Detectives and Criminal Investigators

33-3021.00
Median wage $93,790/yr114,430 employed (US)Rank #777 of 923 scored · top 84% by substitution

Conduct investigations related to suspected violations of federal, state, or local laws to prevent or solve crimes.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution16
Exposure17
Augmentation44

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

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

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

Tasks on the substitution scale

30 rated tasks, binned by substitution score.

Position among all scored occupations

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

Tasks with substitution ≥ 70

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%17

panel mean rating 1.7/5 → substitution pressure 17/100

Technical feasibility todayw 20%16

panel mean rating 1.6/5 → substitution pressure 16/100

Cost vs. human wagew 15%19

panel mean rating 1.8/5 → substitution pressure 19/100

Adoption barriersw 20%inverted — strong barriers lower the score11

panel mean rating 4.5/5 (barrier strength) → substitution pressure 11/100

Sector adoption velocityw 10%13

panel mean rating 1.5/5 → substitution pressure 13/100

Task breakdown (30 tasks)

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

Examine records and governmental agency files to find identifying data about suspects.

60

CI 46–74 · exposure 62 · augmentation 75 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Law enforcement and federal agencies have been rapidly adopting automated records-search systems, facial recognition, and AI-assisted data matching over the past five years, with substantial production deployments in major departments and federal services.
Sector adoption velocityclaude-sonnet-52/5Law enforcement agencies are historically slow technology adopters due to budget constraints, procurement processes, and legal/privacy scrutiny, so despite available tools, actual deployment remains uneven and cautious.
Augmentation potentialclaude-haiku-4-5-202510014/5AI dramatically assists investigators by surfacing matches, aggregating cross-agency data, and highlighting patterns that would take hours to manually compile, freeing investigators to focus on analysis and followup—while the human remains essential for final judgment and legal review.
Augmentation potentialclaude-sonnet-54/5AI-powered search, record aggregation, and pattern-matching tools significantly speed up a detective's ability to locate and cross-reference identifying data, meaningfully boosting investigative productivity while the detective retains judgment and legal responsibility.
Task automatabilityclaude-haiku-4-5-202510014/5AI systems can automatically extract, match, and cross-reference identifying data across records and files at scale with high speed and accuracy, easily exceeding 50% time savings. However, the task retains some investigative judgment in prioritizing which records are relevant and interpreting ambiguous matches, preventing a full 5-rating.
Task automatabilityclaude-sonnet-53/5AI can rapidly search, cross-reference, and extract identifying data from structured records and databases, but integrating disparate governmental systems, verifying accuracy, and legal chain-of-custody still require human involvement, limiting full automation to roughly half the workflow.
Adoption barriersclaude-haiku-4-5-202510013/5Adoption is constrained by regulatory oversight (FCRA, ECPA, state privacy laws governing records access), government procurement friction, and the requirement that findings typically be reviewed and verified by a licensed investigator before use in prosecution—these provide moderate friction without a hard legal prohibition on automation.
Adoption barriersclaude-sonnet-54/5Access to many governmental and criminal justice databases is legally restricted to authorized personnel, and chain-of-custody/evidentiary rules require documented human accountability for investigative findings.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI-powered record search and data extraction costs a fraction of human investigator time per query; the cost per identified record is orders of magnitude lower than paying a detective to manually search files.
Cost vs. human wageclaude-sonnet-54/5Automated database queries and record matching are far cheaper per search than manual file review by a detective, though licensing fees for specialized databases and oversight costs temper the savings somewhat.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed products for records management, data extraction, and identity matching (used by law enforcement and government agencies) perform this task reliably at scale, though integration with legacy governmental systems and occasional disambiguation challenges keep it below a perfect 5.
Technical feasibility todayclaude-sonnet-53/5Records-search and entity-resolution tools are deployed in law enforcement and OSINT products (e.g., Palantir, Clearview, LexisNexis Accurint), but accuracy varies and access to siloed government databases is inconsistent, so reliability in production is moderate rather than uniform.

Analyze completed police reports to determine what additional information and investigative work is needed.

51

CI 23–80 · exposure 58 · augmentation 88 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Police departments are adopting AI tools at middling pace—some large metro departments run pilots on report analysis and case triage, but production rollout is slow due to organizational conservatism, budget constraints, and union concerns. Broader law enforcement adoption lags compared to finance or tech sectors.
Sector adoption velocityclaude-sonnet-51/5Law enforcement is a slow-adopting, highly regulated sector with limited AI deployment for core investigative decision-making, constrained by budgets, legal risk, and public scrutiny.
Augmentation potentialclaude-haiku-4-5-202510015/5AI can dramatically assist detectives by automatically flagging missing information, suggesting investigative leads, and organizing case facts—allowing investigators to focus on complex judgment calls and interviews rather than rereading lengthy reports. This transformative augmentation is already demonstrated in practice.
Augmentation potentialclaude-sonnet-54/5AI can effectively assist by summarizing reports, cross-referencing databases, and highlighting inconsistencies or missing elements, meaningfully speeding up a detective's review process.
Task automatabilityclaude-haiku-4-5-202510015/5AI can systematically review police reports, identify gaps, flag missing witness statements or evidence, and recommend follow-up steps faster and more consistently than human investigators. Modern LLMs can extract key facts, cross-reference against investigation checklists, and generate prioritized lists of needed information with >50% time savings at comparable quality.
Task automatabilityclaude-sonnet-52/5AI can summarize and flag gaps in reports but determining investigative next steps requires legal judgment, contextual case knowledge, and accountability that current systems cannot reliably provide end-to-end.'
Adoption barriersclaude-haiku-4-5-202510013/5Police agencies have custody requirements and procedural protocols around evidence and investigative decisions; some jurisdictions require a sworn officer to sign off on investigation direction. However, AI analysis as a preliminary screening or recommendation layer faces modest adoption friction rather than hard legal bars.
Adoption barriersclaude-sonnet-54/5Investigative decisions carry legal and evidentiary weight, chain-of-custody and due-process requirements, and typically require sworn officer authorization, creating strong barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510015/5Inference costs for analyzing a police report (typically <$0.01 per document) plus integration overhead are orders of magnitude cheaper than paying a detective 1–2 hours of labor ($40–80 loaded cost) to perform the same review.
Cost vs. human wageclaude-sonnet-52/5AI could cheaply flag missing fields or inconsistencies, but the core judgment task still requires detective review, keeping overall cost savings modest.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed AI systems (LLM-based document analysis tools, legal document review platforms) reliably perform structured analysis of police reports in production environments. Some agencies have pilot programs; error rates on gap-detection are low, though edge cases in complex investigations may require human review.
Technical feasibility todayclaude-sonnet-52/5Some police departments pilot AI report-review/summarization tools, but no mature product reliably determines investigative next steps at scale in production.

Notify command of situation and request assistance.

49

CI 23–75 · exposure 58 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Law enforcement agencies are adopting AI-assisted case management and alert systems moderately; pilots and partial deployments are common, but full automation of command notification remains primarily supervisory and oversight-heavy rather than fully hands-off in production.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field operations are a low-digitization, physically embedded sector with minimal AI agent adoption for real-time tactical communication tasks.
Augmentation potentialclaude-haiku-4-5-202510014/5AI substantially assists detectives by auto-generating structured incident summaries, flagging relevant case facts, and routing assistance requests to the right command channel, enabling faster and more accurate notifications while the detective retains decision authority over what is reported and when.
Augmentation potentialclaude-sonnet-52/5AI-enabled dispatch and transcription tools can log or relay information, but they offer limited productivity enhancement for this instantaneous, judgment-driven communication task.
Task automatabilityclaude-haiku-4-5-202510015/5Notifying command of situation and requesting assistance is a straightforward information relay task that AI systems can handle reliably by converting case details into structured incident reports and routing requests to appropriate oversight channels, easily achieving >50% time savings over manual formulation and submission.
Task automatabilityclaude-sonnet-52/5This is a brief communicative act requiring real-time judgment about situational urgency and tactical details; AI could draft or relay messages but cannot independently assess and decide when/how to request assistance in dynamic field situations.dns.gov.br.gov.br .
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: command-and-control hierarchies, chain-of-custody and accountability requirements, legal liability for mischaracterized situations, and organizational protocols that typically require a sworn officer to formally request assistance and sign off on incident reports, preventing full substitution.
Adoption barriersclaude-sonnet-54/5Chain-of-command protocols, officer safety, and legal accountability for law enforcement actions require a sworn human to make and communicate this real-time judgment call, creating strong organizational and legal barriers.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automating notification and assistance requests costs a fraction of a detective's loaded wage—essentially API calls and minor integration overhead—making it at least an order of magnitude cheaper than paying a human to compose and route the same information.
Cost vs. human wageclaude-sonnet-52/5Human officers already perform this as a near-zero-marginal-cost verbal action integrated into their duties; deploying AI to replace this offers no meaningful cost savings and adds integration overhead.
Technical feasibility todayclaude-haiku-4-5-202510014/5Mature deployed products (AI-powered incident management systems, case management automation, alert routing) perform this reliably in law enforcement organizations today, though some customization and human oversight remain standard practice in production deployments.
Technical feasibility todayclaude-sonnet-52/5Radio/dispatch communication systems exist but no deployed AI product autonomously performs the judgment-based act of notifying command during active investigations or incidents.

Examine records to locate links in chains of evidence or information.

