector adoption velocityw 10%17
panel mean rating 1.7/5 → substitution pressure 17/100
Task breakdown (11 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.
Determine duty and taxes to be paid on goods.
56CI 55–57 · exposure 66 · augmentation 88 · importance 4.0/5 · click for rater detail
Determine duty and taxes to be paid on goods.
56| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Customs automation has been slow relative to other government and commercial sectors. Legacy systems dominate; CBP processes remain human-intensive by design. Although some tariff lookup tools exist, deployment of AI agents for end-to-end duty determination remains experimental and not yet normalized in practice. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Customs agencies and trade compliance firms have adopted automated classification and duty calculation tools, but government border agencies tend to be slower, more regulated adopters than private-sector finance or tech. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can powerfully assist by automating tariff lookups, suggesting HS codes, flagging high-risk classifications, and pre-populating duty estimates for officer review. A CBP officer using such tools experiences major productivity gains while retaining final responsibility for determinations and handling edge cases. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-assisted tariff classification and duty calculation tools substantially speed up officers' work by suggesting codes and computing amounts, while the officer retains authority for final determination and enforcement. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Duty and tax determination relies heavily on tariff schedules, HS codes, and rule-based classification—all machine-readable and automatable. Most of the work involves lookups and formula application rather than subjective judgment, though complex classification edge cases or value disputes still require expert review. |
| Task automatability | claude-sonnet-5 | 4/5 | Duty/tax determination is a rules-based classification and calculation task (tariff codes, valuation, applicable rates) that AI systems can perform with high accuracy given structured input data, saving significant time over manual lookup. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Legal and regulatory barriers are substantial: duty determinations have legal consequences, tariff schedules are statutory, and misclassification carries financial and compliance risk. CBP officers have delegated authority and accountability; substituting that with autonomous AI without explicit regulatory framework and human sign-off is not yet permitted, and liability exposure deters full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Final duty determinations carry legal and revenue implications, often requiring authorized government officer sign-off, audit trails, and accountability under customs law, limiting full automation without human oversight. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Inference cost for lookups and basic rule application is trivial; integration and oversight are modest. The loaded cost of a Customs officer's time to perform routine duty calculations is substantially higher than a streamlined AI-assisted process, though complex cases still need paid human expert review. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated tariff/duty calculation engines process transactions at a fraction of the cost of manual officer review, especially at scale for routine shipments. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Commercial systems exist (e.g., tariff classification APIs, customs compliance software) but operate in narrow scopes or require significant human oversight due to legal and correctness stakes. No fully autonomous end-to-end duty calculation system has reached widespread reliable production deployment without human verification. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Trade compliance software and customs classification tools (e.g., HS code classifiers, automated duty calculators) exist and are used commercially, but government CBP systems still require human review for edge cases, disputes, and enforcement decisions. |
Record and report job-related activities, findings, transactions, violations, discrepancies, and decisions.
51CI 37–65 · exposure 58 · augmentation 75 · importance 4.2/5 · click for rater detail
Record and report job-related activities, findings, transactions, violations, discrepancies, and decisions.
51| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | U.S. federal agencies, including DHS and CBP, adopt automation slowly due to legacy systems, security classification concerns, and procurement friction. While pilots exist, production-scale displacement of reporting tasks remains limited and confined to internal administrative workflows. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Government law enforcement agencies, especially federal security agencies, are slow adopters of AI due to security, legal, and procurement constraints, with pilots rare and production deployment rarer still. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-assisted drafting of incident reports, auto-population of structured forms from inspection notes, and violation classification tools substantially boost officer productivity without removing human judgment. Officers benefit from reduced manual transcription and improved record completeness. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by transcribing, summarizing, and pre-filling structured reports from officer dictation or notes, saving significant drafting time while the officer retains responsibility for accuracy and final decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI systems can reliably extract, classify, and document violations, transactions, and findings from structured inspections and interviews, then generate automated reports that meet ≥50% time savings. Human review of edge cases or high-stakes decisions would still be needed, but the bulk of recording and routine reporting is highly automatable with current NLP and document generation tools. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft structured reports and logs from officer notes or transaction data, but capturing accurate real-time findings, violations, and decisions requires human judgment and verification, limiting full automation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Legal and regulatory requirements vary: some border agencies require human sign-off on violations and formal decisions, creating oversight friction. However, recording and initial reporting are not universally gatekept by licensing; bureaucratic and procedural inertia pose moderate adoption barriers. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Official government records of violations and legal decisions carry strong authentication, chain-of-custody, and accountability requirements that typically mandate sworn officer authorship and sign-off. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Inference costs for document classification and report generation are negligible per transaction (~cents), while a CBP officer's loaded wage is ~$60–80/hour. At scale, even with integration and oversight overhead, automation costs are orders of magnitude cheaper per documented activity. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted documentation tools could cut time spent on report writing, but integration with secure government systems and required accuracy oversight keeps costs roughly comparable to current officer time allocation. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed AI systems (document automation, OCR, NLP classifiers) are already performing similar compliance logging and report generation in financial and healthcare contexts. Customs agencies have begun piloting automated violation flagging and incident documentation; these systems work reliably on structured data but require human oversight on ambiguous cases. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Deployed products (transcription, form-filling, case management assistants) exist in some law enforcement contexts, but no mature CBP-specific product reliably automates end-to-end incident/violation reporting today. |
Inspect cargo, baggage, and personal articles entering or leaving U.S. for compliance with revenue laws and U.S. customs regulations.