37

CI 29–45 · exposure 38 · augmentation 75 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Law enforcement agencies are adopting e-discovery and data analytics tools, but integration is uneven, pilots are common, and full agent-based automation in production remains rare due to liability and institutional conservatism in criminal justice.
Sector adoption velocityclaude-sonnet-52/5Law enforcement agencies are generally slow adopters of advanced AI tools due to budget constraints, procurement processes, and legal/ethical scrutiny, though some large agencies use analytic software.
Augmentation potentialclaude-haiku-4-5-202510014/5AI excels at rapidly searching and cross-referencing large volumes of records to surface candidate connections; investigators then examine and validate these suggestions, materially accelerating the evidence-linking workflow while keeping human judgment in the loop.
Augmentation potentialclaude-sonnet-54/5AI-powered link analysis and data mining tools significantly speed up an investigator's ability to surface connections across disparate records, materially boosting productivity while the investigator retains interpretive and legal responsibility.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist in searching and pattern-matching across structured records, but examining chains of evidence requires contextual judgment, understanding of legal admissibility, and domain expertise that AI struggles with reliably. Current systems cannot autonomously verify evidentiary integrity or make binding determinations about linkage sufficiency.
Task automatabilityclaude-sonnet-53/5AI can search, cross-reference, and flag connections across large record sets (phone logs, financial records, databases) quickly, but validating evidentiary chains for legal use requires human judgment and contextual investigative reasoning that current systems cannot fully replace.
Adoption barriersclaude-haiku-4-5-202510014/5Criminal investigation is heavily regulated; chains of evidence must meet strict legal and procedural standards, and investigators must take responsibility for findings in courtroom contexts. Liability asymmetry and legal/regulatory requirements create strong friction against full automation.
Adoption barriersclaude-sonnet-54/5Chain-of-evidence handling is subject to legal admissibility rules, chain-of-custody requirements, and departmental accountability, requiring sworn personnel to attest to findings, which strongly limits full automation.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI-powered search and linkage tools reduce manual record review labor, making costs comparable to or modestly cheaper than human-only review, but investigators remain required for validation and judgment, so full replacement economics do not apply.
Cost vs. human wageclaude-sonnet-53/5Software licensing and data integration costs are substantial, and human review/oversight is still required, making the cost savings moderate rather than dramatic compared to investigator time.
Technical feasibility todayclaude-haiku-4-5-202510012/5While e-discovery and database search tools exist and are deployed, they serve as aids requiring investigator interpretation rather than end-to-end autonomous performance. AI products cannot reliably examine evidence chains without human verification of legal and investigative standards.
Technical feasibility todayclaude-sonnet-53/5Link-analysis and data-fusion tools (e.g., Palantir, i2 Analyst's Notebook, AI-enhanced case management systems) are deployed in law enforcement, but they assist analysts rather than autonomously determining evidentiary links reliably.

Prepare reports that detail investigation findings.

31

CI 25–37 · exposure 33 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement remains relatively early in digital transformation and AI adoption; regulatory conservatism, liability concerns, and institutional resistance to outsourcing investigative judgment limit production deployment of report automation.
Sector adoption velocityclaude-sonnet-52/5Law enforcement is a traditionally slow-adopting sector with strict data security, evidentiary, and procedural constraints, resulting in limited and cautious AI tool rollout so far.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist investigators by drafting sections, organizing evidence, and suggesting structure, boosting report-writing speed and consistency. However, the investigator must validate findings and reasoning, limiting the transformative impact.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully speed up report drafting by transcribing interviews, organizing timelines, and generating structured summaries for the investigator to verify and finalize.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with structuring findings and generating draft text, the task requires synthesizing complex investigative judgment, evidence interpretation, and legal/evidentiary standards that demand human review and sign-off. Current AI cannot reliably replace the investigator's core analytical work.
Task automatabilityclaude-sonnet-53/5AI can draft narrative reports from structured notes, transcripts, or evidence logs, saving significant time, but final reports require judgment about legal sufficiency, chain of custody nuances, and case-specific interpretation that still needs human review.
Adoption barriersclaude-haiku-4-5-202510014/5Investigation reports are legally binding documents that form evidence in court and must be authored/certified by licensed law enforcement personnel; liability, evidentiary rules, and chain-of-custody standards create strong legal and organizational barriers to full automation.
Adoption barriersclaude-sonnet-54/5Investigative reports are legal documents that may be used in court, requiring accuracy, chain-of-custody integrity, and officer accountability/signature, creating strong institutional and legal barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI writing assistance is cheap, but the human investigator remains the primary cost; oversight and revision time offset savings from draft automation, keeping total cost per report near or above the investigator's loaded wage.
Cost vs. human wageclaude-sonnet-53/5AI drafting tools are cheap per report, but mandatory human review, editing for accuracy, and liability checks add cost, making total cost roughly comparable to a detective's own drafting time saved.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI writing tools exist and can help draft sections, but no deployed system reliably produces investigation reports meeting law enforcement standards without substantial human rework. Products perform narrow subtasks (formatting, summarization) rather than the full investigative reporting task.
Technical feasibility todayclaude-sonnet-52/5Some law enforcement agencies pilot AI-assisted report writing tools (e.g., transcription-to-report drafting), but adoption is limited, error-prone, and not yet standard in production across most departments.

Prepare charges or responses to charges, or information for court cases, according to formalized procedures.

26

CI 20–32 · exposure 33 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While some prosecutor offices pilot AI drafting tools, production deployment for autonomous charge preparation remains rare. Regulatory and professional liability concerns, combined with the legal profession's slower digitization, keep real-world adoption limited to pilot phases in forward-thinking jurisdictions.
Sector adoption velocityclaude-sonnet-52/5Law enforcement and criminal justice systems are historically slow to adopt AI due to regulatory scrutiny, evidentiary standards, and institutional caution, with pilots emerging but production use rare.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist prosecutors and detectives by drafting boilerplate sections, suggesting relevant charges based on statute search, and summarizing case facts for document preparation. However, the human investigator or prosecutor must retain full control over legal judgment and final formulation, making augmentation helpful but bounded by the requirement for human expertise.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully speed up drafting of formalized language, summarizing case files, and ensuring procedural completeness, significantly aiding investigators while they retain responsibility for accuracy and judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can draft legal documents and assist with legal research, preparing formal charges and court responses requires human judgment on legal strategy, precedent application, and compliance with jurisdiction-specific procedures. Current AI systems cannot reliably handle the full end-to-end task with the consistency and liability tolerance courts demand.
Task automatabilityclaude-sonnet-53/5AI can draft charge language, summarize evidence, and populate formalized legal templates from structured case data, but final accuracy, evidentiary sufficiency, and legal judgment require human verification, limiting full end-to-end automation.-
Adoption barriersclaude-haiku-4-5-202510015/5Preparing formal charges and court documents is a licensed legal function; prosecutors and their legal staff are often required by law to review and certify these documents. Liability for errors in court filings creates hard barriers to full automation, and human attorney sign-off is typically mandatory.
Adoption barriersclaude-sonnet-55/5Formal charges and court filings require certified, authorized personnel (detectives, prosecutors) to prepare and sign off, with strict legal and procedural liability making human authorization mandatory.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI legal tools still require significant human oversight to ensure correctness and liability-appropriate output. The all-in cost of AI inference plus mandatory human review and legal responsibility remains comparable to or exceeds the cost of an attorney or prosecutor handling the task directly.
Cost vs. human wageclaude-sonnet-53/5AI drafting tools are cheap per document, but required legal review, chain-of-custody accuracy, and oversight by trained investigators/prosecutors keep total cost roughly comparable to current human-driven processes.
Technical feasibility todayclaude-haiku-4-5-202510012/5Legal drafting tools exist and some prosecutors use AI-assisted document generation, but no mature product reliably prepares formal charges or court responses without substantial human oversight and revision. Error rates remain material in jurisdictional compliance and evidentiary framing.
Technical feasibility todayclaude-sonnet-52/5Legal drafting assistants exist and are used for document preparation, but reliable, production-grade tools specifically validated for criminal charge preparation in law enforcement workflows are not widely deployed.

Identify case issues and evidence needed, based on analysis of charges, complaints, or allegations of law violations.

25

CI 25–25 · exposure 25 · augmentation 63 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement adoption of AI for investigative decision-making remains nascent; most agencies still rely on manual case review and detective judgment. Pilots exist but production deployment of AI-driven case analysis is rare, reflecting risk aversion, budget constraints, and lack of proven solutions at scale.
Sector adoption velocityclaude-sonnet-52/5Law enforcement and criminal justice sectors are historically slow to adopt AI due to legal, ethical, and public trust concerns, with pilots more common than deployed production use.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully assist detectives by organizing documents, flagging patterns, surfacing prior complaints, and highlighting potential gaps in evidence collection—productivity gains without replacing investigator judgment. However, the assistance is partial and context-dependent rather than transformative.
Augmentation potentialclaude-sonnet-54/5AI tools can meaningfully assist by organizing case files, cross-referencing statutes, and surfacing similar prior cases or evidence patterns, significantly aiding investigators while they retain decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist in document review and pattern recognition across case files, the task requires judgment about legal sufficiency of charges, relevance of evidence under context-specific law, and investigative priorities—decisions that involve discretion, legal interpretation, and domain knowledge that current AI cannot reliably execute end-to-end. AI might flag potential issues but cannot independently determine case strategy or evidence priorities.
Task automatabilityclaude-sonnet-52/5AI can help summarize complaints and flag relevant statutes or evidence categories, but determining actual case strategy and evidentiary needs requires contextual judgment, legal reasoning, and investigative experience that current systems cannot reliably replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Criminal investigation findings often form the basis for prosecution and arrest, creating high liability stakes and regulatory oversight by prosecutors, courts, and law enforcement hierarchies. Human supervisory sign-off, chain-of-custody rules, and evidentiary standards create strong legal and organizational barriers to unsupervised automation.
Adoption barriersclaude-sonnet-54/5Investigative determinations often carry legal and constitutional weight (e.g., probable cause, chain of custody), requiring sworn officers or licensed investigators, creating strong institutional and legal barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Integration costs (case management systems, legal domain tuning, oversight) and the need for human review of AI recommendations keep total cost per analysis comparable to or higher than a junior investigator's time, especially when errors trigger liability or rework.
Cost vs. human wageclaude-sonnet-52/5While AI-assisted document review is cheap per page, the human oversight, verification, and legal accountability required for investigative analysis keep the effective all-in cost close to human-level for this judgment-heavy task.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs independent case analysis and evidence prioritization at the level required by law enforcement. Tools exist for document management and keyword search, but production systems do not autonomously identify legal issues or recommend investigative direction with the judgment and accuracy criminal cases demand.
Technical feasibility todayclaude-sonnet-52/5Some legal-tech and investigative-support tools exist for document review and case triage, but no deployed product independently identifies case issues and evidentiary needs at production-grade reliability in law enforcement settings.

Record progress of investigation, maintain informational files on suspects, and submit reports to commanding officer or magistrate to authorize warrants.

23

CI 20–25 · exposure 25 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement adoption of AI for investigative workflows remains limited and cautious; most agencies use legacy case management systems with minimal AI integration. Regulatory and liability concerns keep adoption slow despite digitization of records.
Sector adoption velocityclaude-sonnet-52/5Law enforcement is a traditionally slow-adopting sector for AI due to legal, evidentiary, and public trust concerns, with pilots for report-writing assistance emerging but production use still limited.
Augmentation potentialclaude-haiku-4-5-202510013/5AI could assist investigators by summarizing case notes, flagging inconsistencies, and drafting report outlines, moderately improving their productivity in documentation tasks. However, the human must retain full control over findings and legal conclusions, limiting the degree of transformation.
Augmentation potentialclaude-sonnet-53/5AI can help draft narrative summaries, organize case files, and speed up documentation, providing real but partial productivity gains while the officer remains responsible for accuracy and legal compliance.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist with organizing case notes and drafting report templates, but the task requires substantive judgment about investigative progress, suspect assessment, and legal sufficiency that investigators must validate. End-to-end automation without human review would introduce liability and evidentiary risks that prevent meaningful time savings at equal quality.
Task automatabilityclaude-sonnet-52/5Drafting portions of reports and organizing case notes can be assisted by AI, but synthesizing investigative judgment, verifying facts, and preparing legally sufficient warrant affidavits requires human expertise and accountability that current AI cannot fully replace end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Warrants must be authorized by a magistrate or commanding officer based on probable cause affidavits prepared by the investigator; legal liability for inadequate or false statements in warrant applications creates a hard requirement that a human investigator review and take responsibility for the report content. Rules of evidence and chain-of-custody documentation further restrict automation.
Adoption barriersclaude-sonnet-55/5Warrant applications and sworn investigative reports must legally be authored and certified by a licensed officer and often require magistrate review, making this a hard legal/procedural barrier to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI systems for legal document and investigative reporting require significant human review and integration overhead, making their all-in cost comparable to or higher than a detective spending 30–40% of their time on paperwork. Automation savings are marginal.
Cost vs. human wageclaude-sonnet-52/5AI tools can cut some drafting/transcription time cheaply, but the need for a sworn officer's review, accuracy verification, and legal accountability means overall cost savings are modest once oversight is factored in.
Technical feasibility todayclaude-haiku-4-5-202510012/5Document management and basic report generation tools exist, but no deployed system reliably handles the investigative judgment, legal standards for warrant affidavits, and case-specific reasoning required. Products in this space remain limited in scope and require substantial human oversight.
Technical feasibility todayclaude-sonnet-52/5Some law enforcement agencies use case management software with AI-assisted summarization or transcription, but no deployed product reliably drafts complete, legally sound investigative reports or warrant applications at scale.

Provide information to lab personnel concerning the source of an item of evidence and tests to be performed.

21

CI 18–25 · exposure 20 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement remains a relatively low-digitization sector with strong institutional resistance to algorithmic autonomy in criminal procedures. AI adoption in investigations is still largely pilot-stage, not production-at-scale displacement.
Sector adoption velocityclaude-sonnet-51/5Law enforcement and forensic evidence handling are low-digitization, highly procedural environments with minimal AI agent adoption for such liaison tasks.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist detectives by retrieving relevant test protocols, drafting routine evidence communication templates, and suggesting standard lab procedures based on evidence type. However, augmentation is limited because final decisions require detective expertise and legal accountability.
Augmentation potentialclaude-sonnet-53/5AI can help draft evidence summaries, structure requests, or auto-fill lab request forms, improving efficiency while the investigator still supplies and verifies the substantive information.
Task automatabilityclaude-haiku-4-5-202510012/5While AI could draft routine communications about evidence sources and standard tests, criminal investigation requires context-specific judgment, chain-of-custody integrity, and nuanced understanding of case-specific factors. Current AI cannot reliably handle the full task end-to-end with 50% time savings at equal quality.
Task automatabilityclaude-sonnet-52/5This task requires an investigator to communicate case-specific context, chain-of-custody details, and judgment calls about testing priorities, which AI cannot originate though it could help draft the communication.'} ...cannot fully replace the human decision and liaison role.'.
Adoption barriersclaude-haiku-4-5-202510014/5Criminal evidence handling is heavily regulated; detectives are legally and professionally accountable for evidence integrity and proper documentation. Chain-of-custody requirements and evidentiary admissibility standards create hard barriers to full automation and require licensed investigator sign-off.
Adoption barriersclaude-sonnet-54/5Chain-of-custody, evidentiary integrity, and legal accountability requirements mean a sworn investigator must be the one providing and vouching for this information.
Cost vs. human wageclaude-haiku-4-5-202510012/5The cost of AI-generated communication errors in criminal investigation (chain-of-custody breaks, misidentified evidence) far exceeds savings, and human review remains essential. Integration and oversight costs are substantial relative to the narrow time savings possible.
Cost vs. human wageclaude-sonnet-52/5Since AI cannot independently perform this liaison task, any cost comparison favors the human investigator who must remain involved for legal and procedural accuracy.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs this task autonomously. AI systems can assist with information retrieval and drafting, but the task demands human detective judgment, legal compliance verification, and authenticated communication that deployed AI systems cannot handle independently.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously interfaces with forensic lab personnel to convey evidentiary context and testing requirements in real investigative workflows.

Obtain and verify evidence by interviewing and observing suspects and witnesses or by analyzing records.

18

CI 11–25 · exposure 17 · augmentation 63 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement is traditionally slow to adopt automation, and evidence gathering remains heavily human-driven in practice. While some departments pilot AI for record search and case file management, interviewing and observation remain core investigator functions with minimal displacement today.
Sector adoption velocityclaude-sonnet-52/5Law enforcement is a traditionally slow-adopting, highly regulated sector with cautious integration of AI due to legal and civil liberties concerns.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by organizing records, flagging inconsistencies in witness statements, summarizing case files, and suggesting follow-up lines of inquiry, raising investigator efficiency. However, the human remains responsible for the actual interview, credibility judgment, and evidence synthesis.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist with transcription, record cross-referencing, pattern detection, and summarizing evidence, boosting investigator efficiency while humans retain interview and judgment duties.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time human interaction, judgment about credibility, and legal compliance that cannot be substituted by current AI systems end-to-end. Interviewing involves reading nonverbal cues, establishing rapport, and making tactical decisions about questioning that remain fundamentally human capabilities.
Task automatabilityclaude-sonnet-52/5Interviewing suspects/witnesses and reading human behavioral cues requires judgment, adaptive questioning, and legal sensitivity that current AI cannot reliably replicate; record analysis portions are more automatable but the core interpersonal task is not.
Adoption barriersclaude-haiku-4-5-202510014/5Legal and liability barriers are substantial: statements obtained by an AI agent lack evidentiary weight, interviews must be conducted by authorized personnel for admissibility, and rules of evidence require human judgment. Regulatory frameworks (Fourth Amendment, evidence rules, interrogation law) require a licensed investigator in control.
Adoption barriersclaude-sonnet-55/5Evidence gathering and interviews are governed by strict legal, chain-of-custody, and due-process requirements, typically requiring sworn officers with legal authority.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for evidence analysis and records review are cost-effective for narrow document-processing subtasks, but the full task of interviewing, observing, and synthesizing evidence still depends on investigator labor. Integration overhead and need for human oversight keep total cost near or above human-alone approaches.
Cost vs. human wageclaude-sonnet-52/5AI tools can cheaply process documents, but the interview/verification component still requires a trained investigator, so overall cost savings are limited.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI can assist with record analysis and document review, but no deployed product reliably conducts interviews or credibility assessment independent of human investigators. Narrow applications (record search, initial document classification) exist, but the core task—obtaining evidence through interaction and synthesis—requires human execution.
Technical feasibility todayclaude-sonnet-52/5Some products assist with record/document analysis and transcription, but no deployed system conducts investigative interviews or verifies evidence credibility reliably in real casework.

Note, mark, and photograph location of objects found, such as footprints, tire tracks, bullets and bloodstains, and take measurements of the scene.