24CI 20–28 · exposure 25 · augmentation 63 · importance 4.6/5 · click for rater detail
Inspect cargo, baggage, and personal articles entering or leaving U.S. for compliance with revenue laws and U.S. customs regulations.
24| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | CBP has adopted risk-assessment and screening tools and pilot programs, but meaningful deployment remains limited; agencies continue to rely heavily on human inspectors and adoption velocity is constrained by regulatory requirements and legacy processes. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Government/security agencies adopt AI slowly for core enforcement functions due to legal, chain-of-custody, and accountability constraints, though scanning tech adoption is ongoing. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist with document pre-screening, risk flagging, and regulatory lookup to prioritize inspection focus, meaningfully raising inspector productivity without removing human judgment from the compliance decision. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-enhanced imaging, anomaly detection, and risk-scoring significantly help officers prioritize inspections and spot anomalies, meaningfully boosting efficiency while humans retain final judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with document scanning and risk assessment flagging, the task fundamentally requires physical inspection of items and nuanced judgment about regulatory compliance that varies with context. Current AI systems cannot reliably perform the full end-to-end inspection and compliance determination with 50% time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | Physical inspection of cargo and baggage requires manipulation, sensory judgment, and often on-site decision-making that current AI cannot perform end-to-end; AI can flag risk but not conduct the inspection itself. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | This task is legally restricted to authorized Customs and Border Protection officers; international law and U.S. federal statute require human inspection and sign-off for customs clearance, creating hard barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This is a sworn federal law enforcement function requiring statutory authority to search, seize, and enforce customs law—only credentialed officers can legally perform it. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted screening tools are deployed, but their total cost (infrastructure, integration, human oversight) does not yet undercut the fully loaded cost of border officers performing the task, especially given liability asymmetry and regulatory requirements. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Scanning and detection systems reduce some labor costs, but human officers are still required for hands-on inspection, verification, and legal authority, keeping overall costs comparable to human labor. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Document analysis and risk scoring tools exist in deployed systems, but the core requirement—reliable cargo inspection for compliance—depends on human physical examination and contextual legal judgment that AI cannot yet replicate in production at scale. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | X-ray/scanning analytics and risk-targeting systems are deployed to assist, but no product independently performs full physical inspection and compliance determination reliably at scale. |
Investigate applications for duty refunds and petition for remission or mitigation of penalties when warranted.
23CI 20–25 · exposure 25 · augmentation 63 · importance 3.1/5 · click for rater detail
Investigate applications for duty refunds and petition for remission or mitigation of penalties when warranted.
23| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | CBP and customs agencies operate in heavily regulated, risk-averse environments where process automation is slow. Although digitization of applications is underway, adoption of AI-driven investigation or penalty decisions remains in pilot or limited deployment phases rather than enterprise-wide production. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Government law enforcement and customs agencies are typically slow adopters of AI for discretionary legal decisions due to accountability and due-process requirements. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist officers by summarizing application materials, flagging precedent cases, and organizing financial records, materially reducing manual document review time. However, the human officer must retain investigative judgment and final discretionary authority over remission determinations. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly assist by extracting relevant data, flagging inconsistencies, summarizing case files, and suggesting precedent-based recommendations, speeding up the officer's review process. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can extract and organize application data and flag pattern anomalies, the core task requires nuanced judgment about warrant and discretionary mitigation decisions that depend on legal precedent, agency policy, and case-specific context. Current systems cannot reliably make the end-to-end determination at equal quality to a trained officer. |
| Task automatability | claude-sonnet-5 | 2/5 | The investigation involves reviewing documentation, applying regulatory judgment, and making discretionary decisions about penalty mitigation that require contextual and legal judgment beyond current AI capability to fully automate.atural |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Duty refund and penalty remission authority is vested in federal officers and subject to regulatory review and statutory limits. Legal liability for incorrect refunds or inappropriate penalty waiver falls on the agency, creating high error-cost asymmetry and a requirement that human officers retain accountability and sign-off on determinations. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This is a governmental enforcement and adjudicative function requiring authorized federal officers with legal authority to grant remission or mitigation, making it a hard legal/regulatory barrier. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce administrative overhead in document preparation and case assembly, but the investigative expertise and legal judgment command significant human labor cost. Integration and oversight of AI-assisted workflows would not yet achieve clear cost advantage over skilled officer review. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply process documents but the investigation and discretionary decision-making requires significant human oversight and legal accountability, keeping all-in costs comparable to human labor. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some document processing and preliminary application triage tools exist, but no deployed system reliably performs the full investigation and penalty mitigation decision at production quality. The task involves statutory interpretation and discretionary authority that remains human-centered in practice. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI document review and case-summarization tools exist but no deployed product independently investigates duty refund claims and adjudicates penalty remission in production customs environments. |
Examine immigration applications, visas, and passports and interview persons to determine eligibility for admission, residence, and travel in the U.S.
20CI 20–20 · exposure 25 · augmentation 50 · importance 4.8/5 · click for rater detail
Examine immigration applications, visas, and passports and interview persons to determine eligibility for admission, residence, and travel in the U.S.
20| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | CBP has explored AI-assisted document screening and biometric tools, but actual adoption of autonomous decision-making remains minimal due to legal and liability constraints. Most initiatives are confined to supplementary scanning and flagging rather than replacing officer judgment. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Border agencies have adopted biometric and e-gate technology for pre-screening, but adoption of AI for the core interview and decision-making function remains slow and cautious due to security and legal stakes. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist officers by pre-screening documents, flagging inconsistencies, and summarizing application data, meaningfully reducing manual review time. However, the interview and final determination remain human-centric, limiting the productivity multiplier compared to tasks with less inherent human judgment. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI tools assist officers via document authentication, watchlist cross-referencing, and translation aids, improving throughput and accuracy while officers retain final decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Document review (applications, visas, passports) can be partially automated via OCR and rule-based verification, but the core task—interviewing persons to assess eligibility, detect fraud, and evaluate credibility—requires nuanced human judgment and adaptive questioning that current AI cannot reliably perform end-to-end at the required quality threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | Document verification and database checks can be partially automated, but the interview and admissibility judgment require contextual reasoning, deception detection, and discretionary legal judgment that current AI cannot reliably replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Immigration decisions are governed by federal law and must be made by authorized government officers; admission/residence determinations cannot legally be delegated to automated systems without human sign-off. Regulatory and legal barriers are absolute and unambiguous. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This is a sovereign law-enforcement function requiring sworn, legally authorized federal officers; admissibility decisions carry major legal and national security consequences, making human authority essentially non-negotiable. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | While document processing automation is cheap, the labor savings are offset by the need for human review, interview oversight, and liability management. AI integration costs and the requirement for trained human officers to conduct interviews keep the cost ratio unfavorable to full automation. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Automated document scanning is cheap, but the interview/judgment component still requires trained officers, keeping overall costs comparable to human labor when full task scope is considered. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed AI systems can assist with document classification and basic rule-checking, but no production system reliably handles the full interview and eligibility determination without substantial human oversight. The stakes and variability make current AI too error-prone for independent deployment. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Deployed systems like automated passport gates (e.g., e-gates, facial recognition kiosks) exist for routine document checks, but full interview-based eligibility determination is not handled autonomously by any production AI system. |
Interpret and explain laws and regulations to travelers, prospective immigrants, shippers, and manufacturers.
19CI 18–20 · exposure 25 · augmentation 63 · importance 4.4/5 · click for rater detail
Interpret and explain laws and regulations to travelers, prospective immigrants, shippers, and manufacturers.
19| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Government agencies, particularly those with statutory enforcement roles, adopt AI slowly and conservatively. CBP operates in a heavily regulated, legally-accountable environment where autonomous or AI-primary interpretation of law is not feasible or permitted under current governance structures. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Government/border security agencies are typically slow adopters of AI for frontline legal/enforcement interactions due to regulatory, security, and liability constraints. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI could assist officers by retrieving relevant regulatory text, summarizing case law, or drafting explanations that the officer then reviews and personalizes; this would improve reference speed and consistency without removing human judgment from the authoritative explanation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can meaningfully assist officers by quickly retrieving relevant regulations, precedents, and multilingual explanations, improving speed and consistency while the officer retains final authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can retrieve and summarize relevant laws and regulations, it struggles with the nuanced, context-dependent interpretation and real-time explanation required for diverse travelers and complex compliance scenarios. The task demands judgment about applicability and clear communication tailored to non-expert audiences, which current systems do poorly without extensive human oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft explanations of regulations but interpreting nuanced legal situations for specific travelers/shippers requires judgment, context verification, and authority that current systems cannot fully replace end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | CBP officers must be federally commissioned employees with statutory authority to enforce laws; only a licensed government official can legally interpret and apply regulations to travelers and shippers in an official capacity. Liability and legal accountability for incorrect guidance create near-absolute barriers to AI substitution. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This is a government law-enforcement function requiring sworn, authorized officers; legal interpretation and enforcement decisions must be made by credentialed personnel with legal authority. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Even assuming functional AI systems, the cost of building compliance-grade infrastructure, legal review, liability coverage, and required human oversight for correctness would approach or exceed the loaded cost of employing a CBP officer for this function. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI text generation is cheap, the need for human oversight, liability, and accuracy verification in legal interpretation keeps effective costs closer to human labor costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI chatbots can provide general regulatory information, but deployed solutions lack the legal authority, situational judgment, and accountability needed for official border/customs interpretation. No mature product reliably handles the full spectrum of edge cases and legal liability inherent in official regulatory guidance. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Chatbots and info portals exist for general customs guidance, but no deployed product reliably handles live, case-specific legal interpretation and explanation for individuals at ports of entry. |
Collect samples of merchandise for examination, appraisal, or testing.
13CI 0–25 · exposure 13 · augmentation 38 · importance 3.8/5 · click for rater detail
Collect samples of merchandise for examination, appraisal, or testing.
13| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Customs agencies operate in highly regulated, risk-averse sectors with slow AI adoption. While some risk-scoring and analytics tools are being piloted, actual deployment of automation in sample collection is minimal and moving slowly. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Border security and customs enforcement remain physically-grounded, government sectors with low digitization of frontline physical inspection tasks and minimal AI agent deployment for hands-on sampling. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI could assist officers by predicting which shipments warrant sampling, flagging high-risk containers, or organizing documentation, thereby raising officer efficiency. However, the assistance is limited to decision support rather than transforming the core collection task itself. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI can assist with logistics, risk-flagging of shipments, or documentation after sampling, but offers little direct assistance to the physical act of collecting the sample itself. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI systems could assist in identifying which merchandise to sample and prioritizing based on risk profiles, the actual physical collection of samples requires human presence at borders and ports. Current AI cannot perform the hands-on sampling, verification, and chain-of-custody documentation independently. |
| Task automatability | claude-sonnet-5 | 1/5 | Physically collecting samples of merchandise requires physical manipulation, mobility, and presence at inspection sites, which current AI systems cannot perform without robotic embodiment far beyond deployed capability. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Federal law and CBP regulations require authorized officers to collect and certify samples for legal and evidentiary purposes. Chain-of-custody requirements and the need for human attestation create hard legal barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | This task is tied to law enforcement authority, chain-of-custody requirements, and legal search/seizure protocols that require an authorized officer to perform or directly supervise it. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The loaded cost of a Customs officer (~$80–100K annually) vastly exceeds any current AI system's cost for the sampling-specific components. Physical presence and legal authority cannot be substituted, keeping human labor the dominant cost. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this physical task, so cost comparison favors the human by default since AI cannot yet deliver the output at all. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed AI systems independently perform physical sample collection at borders today. Risk assessment and prioritization tools exist in pilot form at some agencies, but the core task of physically selecting and collecting samples remains human-dependent with limited automation. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs physical sample collection of merchandise for customs inspection; this remains a manual, hands-on task performed by officers. |
Locate and seize contraband, undeclared merchandise, and vehicles, aircraft, or boats that contain such merchandise.