17

CI 9–25 · exposure 20 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Forensic agencies and police departments adopt AI and automation more slowly than white-collar sectors; budgets are tight, processes are conservative due to legal risk, and integration into existing case management systems is slow. Most departments still rely on manual photography and documentation by trained technicians.
Sector adoption velocityclaude-sonnet-51/5Law enforcement and forensic fieldwork are low-digitization, physically embedded tasks with minimal AI adoption for on-scene evidence collection.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-assisted tools for object detection, scene reconstruction visualization, and automated measurement assistance can meaningfully improve a forensic technician's speed and thoroughness in documentation. However, the augmentation is moderate because human judgment and expertise remain central to deciding what warrants documentation and interpreting the significance of evidence.
Augmentation potentialclaude-sonnet-53/5AI-assisted tools like 3D scene reconstruction, photogrammetry, and image analysis software can help document and analyze evidence, improving accuracy and speed of the recording process.
Task automatabilityclaude-haiku-4-5-202510012/5While AI-powered image recognition can identify and classify objects like footprints or bloodstains in photographs, the task requires judgment about what to photograph, how to position evidence, and coordination with the overall investigation. Photography and measurement can be partially automated with robotics and structured light systems, but current systems cannot reliably perform the full reasoning, prioritization, and scene assessment end-to-end without substantial human oversight.
Task automatabilityclaude-sonnet-52/5Physical presence at a crime scene is required to locate, mark, and physically photograph evidence and take measurements; AI cannot perform the on-site physical actions, though it can assist with documentation and analysis afterward.'
Adoption barriersclaude-haiku-4-5-202510014/5Chain-of-custody and evidence integrity requirements mean that forensic documentation must be performed or signed off by a licensed investigator or forensic technician; automated systems cannot legally serve as the primary record without human verification and legal accountability. Regulatory and liability standards in criminal proceedings create strong barriers to full automation.
Adoption barriersclaude-sonnet-55/5Crime scene processing requires trained, often certified personnel whose evidence handling is subject to strict chain-of-custody and legal admissibility rules, making unauthorized automation legally untenable.
Cost vs. human wageclaude-haiku-4-5-202510012/5Specialized crime scene documentation equipment, AI software licenses, and integration into existing forensic workflows carry material upfront costs. The human investigator's time remains essential for scene interpretation and judgment, making the all-in cost of AI-assisted documentation comparable to or exceeding a detective's hourly rate when training and oversight are included.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing the physical evidence-gathering task, so cost comparison favors the human investigator entirely.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed products exist for automated crime scene photography and object detection (e.g., blood pattern analysis software), but they require human curation of what to photograph, manual setup of measurement reference objects, and expert interpretation of results. No production system reliably performs the full task independently; most systems are supportive tools rather than autonomous end-to-end solutions.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously conducts crime scene evidence collection, marking, or physical measurement; this remains a manual forensic task performed by trained personnel.

Maintain surveillance of establishments to obtain identifying information on suspects.

16

CI 7–25 · exposure 13 · augmentation 50 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption remains cautious due to legal complexity, civil liberties concerns, and high-profile accuracy failures. While some law enforcement agencies pilot facial recognition, widespread production deployment is slow. Budget constraints and regulatory pushback limit velocity in most jurisdictions.
Sector adoption velocityclaude-sonnet-52/5Law enforcement is a traditionally slow-adopting sector for autonomous AI in field operations, though camera analytics and facial recognition tools are seeing gradual pilot adoption.
Augmentation potentialclaude-haiku-4-5-202510013/5AI video analysis and facial recognition tools can assist investigators by flagging potential matches and reducing review time, but the human investigator must validate findings and maintain situational awareness. Productivity gain is meaningful but modest because core judgment and legal compliance remain human responsibilities.
Augmentation potentialclaude-sonnet-53/5AI-enhanced cameras, license plate readers, and facial recognition can assist investigators in identifying suspects faster, though the core surveillance task remains human-led.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with video analysis and facial recognition to identify suspects, the task requires human judgment in real-time surveillance decisions, suspect interaction, and interpretation of context—elements that cannot be fully automated today to meet the 50% time-saving bar. Current systems excel at flagging potential matches but struggle with contextual reasoning and legal/ethical compliance.
Task automatabilityclaude-sonnet-51/5Physical surveillance of establishments requires in-person presence, real-time judgment, and adaptability to unpredictable human behavior that current AI cannot autonomously execute end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Strong legal and regulatory barriers exist: surveillance must comply with Fourth Amendment protections, warrant requirements, state wiretapping laws, and agency-specific authorization protocols. Liability for false identifications and wrongful investigation is high. A licensed investigator or warrant typically must authorize and oversee surveillance activities.
Adoption barriersclaude-sonnet-54/5Surveillance for criminal investigation often requires legal authorization, chain-of-custody integrity, and sworn officer accountability, creating strong procedural and legal barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI surveillance infrastructure (cameras, servers, software licensing, integration) is substantial upfront; ongoing operational and compliance costs are high. When accounting for false positive investigation overhead and human review requirements, the cost per reliable identification approaches or exceeds that of skilled human surveillance operators.
Cost vs. human wageclaude-sonnet-51/5AI cannot replace the human presence and mobility required; any camera/analytics support still requires human operatives on-site, so AI does not reduce overall cost below human labor.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed facial recognition and surveillance systems exist but have documented accuracy limitations, bias issues, and narrow operational scope. Production deployments are limited; most real-world surveillance still relies heavily on human observers. Reliability remains materially compromised for mission-critical identification.
Technical feasibility todayclaude-sonnet-51/5No deployed product independently conducts covert physical surveillance operations; AI is at most used in fixed camera analytics, not full task execution.

Collaborate with other offices and agencies to exchange information and coordinate activities.

15

CI 5–25 · exposure 17 · augmentation 50 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement and criminal investigations are among the slowest-adopting sectors for AI automation, particularly for tasks involving inter-agency coordination where legal and institutional barriers remain high.
Sector adoption velocityclaude-sonnet-52/5Law enforcement is a traditionally slow-adopting sector for AI in core investigative coordination, constrained by legal, privacy, and interagency trust issues, though some digital case-sharing tools are emerging.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by organizing case files, flagging relevant information for sharing, scheduling coordination meetings, and tracking multi-agency communications—useful support that enhances human efficiency without replacing human judgment.
Augmentation potentialclaude-sonnet-53/5AI can help by summarizing case files, flagging relevant connections across databases, and drafting communications, meaningfully aiding investigators while humans retain coordination responsibility.
Task automatabilityclaude-haiku-4-5-202510011/5This task fundamentally requires human judgment, relationship-building, and strategic coordination across organizational boundaries. AI cannot independently decide what information to share, negotiate interagency priorities, or take responsibility for coordinated activities.
Task automatabilityclaude-sonnet-52/5Coordination involves relationship-building, judgment calls on information sensitivity, and negotiation across jurisdictions that current AI cannot autonomously perform, though it can assist with drafting and tracking.
Adoption barriersclaude-haiku-4-5-202510015/5Legal authority, confidentiality requirements, inter-agency protocols, and liability concerns all mandate that human law enforcement officials retain decision-making control over information sharing and coordination with external agencies.
Adoption barriersclaude-sonnet-54/5Interagency coordination often involves classified or sensitive law enforcement information, chain-of-custody concerns, and legal authority requirements that mandate human officers with proper clearance and accountability.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI systems would require significant oversight and human validation at every step; the cost of integrating such systems plus human supervision would exceed the cost of humans directly managing these relationships and communications.
Cost vs. human wageclaude-sonnet-52/5Human liaison officers remain necessary for trust-based interagency coordination; AI tools reduce some administrative overhead but don't replace the core cost driver of human relationship management.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can assist with information retrieval and scheduling, no deployed system can autonomously manage inter-agency collaboration or coordinate investigation activities across multiple jurisdictions requiring human authority and discretion.
Technical feasibility todayclaude-sonnet-52/5Some case management and information-sharing platforms exist, but no deployed AI product autonomously coordinates cross-agency investigative activities today.

Prepare and serve search and arrest warrants.

14

CI 0–29 · exposure 20 · augmentation 63 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement agencies are adopting AI for investigative support and document management slowly relative to other sectors, with limited production deployment of AI-driven warrant systems. Most jurisdictions still rely on manual drafting and human legal review.
Sector adoption velocityclaude-sonnet-51/5Law enforcement and judicial warrant processes are highly regulated, procedural, and slow to digitize let alone automate with AI agents.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can substantially augment detective work on warrant preparation by accelerating legal research, auto-populating template sections, cross-checking citations and statutes, and flagging compliance issues—allowing investigators to focus on factual investigation and judgment. This support meaningfully raises productivity on the preparation side.
Augmentation potentialclaude-sonnet-53/5AI can help draft affidavit language, search records, or organize case facts to support warrant applications, but the core legal and physical execution remains human-driven.
Task automatabilityclaude-haiku-4-5-202510013/5AI can assist with document preparation, legal citation checking, and procedural compliance for warrant applications, which constitute a significant portion of the work. However, the actual service/execution of warrants requires physical presence and human judgment in dynamic, sometimes adversarial situations that AI cannot fully automate.
Task automatabilityclaude-sonnet-51/5This task requires legal judgment, sworn affidavits, court appearances before a judge, and physical service/execution involving arrest authority—none of which current AI can perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5This task is legally protected: search and arrest warrants must be signed by a judge or magistrate, and service must typically be performed by sworn law enforcement or authorized personnel. Liability and constitutional requirements (Fourth Amendment compliance) create hard barriers to full automation or non-human execution.
Adoption barriersclaude-sonnet-55/5Warrant preparation and service require sworn law-enforcement authority, judicial approval, and legal accountability, representing hard statutory and licensing barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI document and research tools reduce labor on the preparation phase, but the cost of AI systems, legal review, and human sign-off combined with the human wages for the service execution component means the overall cost-benefit is only marginally favorable or comparable to current human processes.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the sworn officer or judicial process, so there is no viable cost comparison—human involvement is mandatory regardless of cost.
Technical feasibility todayclaude-haiku-4-5-202510012/5Legal research and document assembly for warrant preparation have AI tools in production (contract review, legal research platforms), but no deployed system reliably handles the full warrant preparation workflow including jurisdiction-specific requirements and legal arguments. Warrant service execution remains entirely outside current AI capability.
Technical feasibility todayclaude-sonnet-51/5No deployed product prepares legally binding warrant affidavits or serves warrants; this remains a human law-enforcement and judicial function.