6CI 0–11 · exposure 5 · augmentation 50 · importance 4.6/5 · click for rater detail
Locate and seize contraband, undeclared merchandise, and vehicles, aircraft, or boats that contain such merchandise.
6| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Border agencies have adopted AI for initial screening and risk scoring at scale, but actual field seizure operations remain wholly human-driven. Adoption of AI assistance tools is present but slow, with no observable displacement of officer seizure functions in production environments. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Border security and physical law enforcement are low-digitization, high-friction sectors with minimal AI-driven task displacement in the physical seizure function itself. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist officers by prioritizing shipments for inspection, flagging high-risk patterns in manifests, or enhancing imaging during physical searches. However, the core judgment and execution of seizure remain human, and augmentation is limited to pre- and during-inspection support. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-assisted screening, risk scoring, image analysis (e.g., X-ray/scanner interpretation) and data matching can help officers identify likely targets, improving efficiency of the search process even though execution remains human. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Locating and seizing contraband requires real-time physical inspection, judgment under uncertainty, interaction with people, and legal authority to detain—capabilities that current AI systems cannot perform end-to-end in the field. AI can assist with screening data or image analysis, but cannot independently conduct searches or make seizure decisions. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a physical enforcement task requiring in-person inspection, searches, seizure authority, and physical apprehension of goods/vehicles; no AI system today can perform the physical search and seizure itself. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Only licensed Customs and Border Protection officers have legal authority to conduct searches and seize property; no AI system can perform this function without human sign-off. Liability for wrongful seizure, Fourth Amendment constraints, and regulatory authority vested in personnel create hard barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Seizure and search authority is a legally mandated law enforcement power requiring sworn, authorized officers; this is a hard legal/regulatory barrier preventing automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce some screening costs (e.g., image analysis of cargo), but the comprehensive task—physical location, inspection, interdiction, legal seizure—still requires human officers and cannot be delegated to automation. Full-system cost remains dominated by human labor. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the physical officer performing searches and seizures, so there is no viable AI cost basis for this task—human labor remains mandatory. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI systems can support contraband detection through imaging and pattern recognition in controlled lab settings, deployed systems lack the contextual judgment, legal standing, and reliability needed for actual field seizure operations. Some ports use screening AI as an input, but humans remain the decision-maker and executor. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously locates and seizes contraband; AI is at most used for risk-flagging/screening in supporting systems, not for the physical task described. |
Detain persons found to be in violation of customs or immigration laws and arrange for legal action, such as deportation.
0CI 0–0 · exposure 0 · augmentation 38 · importance 4.6/5 · click for rater detail
Detain persons found to be in violation of customs or immigration laws and arrange for legal action, such as deportation.
0| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | No automation or replacement is occurring or will occur in the near term; detention and deportation decisions are legally protected as human functions requiring accountability and discretion in a sensitive law enforcement context. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Border security and law enforcement are low-digitization, physically-grounded sectors with minimal AI adoption for the enforcement act itself, though supporting data systems are used. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI might assist with document review, case history lookups, or generating deportation paperwork templates, but the core decision-making and custody authority remain entirely human-driven; augmentation is limited to peripheral support. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist by flagging risk profiles, cross-referencing watchlists, and drafting paperwork for legal action, but the actual detention and adjudication decision remains human-driven. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires exercising law enforcement authority, making discretionary judgment calls on custody, legal status, and deportation proceedings. Current AI cannot make binding legal determinations or lawfully detain individuals; these are inherently human functions requiring legal accountability and the ability to exercise discretion in complex, fact-specific cases. |
| Task automatability | claude-sonnet-5 | 1/5 | Detaining persons and initiating legal proceedings requires physical enforcement authority, judgment about legal thresholds, and coercive power that AI cannot exercise; no automation of the core act is possible. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | This task has the highest barriers: only licensed law enforcement officers with federal authority can lawfully detain persons or initiate deportation. Legal liability, constitutional protections, and statutory requirements (8 U.S.C. § 1357) strictly limit who can perform these functions. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This is a sworn law-enforcement power requiring statutory authority, badge/certification, and legal accountability; only authorized officers may detain and initiate deportation, an absolute legal barrier. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI cannot perform the core legal and enforcement functions of this task at all, making cost comparison moot; a human officer is mandatory for detention authority and legal proceedings. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this function, so no meaningful cost comparison exists; the human officer is the only viable option. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product can legally or reliably perform detention, custody decisions, or arrange deportation proceedings. These actions require licensed law enforcement authority, legal review, and court involvement that AI systems cannot discharge. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product performs physical detention or independently arranges deportation proceedings; this remains entirely a human law-enforcement function. |
Institute civil and criminal prosecutions and cooperate with other law enforcement agencies in the investigation and prosecution of those in violation of immigration or customs laws.
0CI 0–0 · exposure 0 · augmentation 38 · importance 4.3/5 · click for rater detail
Institute civil and criminal prosecutions and cooperate with other law enforcement agencies in the investigation and prosecution of those in violation of immigration or customs laws.
0| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | No meaningful AI adoption is occurring in prosecution or inter-agency criminal coordination because these are legally reserved functions. Even information-sector law firms rely on human attorneys for all prosecution decisions. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Federal law enforcement and prosecutorial functions are heavily regulated and slow to adopt automation for core legal authority tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI can modestly assist with legal research, document review, or evidence organization prior to prosecution, but the core task—deciding whether to prosecute, negotiating with agencies, and courtroom argument—remains entirely human-driven. Augmentation potential is limited by the task's irreducibly legal and discretionary nature. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with case documentation, evidence organization, and research support, but the core prosecutorial and investigative coordination remains human-driven. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires legal judgment, prosecutorial discretion, courtroom testimony, and complex interpersonal negotiation with other agencies. Current AI cannot independently decide whether to prosecute, what charges to bring, or how to cooperate strategically with law enforcement—these demand human legal expertise and accountability. |
| Task automatability | claude-sonnet-5 | 1/5 | Instituting prosecutions and cross-agency criminal investigation cooperation requires legal authority, judgment, and interagency human coordination that AI cannot perform end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | This task is protected by hard legal barriers: only licensed attorneys or authorized government officials can prosecute; prosecutorial decisions carry personal and institutional liability; and immigration/customs prosecution is explicitly regulated. A human must legally perform and sign off on all material decisions. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This is a law-enforcement and legal function requiring sworn authority, statutory power, and judicial oversight—hard legal barriers prevent AI from instituting prosecutions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI cannot perform this task at all, so cost comparison is inapplicable; a human prosecutor/officer is mandatory by law and function. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this function, so cost comparison favors the human officer/legal system entirely. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product can autonomously institute civil or criminal prosecutions or lead multi-agency investigation coordination. These are inherently human-performed legal and executive functions that require attorney licensure, prosecutorial authority, and institutional accountability. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product initiates or conducts prosecutions or coordinates law enforcement investigations autonomously; this remains firmly a human institutional function. |
Testify regarding decisions at immigration appeals or in federal court.
0CI 0–0 · exposure 0 · augmentation 38 · importance 4.2/5 · click for rater detail
Testify regarding decisions at immigration appeals or in federal court.
0| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | There is no adoption velocity to measure because the task cannot be automated; it is legally mandated to be performed by a qualified human officer in a court or appellate setting. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Judicial and immigration appeal processes are highly conservative, procedure-bound, and have not adopted AI for testimonial roles. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI can assist by summarizing case files, drafting background documents, or organizing evidence before testimony, but it cannot augment the core act of testifying itself, which remains entirely human. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help officers prepare testimony by organizing case files, summarizing prior decisions, or drafting talking points, but it does not touch the act of testifying itself. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Testimony in court or immigration appeals requires a human expert to be present, answer questions under oath, and make real-time credible judgments. No AI system can legally or procedurally substitute for a human witness or decision-maker in this context. |
| Task automatability | claude-sonnet-5 | 1/5 | Testifying requires a real officer physically present, sworn under oath, subject to cross-examination and personal accountability, none of which AI can perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Federal law and court rules explicitly require a human witness to testify in person, answer questions, and subject themselves to cross-examination. This is a hard legal and procedural barrier that cannot be circumvented. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Legal systems require testimony from an authorized human witness under oath, with perjury liability and due-process rules that categorically bar AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The task requires a qualified officer to physically appear and testify; there is no AI cost comparison because no AI substitute exists or is legally permissible. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute for sworn testimony, so cost comparison is moot; any attempted automation would still require human presence, yielding no savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | While AI can draft documents or summarize records, it cannot serve as a legal witness or testify under oath in any deployed system. This is fundamentally a human-only legal requirement. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product provides courtroom or tribunal testimony on behalf of a government officer; this is purely research-stage or nonexistent territory. |
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.