Notify, or request notification of, medical examiner or district attorney representative.

13

CI 4–23 · exposure 13 · augmentation 38 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5This task sits at the core of investigative authority and legal procedure; adoption of autonomous systems is near-zero because the legal and accountability structure requires human discretion and accountability.
Sector adoption velocityclaude-sonnet-51/5Law enforcement and criminal justice sectors are notably slow to adopt AI for operational/legal notification tasks due to institutional inertia, legal risk, and lack of digitization in many police departments.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could assist by flagging cases that may warrant notification or drafting template messages, but the decision and act of notification must remain with the investigator to ensure proper legal standing and accountability.
Augmentation potentialclaude-sonnet-53/5AI can help by auto-drafting notification messages, tracking deadlines, or flagging when notification is required based on case status, providing moderate workflow assistance.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires understanding situational context, legal authority, and timing judgment—determining when notification is warranted and to whom. Current AI systems cannot reliably make these discretionary legal decisions or initiate official communications with proper legal standing.
Task automatabilityclaude-sonnet-52/5This is a simple notification/communication step, but it requires judgment about timing, jurisdiction, and case-specific protocol that ties into broader investigative context, limiting full automation despite the mechanical act of notifying being simple.'
Adoption barriersclaude-haiku-4-5-202510015/5This task is protected by legal and liability barriers: only authorized law enforcement personnel can officially notify medical examiners or prosecutors; improper notification could compromise investigations or evidence chain-of-custody requirements.
Adoption barriersclaude-sonnet-54/5Notification protocols in criminal investigations are often governed by legal and procedural requirements, chain-of-custody concerns, and accountability structures that require a human investigator to initiate and document the action.
Cost vs. human wageclaude-haiku-4-5-202510012/5The human cost of a detective making a notification is low (minutes of a salaried officer's time), and automating the decision logic and communication would require oversight that may exceed the time savings.
Cost vs. human wageclaude-sonnet-52/5The task itself is low-cost for a human to perform (a phone call or message), so AI automation offers little cost advantage once integration and oversight are factored in.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product performs this task independently; it remains a human-judgment and authorization responsibility. AI cannot legally act as the notifying party or hold responsibility for ensuring proper official channels are followed.
Technical feasibility todayclaude-sonnet-52/5Basic communication tools (email, messaging, case management systems) can facilitate notification, but no deployed AI product autonomously decides when and how to notify a medical examiner or DA in active investigations.

Preserve, process, and analyze items of evidence obtained from crime scenes and suspects, placing them in proper containers and destroying evidence no longer needed.

13

CI 0–25 · exposure 13 · augmentation 38 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement and forensic labs are moderate-to-laggard adopters of AI automation. While digital case management is common, physical evidence processing remains labor-intensive and heavily regulated, with limited incentive or regulatory allowance for AI-driven automation.
Sector adoption velocityclaude-sonnet-51/5Law enforcement evidence handling is a low-digitization, physically-grounded process with minimal AI adoption for the core physical preservation/destruction steps.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist technicians with evidence inventory management, documentation templates, and automated alerts for expiration or destruction deadlines, improving workflow efficiency. However, the physical handling and judgment-critical decisions limit transformative augmentation potential.
Augmentation potentialclaude-sonnet-52/5AI can assist with evidence logging, metadata tracking, or digital forensic analysis of certain evidence types, but offers little help with the core physical preservation and destruction actions.
Task automatabilityclaude-haiku-4-5-202510012/5Evidence handling involves physical manipulation of potentially contaminated items, chain-of-custody documentation, and context-dependent decisions about preservation methods. While AI could assist with documentation and sorting, the manual handling, proper containerization, and legal compliance aspects require human judgment and physical presence that current systems cannot provide at scale.
Task automatabilityclaude-sonnet-51/5This is a physical, chain-of-custody task requiring hands-on handling of physical evidence, proper containment, and legally mandated destruction procedures—none of which current AI systems can perform.
Adoption barriersclaude-haiku-4-5-202510014/5Strict chain-of-custody legal requirements, court admissibility standards, and regulatory oversight of evidence handling create strong barriers. Most jurisdictions require a qualified human to certify proper handling, and destruction often requires authorized sign-off, preventing full automation.
Adoption barriersclaude-sonnet-55/5Chain-of-custody law, evidentiary admissibility rules, and criminal procedure require certified personnel to handle, document, and dispose of evidence, creating hard legal barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5The labor cost for trained evidence technicians is relatively modest per item processed, and AI solutions for partial automation (documentation, cataloging) still require human oversight and physical work, keeping total cost savings limited compared to full technician wages.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute for physical evidence handling, so cost comparison favors the human by default since AI cannot perform the task at all.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some AI-based inventory and documentation systems exist in labs, but end-to-end evidence processing—physical handling, contamination prevention, proper storage, and destruction authorization—remains primarily manual. No deployed product reliably performs the full task independently.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product handles, packages, or physically disposes of forensic evidence; this remains entirely a human physical and procedural task.

Determine scope, timing, and direction of investigations.

13

CI 0–25 · exposure 13 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement has invested in data systems and evidence management tools, but adoption of AI for autonomous investigation-direction decisions remains very limited. Most deployments are for narrower tasks (facial recognition, crime forecasting) rather than scope/timing decisions, and cultural and legal conservatism in policing slows adoption.
Sector adoption velocityclaude-sonnet-51/5Law enforcement is a slow-adopting, highly regulated sector with minimal AI deployment in core investigative decision-making due to legal and public accountability constraints.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully assist by organizing evidence, flagging patterns, recommending resource allocation scenarios, and tracking case timelines, improving investigator productivity. However, the core judgment remains with the human investigator, making this a moderate augmentation scenario rather than transformative.
Augmentation potentialclaude-sonnet-53/5AI can assist by organizing case data, flagging patterns, or summarizing evidence, helping investigators make more informed scope and timing decisions, though the judgment itself remains human.
Task automatabilityclaude-haiku-4-5-202510012/5Determining investigation scope and timing requires weighing evidence, risk assessment, resource allocation, and legal/procedural judgment. While AI can assist with data analysis and pattern detection, it cannot reliably make the complex discretionary decisions about prioritization and legal direction that define this task, and cannot meet the ≥50% time-saving threshold without human oversight remaining dominant.
Task automatabilityclaude-sonnet-51/5Determining investigation scope, timing, and direction requires contextual judgment, legal strategy, and adaptive decision-making based on evolving evidence that current AI cannot perform end-to-end reliably.
Adoption barriersclaude-haiku-4-5-202510014/5Investigation direction carries high liability; decisions affect suspects' rights, case admissibility, and resource allocation. Regulatory frameworks (Fourth Amendment, investigative standards, chain of custody) and organizational accountability structures require human sign-off and discretion, creating strong legal and institutional barriers.
Adoption barriersclaude-sonnet-55/5This is a core law enforcement function requiring sworn authority, legal accountability, and chain-of-custody responsibility that only credentialed detectives can exercise.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI assistance for data compilation and analysis is relatively cheap, but does not replace the investigator's judgment role. The cost of oversight, error correction, and liability remains substantial, keeping total cost comparable to or above the loaded cost of a human investigator.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this decision task, so cost comparison favors the human investigator entirely.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs this task end-to-end. Tools exist for evidence organization and recommendation, but actual investigation direction decisions remain human-led in all known production settings, and the liability and legal stakes prevent autonomous systems from making these calls.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously directs criminal investigations; AI tools at best support research or data analysis, not strategic case direction.

Obtain facts or statements from complainants, witnesses, and accused persons and record interviews, using recording device.

11

CI 5–16 · exposure 5 · augmentation 50 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement has been slow to adopt AI for investigative interviewing; most adoption is limited to transcription and data management tools, not autonomous statement-gathering, reflecting both regulatory caution and skepticism about AI credibility in criminal proceedings.
Sector adoption velocityclaude-sonnet-51/5Law enforcement is a low-digitization, physically grounded, highly regulated sector with slow AI adoption for core investigative interviewing functions.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by transcribing interviews, flagging inconsistencies, and organizing statements, but the human investigator remains essential for conducting the actual interview, assessing credibility, and making on-the-spot judgment calls.
Augmentation potentialclaude-sonnet-53/5AI transcription, translation, and note-summarization tools can meaningfully assist investigators in recording and organizing statements, even though the interview itself remains human-led.
Task automatabilityclaude-haiku-4-5-202510011/5Recording audio or video is trivial for machines, but 'obtaining facts or statements' requires interpersonal judgment, rapport-building, and real-time responsiveness to emotional cues and inconsistencies that current AI cannot reliably perform in investigative contexts where accuracy and admissibility are critical.
Task automatabilityclaude-sonnet-51/5Obtaining facts and statements requires in-person interaction, reading demeanor, adapting questioning strategy, and building rapport/trust with witnesses or suspects—none of which current AI can perform end-to-end. Recording audio is trivial but is a minor sub-component of the task.
Adoption barriersclaude-haiku-4-5-202510014/5Legal and evidentiary requirements are substantial: statements must be properly obtained by authorized personnel, chain of custody and admissibility rules apply, and liability for mishandled interviews falls on the department; these create high friction against full automation.
Adoption barriersclaude-sonnet-54/5Legal, evidentiary, and due-process requirements (Miranda warnings, chain-of-custody, admissibility rules) generally require a sworn officer or authorized investigator to conduct and document these interviews, creating strong institutional and legal barriers.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI transcription is cheap, but the core task—extracting truthful, legally sound statements—still requires a trained human investigator; the marginal cost savings from automation are minimal relative to the investigator's loaded wage.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute for conducting the interview itself, so cost comparison is moot for the core task; human investigators remain necessary.
Technical feasibility todayclaude-haiku-4-5-202510012/5While transcription and basic recording devices exist and work, no deployed AI system reliably obtains statements from witnesses and suspects in ways that meet investigative and legal standards; human investigators must conduct the interviews themselves, though AI can assist with transcription post-hoc.
Technical feasibility todayclaude-sonnet-51/5No deployed product conducts investigative interviews or interrogations autonomously; AI transcription tools exist but do not perform the interviewing itself.

Search for and collect evidence, such as fingerprints, using investigative equipment.

10

CI 0–20 · exposure 13 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement adoption of investigative AI remains limited; agencies primarily use forensic software for post-collection analysis rather than autonomous field collection. Human investigators remain central to evidence work across most jurisdictions.
Sector adoption velocityclaude-sonnet-51/5Law enforcement and forensic fieldwork are low-digitization, physically grounded sectors with minimal AI agent deployment for on-scene evidence gathering.
Augmentation potentialclaude-haiku-4-5-202510014/5AI-powered forensic image analysis, fingerprint enhancement, and database matching substantially assist human investigators in processing and interpreting collected evidence, accelerating identification workflows and reducing analyst workload significantly.
Augmentation potentialclaude-sonnet-52/5AI-enabled tools (e.g., digital fingerprint matching software, image analysis) can assist in processing or analyzing collected evidence, but offer little help with the physical search and collection process itself.
Task automatabilityclaude-haiku-4-5-202510012/5While AI-assisted image analysis can help classify fingerprints post-collection, the physical act of searching for and collecting evidence requires human judgment, dexterity, and presence at crime scenes. Current systems cannot autonomously locate, photograph, and preserve forensic evidence in uncontrolled environments with <50% time savings at equal quality.
Task automatabilityclaude-sonnet-51/5Physical evidence collection requires manual dexterity, scene navigation, and chain-of-custody handling that current AI systems cannot perform; this is a physical-world manipulation task, not a cognitive/data task.dans
Adoption barriersclaude-haiku-4-5-202510015/5Chain of custody and evidentiary standards legally require documented human handling and authority over evidence collection; courts mandate a credible human investigator to testify and certify procedures, creating a hard regulatory and legal barrier to automation.
Adoption barriersclaude-sonnet-55/5Evidence collection has strict legal chain-of-custody, admissibility, and licensing/certification requirements for who may collect and handle forensic evidence, creating hard legal barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Investigative equipment, crime-scene imaging, and forensic analysis tools are capital-intensive; human investigators remain cheaper than the combined infrastructure, maintenance, and human oversight required for any automated collection system.
Cost vs. human wageclaude-sonnet-51/5There is no AI system capable of substituting for the human physical labor involved, so cost comparison is moot—AI cannot perform the task at all, making it effectively infinitely more expensive or simply infeasible.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed forensic software (AFIS, forensic imaging tools) assists with fingerprint classification and matching, but no autonomous system reliably searches for, locates, or physically collects evidence. Products exist for analysis post-collection, but reliable end-to-end autonomous collection is research-stage only.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously searches crime scenes and physically collects fingerprint or forensic evidence; this remains entirely a human forensic task, sometimes aided by imaging tools but not performed by AI.

Question individuals or observe persons and establishments to confirm information given to patrol officers.

3

CI 0–5 · exposure 5 · augmentation 25 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement remains a laggard sector in AI adoption for core investigative functions due to legal, liability, and union considerations; AI deployment in detective work is minimal and largely limited to back-office analysis rather than field-level questioning.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field investigation is a low-digitization, physically embedded sector with minimal AI agent deployment for actual questioning or surveillance work.
Augmentation potentialclaude-haiku-4-5-202510012/5AI can provide limited assistance by preparing case summaries or flagging inconsistencies in prior statements for a detective to follow up on, but it does not meaningfully augment the core task of live questioning and observational confirmation.
Augmentation potentialclaude-sonnet-52/5AI can help with case note transcription, background checks, or flagging inconsistencies in prior statements, but offers little assistance to the live questioning/observation process itself.
Task automatabilityclaude-haiku-4-5-202510011/5Questioning individuals and observing establishments to confirm patrol officer information requires real-time interpersonal judgment, discretion, and the ability to read contextual cues and detect deception—capabilities that current AI systems cannot reliably perform in uncontrolled field settings. No AI system can replace the investigative acumen and human presence required for effective questioning.
Task automatabilityclaude-sonnet-51/5This requires physically visiting establishments, in-person surveillance, and interviewing people with adaptive rapport-building and judgment—capabilities far beyond current AI systems' end-to-end scope.
Adoption barriersclaude-haiku-4-5-202510015/5Criminal investigation work is heavily regulated and legally constrained; interrogations and investigative questioning are governed by Miranda rights, constitutional protections, and evidentiary rules that effectively require a licensed human investigator to conduct and sign off on confirmatory questioning.
Adoption barriersclaude-sonnet-55/5Law enforcement authority, chain-of-custody/evidentiary rules, constitutional protections (Miranda, search/seizure), and sworn-officer requirements make this a hard legal barrier to automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of implementing, monitoring, and managing AI for investigative questioning far exceeds the labor cost of a detective performing this task, especially when accounting for liability, integration with case management systems, and the oversight required.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing physical observation or interrogation, so the comparison is moot—human labor is the only viable option today.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can assist with analyzing surveillance footage or generating initial report summaries, no deployed product reliably performs live questioning or independent observational confirmation at the quality standard required for criminal investigations. Some tools exist for data analysis, but the core task remains firmly in human domain.
Technical feasibility todayclaude-sonnet-51/5No deployed product conducts in-person questioning or physical surveillance of suspects/witnesses; this remains firmly a human field task.

Perform undercover assignments and maintain surveillance, including monitoring authorized wiretaps.

1

CI 0–3 · exposure 0 · augmentation 38 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement agencies are extremely conservative adopters of automation in undercover and surveillance roles due to liability, legal standing, and constitutional protections; there is minimal production deployment of AI in these functions.
Sector adoption velocityclaude-sonnet-52/5Law enforcement is a slow-adopting, highly regulated sector; AI is used for transcription/analysis support but not for replacing the human field or interception role.
Augmentation potentialclaude-haiku-4-5-202510012/5AI might assist with post-surveillance analysis, data synthesis, or pattern detection from recorded material, but cannot meaningfully augment the core undercover or real-time surveillance activities, which remain almost entirely human-dependent.
Augmentation potentialclaude-sonnet-53/5AI can assist with transcribing, translating, and flagging relevant content in wiretap audio and analyzing surveillance data patterns, aiding investigators without replacing the human role.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time physical presence, situational judgment, adaptive decision-making under uncertainty, and the ability to respond to unpredictable human interactions—capabilities far beyond current AI. Wiretap monitoring alone is partial and lacks the investigative intuition, legal authority interpretation, and human presence that undercover work demands.
Task automatabilityclaude-sonnet-51/5Undercover work requires physical human presence, deception of suspects via human social interaction, and real-time judgment in dangerous situations that AI cannot perform.
Adoption barriersclaude-haiku-4-5-202510015/5Undercover assignments and authorized wiretap monitoring are tightly regulated by law enforcement hierarchy, require specific legal authorization, carry severe liability and constitutional consequences for error, and mandate human accountability—creating near-total legal and organizational barriers to automation.
Adoption barriersclaude-sonnet-55/5Undercover operations and wiretap authorization are legally mandated to be conducted by sworn, authorized law enforcement personnel under strict judicial oversight and chain-of-custody rules.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of developing, deploying, and maintaining an AI system capable of undercover work (if theoretically possible) would vastly exceed the loaded wage of a human investigator, and current systems offer no cost advantage for this task.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute for the human physical/social component, so cost comparison is moot for the core task, though wiretap monitoring could be partially assisted at lower cost.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product reliably performs undercover operations or manages authorized surveillance in any meaningful sense; these are inherently human-centric law enforcement activities requiring legal standing, physical embodiment, and real-time contextual judgment that current systems cannot replicate.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs undercover impersonation or physical surveillance; this remains entirely a human field operation.

Check victims for signs of life, such as breathing and pulse.

0

CI 0–0 · exposure 0 · augmentation 0 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement remains highly traditional and risk-averse in automation of critical forensic tasks; no measurable AI adoption for victim assessment exists in production policing.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field work involving physical victim assessment shows essentially no AI adoption; this is a low-digitization, physical-presence task.
Augmentation potentialclaude-haiku-4-5-202510011/5AI cannot meaningfully augment the physical examination of victims for signs of life; the task requires direct tactile feedback and immediate clinical judgment that current AI cannot support in real-time field conditions.
Augmentation potentialclaude-sonnet-51/5AI offers no meaningful assistance for the immediate physical act of checking pulse or breathing at a scene.
Task automatabilityclaude-haiku-4-5-202510011/5Checking for signs of life requires physical contact with victims, assessment of subtle physiological signals, and real-time decision-making in complex, unpredictable crime scenes. Current AI cannot perform the physical examination or reliably detect pulse and breathing without specialized hardware not deployable in field conditions.
Task automatabilityclaude-sonnet-51/5This requires direct physical contact and sensory examination of a person at a scene, which current AI systems cannot perform; it is an inherently physical, embodied task.'
Adoption barriersclaude-haiku-4-5-202510015/5This task is legally and procedurally protected: only licensed law enforcement or medical personnel can certify life status at crime scenes, and chain-of-custody and evidentiary standards require human professional judgment and accountability.
Adoption barriersclaude-sonnet-55/5Checking for signs of life at a crime scene involves legal, medical, and evidentiary responsibilities requiring trained, authorized personnel present in person, an absolute hard barrier to automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5Deploying autonomous systems for this would require mobile robotics with fine manipulation, sensing, and decision-making capabilities—orders of magnitude more expensive than a detective performing the task directly.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute providing this physical service, so cost comparison favors the human by default.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product performs this task autonomously in field conditions. Wearable sensors and automated monitoring exist in clinical settings, but they cannot replace the tactile and observational assessment performed by a detective at a crime scene.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs physical vital-sign checks on crime scene victims; this remains entirely a human/first-responder function.

Secure deceased body and obtain evidence from it, preventing bystanders from tampering with it prior to medical examiner's arrival.

0

CI 0–0 · exposure 0 · augmentation 13 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5This is a core law enforcement function tied to human authority, training, and legal accountability; there is no meaningful AI adoption in this specific task in any sector.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field operations, especially physical scene control, are a low-digitization, low-AI-adoption domain with no evidence of automation trends here.
Augmentation potentialclaude-haiku-4-5-202510011/5Current AI systems offer no meaningful assistance in physically securing a scene, preventing tampering, or managing evidence collection in real time.
Augmentation potentialclaude-sonnet-52/5AI could assist with documentation, scene mapping, or evidence logging afterward, but offers minimal help with the core real-time physical securing and bystander control task.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires physical presence at a crime scene, securing a perimeter, and making real-time decisions about evidence handling and bystander management—capabilities that current AI systems lack entirely. No part of the task can be meaningfully automated by software or deployed robotic systems today.
Task automatabilityclaude-sonnet-51/5This is a physical presence and crime-scene security task requiring a human officer to physically guard a body and control bystanders; no AI system can perform this physical, authority-based function.
Adoption barriersclaude-haiku-4-5-202510015/5Only licensed law enforcement personnel can legally secure a crime scene and handle evidence under chain-of-custody requirements; this is a hard legal and regulatory barrier that prevents any non-human substitution.
Adoption barriersclaude-sonnet-55/5Legal chain-of-custody requirements, sworn officer authority, and evidentiary integrity rules mandate a certified law enforcement official perform and document this task.
Cost vs. human wageclaude-haiku-4-5-202510011/5A trained detective or investigator on-site is necessary and far cheaper than any hypothetical robotic or AI system capable of this task, which does not exist at any price.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical security and evidence-preservation function, so cost comparison favors the human by default since AI cannot do the task at all.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product performs this task; it requires embodied agents with legal authority, situational judgment, and physical intervention capabilities that do not exist in production systems.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product secures physical locations, guards evidence, or controls crowds; this remains entirely a human/robotic-physical task outside current AI product scope.

Obtain summary of incident from officer in charge at crime scene, taking care to avoid disturbing evidence.

0

CI 0–0 · exposure 0 · augmentation 25 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement operates in a heavily regulated, hierarchical, and evidence-dependent domain with strict procedural requirements. Adoption of AI automation for crime scene investigative tasks remains virtually nonexistent in production deployments.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field operations are a low-digitization, physically-grounded sector with minimal AI agent deployment for scene-level evidence handling.
Augmentation potentialclaude-haiku-4-5-202510012/5AI might assist with post-incident documentation or transcription of officer summaries already recorded, but it offers minimal real-time assistance during the critical scene interaction and evidence-preservation phase where judgment and physical presence are paramount.
Augmentation potentialclaude-sonnet-52/5AI could help transcribe or summarize verbal reports after the fact, but it offers little assistance during the actual in-person, evidence-sensitive information gathering.
Task automatabilityclaude-haiku-4-5-202510011/5This task fundamentally requires real-time presence at a crime scene, navigation of sensitive physical environments, and dynamic human interaction with law enforcement personnel. AI has no capability to physically attend scenes, assess evidence preservation, or conduct nuanced in-person interviews today.
Task automatabilityclaude-sonnet-51/5This requires physically being present at a crime scene, engaging in real-time verbal interaction with officers, and making judgment calls about evidence handling—none of which current AI can perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Criminal investigation and evidence handling are heavily regulated by law, chain-of-custody requirements, and liability frameworks. Only authorized law enforcement personnel can legally conduct crime scene investigations and obtain evidentiary summaries; this is a hard legal and procedural barrier.
Adoption barriersclaude-sonnet-55/5Chain-of-custody, evidentiary integrity, and legal authority requirements mean only sworn, authorized personnel can perform this task, creating hard legal and procedural barriers.
Cost vs. human wageclaude-haiku-4-5-202510011/5Since the task cannot be meaningfully automated by current AI, there is no cost comparison favorable to AI; the human detective remains essential and irreplaceable for this function.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical, in-person task, so cost comparison favors the human by default since AI cannot deliver the output at all.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI system can conduct physical crime scene attendance or obtain summaries from officers through in-person interaction while managing evidence protocols. This requires embodied presence and real-world coordination that does not exist in production systems.
Technical feasibility todayclaude-sonnet-51/5No deployed product conducts in-person crime scene briefings or physically navigates evidence-sensitive environments; this remains entirely a human field task.

Secure persons at scene, keeping witnesses from conversing or leaving the scene before investigators arrive.

0

CI 0–0 · exposure 0 · augmentation 13 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5There is no adoption of AI for this task because it requires legal authority and physical presence that cannot be replicated by software or current robotics deployed in law enforcement contexts.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field operations are a low-digitization, physical-presence-dependent sector with minimal AI adoption for on-scene control tasks.
Augmentation potentialclaude-haiku-4-5-202510011/5AI cannot meaningfully augment an officer's ability to physically secure a scene or enforce compliance with movement restrictions, as the core task is interpersonal and physical in nature.
Augmentation potentialclaude-sonnet-52/5AI could assist with logistics like recording witness statements or scene documentation afterward, but offers negligible help during the active physical securing of persons and scene.
Task automatabilityclaude-haiku-4-5-202510011/5Securing persons and preventing them from leaving or conversing requires physical presence, authority enforcement, and real-time judgment in unpredictable human interactions. Current AI systems cannot be deployed to physically restrain people or enforce compliance in the field.
Task automatabilityclaude-sonnet-51/5This requires physical presence, authority, and real-time interpersonal control at a scene—current AI systems have no embodied capability to physically secure people or scenes.
Adoption barriersclaude-haiku-4-5-202510015/5This task has hard legal and regulatory barriers: only authorized law enforcement officers can lawfully secure a crime scene and restrict persons' movement and speech. A human officer must legally perform this function.
Adoption barriersclaude-sonnet-55/5This is a legally authorized law enforcement function requiring sworn officer authority, physical presence, and legal accountability, making automation essentially barred.
Cost vs. human wageclaude-haiku-4-5-202510011/5There is no AI alternative to human security personnel at a crime scene; the cost comparison is irrelevant since automation is not feasible at any price point.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical task, so AI cost is not comparable—human officers are the only viable option.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI system can perform physical security, crowd management, or authority-based person control in real crime scenes. This task fundamentally requires human law enforcement presence.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs physical scene security or crowd/witness control; this remains entirely a human, on-site law enforcement function.

Participate or assist in raids and arrests.

0

CI 0–0 · exposure 0 · augmentation 38 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5No measurable adoption of AI for conducting raids or arrests exists; the task remains exclusively human-performed due to legal mandate and physical necessity.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field operations are a low-digitization, physical-world sector with minimal AI deployment for direct tactical actions like raids and arrests.
Augmentation potentialclaude-haiku-4-5-202510012/5AI can assist with pre-raid planning (intelligence analysis, suspect location forecasting) but offers minimal real-time support during the actual raid or arrest execution itself.
Augmentation potentialclaude-sonnet-53/5AI can assist with pre-raid intelligence, surveillance analysis, mapping, and risk assessment, but offers no direct help during the physical act of the raid or arrest itself.
Task automatabilityclaude-haiku-4-5-202510011/5Raids and arrests involve physical presence, real-time judgment calls, force deployment, and direct human interaction in high-stakes legal settings. Current AI cannot physically execute these actions or reliably make split-second safety/legal decisions in unpredictable environments.
Task automatabilityclaude-sonnet-51/5Physical raids and arrests require in-person presence, physical control, use of force, and split-second judgment that no current AI system can perform end-to-end.'
Adoption barriersclaude-haiku-4-5-202510015/5Raids and arrests are among the most heavily regulated law enforcement functions, requiring licensed police authority, legal warrants, liability oversight, and direct human accountability under criminal and civil law.
Adoption barriersclaude-sonnet-55/5Arrests and use of force are strictly regulated, require sworn, legally authorized officers, and carry major liability and constitutional constraints preventing any non-human execution.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI has no economic alternative to a human detective performing physical enforcement; there is no meaningful cost comparison since substitution is not feasible.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical enforcement task, so no meaningful cost comparison exists—AI cannot replace the human labor involved.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product performs raid or arrest operations autonomously today. This task requires licensed law enforcement with legal authority and physical capability that AI systems fundamentally lack.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product physically participates in raids or arrests; this remains entirely a human physical/tactical task.

Organize scene search, assigning specific tasks and areas of search to individual officers and obtaining adequate lighting as necessary.

0

CI 0–0 · exposure 0 · augmentation 25 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement agencies operate within rigid command structures and legal frameworks that require human officers to direct scene management. Digital tools may assist documentation, but the core coordination function remains human-dependent.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field operations are a low-digitization, physically-grounded sector with minimal AI agent deployment for on-scene tactical command.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could potentially assist with scene mapping visualization or search optimization suggestions, but current tools offer minimal meaningful support to the detective's core function of assigning tasks and coordinating personnel in real-time.
Augmentation potentialclaude-sonnet-52/5AI could assist with checklists, scene mapping software, or evidence logging suggestions, but it offers little real-time support for dynamic personnel assignment and lighting logistics decisions.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time coordination of personnel at a physical crime scene, assigning tasks to individual officers, and making context-dependent decisions about lighting and search strategy. Current AI lacks the situational awareness, human coordination capability, and on-site authority necessary to perform this end-to-end.
Task automatabilityclaude-sonnet-51/5This requires physical presence, real-time judgment about scene hazards/evidence, and direct coordination of personnel on-site; no AI system can perform this end-to-end today.
Adoption barriersclaude-haiku-4-5-202510015/5This task is legally and operationally required to be performed by an authorized law enforcement officer. Chain of command, evidence integrity requirements, and liability create hard barriers to automation or substitution.
Adoption barriersclaude-sonnet-55/5Chain-of-custody, evidentiary integrity, and legal admissibility require a sworn, trained officer to physically direct and document scene search activities, creating hard legal and procedural barriers.
Cost vs. human wageclaude-haiku-4-5-202510011/5The task requires an experienced human detective on-site; any AI assistance would require human oversight and decision-making, making the combined cost higher than the human alone performing the task.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical coordination task, so cost comparison favors the human by default since AI cannot deliver the output at all.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI system can reliably organize and direct human officers at crime scenes or make binding task assignments that replace a detective's on-site command role. This requires human authority and real-time physical presence.
Technical feasibility todayclaude-sonnet-51/5No deployed product organizes and directs physical crime scene searches or personnel assignments; this remains entirely a research-fiction scenario for current AI.

Summon medical help for injured individuals and alert medical personnel to take statements from them.

0

CI 0–0 · exposure 0 · augmentation 25 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement operates in highly regulated, human-intensive environments where physical presence, real-time judgment, and legal accountability remain mandatory; adoption of AI for emergency summoning at crime scenes is negligible.
Sector adoption velocityclaude-sonnet-51/5Law enforcement field operations are a low-digitization, physically embedded sector with minimal AI agent deployment for real-time scene management.
Augmentation potentialclaude-haiku-4-5-202510012/5Current AI might assist marginally by logging or documenting statements post-call, but cannot meaningfully augment the core task of assessing injury and directing emergency response, which demands immediate human presence and judgment.
Augmentation potentialclaude-sonnet-52/5AI could help with dispatch logistics or documentation afterward, but offers little assistance during the actual real-time act of summoning help and coordinating statements.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires human judgment to assess whether medical help is needed, direct communication with emergency services, and coordination with medical personnel—none of which can be reliably automated end-to-end by current AI systems in real-world crime scenes.
Task automatabilityclaude-sonnet-51/5This requires physical presence at a scene, real-time judgment about medical urgency, and direct human coordination with emergency responders—none of which current AI can perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Strong legal and operational barriers exist: only humans can legally make triage decisions and call 911; liability for injury assessment and emergency dispatch falls on responsible humans; crime-scene protocols mandate human authority and documentation.
Adoption barriersclaude-sonnet-55/5Emergency response, medical triage decisions, and evidentiary statement-taking involve legal authority, chain-of-custody, and duty-of-care obligations that require a sworn officer or authorized personnel.
Cost vs. human wageclaude-haiku-4-5-202510011/5A human detective on-scene can assess and act immediately at minimal marginal cost; any AI system would require extensive integration, real-time sensing, and human oversight, making it more expensive than the baseline human task.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this physical, time-critical task, so cost comparison favors the human by default.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs the full task of assessing injury severity, summoning appropriate emergency response, and coordinating with medical teams in a crime-scene context; this requires human presence and accountability.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product summons emergency help or coordinates real-time medical response and interviews at an investigation scene; this remains entirely a human field function.

Block or rope off scene and check perimeter to ensure that entire scene is secured.

0

CI 0–0 · exposure 0 · augmentation 13 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Crime scene security is a core law enforcement function with deep institutional and legal requirements. Adoption of automation in this domain has been minimal, and public data shows no meaningful production deployment of AI or robotic systems for independent perimeter control.
Sector adoption velocityclaude-sonnet-51/5Physical scene security in policing is a low-digitization, hands-on task with no meaningful AI/robotic adoption trend.
Augmentation potentialclaude-haiku-4-5-202510012/5While drones or monitoring cameras might provide supplementary situational awareness to investigators, they offer limited assistance to the core task of physically blocking/roping off and patrolling a perimeter, which remains fundamentally manual.
Augmentation potentialclaude-sonnet-51/5AI provides no meaningful assistance to the physical act of roping off and patrolling a perimeter.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires physical presence at a scene to establish perimeters, handle barriers, and make real-time security assessments. Current AI systems cannot physically manipulate rope, barriers, or deploy personnel to secure locations.
Task automatabilityclaude-sonnet-51/5This is a physical task requiring a human to walk terrain, place barriers, and physically assess boundaries—no AI system can perform this end-to-end today.
Adoption barriersclaude-haiku-4-5-202510015/5This task has hard legal and operational barriers: law enforcement authority, liability, and chain-of-custody requirements mandate a licensed human officer physically present to control and verify scene security. Automation would require regulatory change and institutional authority.
Adoption barriersclaude-sonnet-55/5Scene security is a legally mandated law-enforcement function tied to chain-of-custody and evidentiary integrity, requiring sworn personnel physically present.
Cost vs. human wageclaude-haiku-4-5-202510011/5Deploying robots or autonomous systems to secure crime scenes would be significantly more expensive than assigning trained personnel, particularly given the need for reliability and legal accountability at active scenes.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing this physical action, so AI cost is not applicable/comparable—human labor is the only option.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product can autonomously block off crime scenes or physically secure perimeters. While drones or robots might assist, they cannot independently execute the full task of securing an investigative scene with the authority and judgment required.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product physically secures crime scenes; this remains entirely manual work performed by officers or investigators.

Testify before grand juries concerning criminal activity investigations.

0

CI 0–0 · exposure 0 · augmentation 25 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5This task will not see AI adoption because it is legally impossible. No sector trend toward AI grand jury testimony can occur, regardless of digitization or willingness to automate.
Sector adoption velocityclaude-sonnet-51/5Legal proceedings are highly resistant to automation of testimony itself; adoption of AI in this specific act is essentially nonexistent and unlikely to change.
Augmentation potentialclaude-haiku-4-5-202510011/5AI cannot assist a detective in the act of testifying before a grand jury, since the detective must personally answer questions under oath. Background case summaries or evidence preparation might help beforehand, but do not augment testimony itself.
Augmentation potentialclaude-sonnet-53/5AI can help investigators prepare testimony by organizing case files, summarizing evidence, and drafting talking points, though the testimony delivery itself is unaided.
Task automatabilityclaude-haiku-4-5-202510011/5Testifying before a grand jury fundamentally requires a human witness to be present and sworn, answering questions in real time about an investigation. This is inherently a human legal function with no meaningful automation opportunity.
Task automatabilityclaude-sonnet-51/5Testifying is a legally binding act requiring a live, sworn human witness with personal knowledge; AI cannot perform courtroom/grand jury testimony end-to-end today.
Adoption barriersclaude-haiku-4-5-202510015/5Grand jury testimony is protected by strict legal and constitutional requirements: the witness must be physically present, sworn, and subject to questioning. Federal and state rules of evidence and procedure mandate human testimony; no delegation to AI is permissible.
Adoption barriersclaude-sonnet-55/5Grand jury testimony has hard legal and evidentiary requirements demanding a sworn human witness with firsthand knowledge, making AI substitution legally impossible.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI cannot perform this task at all; cost comparison is moot. A detective must appear in person, making the human cost unavoidable.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute to price against human cost since the task legally requires an actual investigator to testify under oath.
Technical feasibility todayclaude-haiku-4-5-202510011/5No AI system can legally substitute for a detective's in-person testimony before a grand jury, which requires a natural person under oath to answer cross-examination. This is not a technical capability gap but a legal and constitutional requirement.
Technical feasibility todayclaude-sonnet-51/5No deployed product substitutes for sworn human testimony before a grand jury; this remains entirely research-inapplicable and legally impossible.

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.