Intelligence Analysts

33-3021.06
Median wage $93,790/yr114,430 employed (US)Rank #517 of 923 scored · top 56% by substitution

Gather, analyze, or evaluate information from a variety of sources, such as law enforcement databases, surveillance, intelligence networks or geographic information systems. Use intelligence data to anticipate and prevent organized crime activities, such as terrorism.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution26
Exposure28
Augmentation69

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

21 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%29

panel mean rating 2.2/5 → substitution pressure 29/100

Technical feasibility todayw 20%28

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

Cost vs. human wagew 15%29

panel mean rating 2.1/5 → substitution pressure 29/100

Adoption barriersw 20%inverted — strong barriers lower the score17

panel mean rating 4.3/5 (barrier strength) → substitution pressure 17/100

Sector adoption velocityw 10%30

panel mean rating 2.2/5 → substitution pressure 30/100

Task breakdown (21 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.

Design, use, or maintain databases and software applications, such as geographic information systems (GIS) mapping and artificial intelligence tools.

57

CI 28–87 · exposure 58 · augmentation 88 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Government, defense, and intelligence agencies have begun integrating AI-assisted development and database tools into production workflows; commercial analytics and GIS sectors show rapid adoption of AI coding assistants and automated schema generation. Adoption is moving from pilots to operational deployment, though security and compliance requirements may slow government sector velocity.
Sector adoption velocityclaude-sonnet-53/5Government and defense-adjacent intelligence sectors are moderately adopting AI-assisted analytics and GIS tools, but adoption is slower and more cautious than in fast-moving private-sector information industries.
Augmentation potentialclaude-haiku-4-5-202510015/5AI tools dramatically enhance analyst productivity in database design, code review, GIS scripting, and application maintenance by generating templates, catching errors, and automating boilerplate work while analysts focus on security requirements, domain logic, and system architecture decisions.
Augmentation potentialclaude-sonnet-54/5AI coding assistants, GIS automation scripts, and ML-based pattern detection meaningfully speed up database design, maintenance, and analytic tool development while analysts retain oversight and design authority.
Task automatabilityclaude-haiku-4-5-202510015/5AI systems can now design database schemas, generate code for software applications, and maintain GIS layers with minimal human oversight. Current LLMs and specialized tools can handle schema design, SQL generation, GIS scripting, and software testing at scale, easily meeting the 50% time-saving threshold for routine maintenance and application design tasks.
Task automatabilityclaude-sonnet-52/5AI coding assistants can help write scripts or queries for GIS and databases, but designing, integrating, and maintaining full analytic systems requires substantial human architecture decisions and domain judgment that current tools cannot fully replace..rated conservatively.
Adoption barriersclaude-haiku-4-5-202510012/5Intelligence contexts may involve security/compartmentalization requirements and organizational preference for human accountability in system design, but there are no legal licensing bars to automation itself. Most barriers are organizational friction and risk mitigation rather than hard legal constraints on substitution.
Adoption barriersclaude-sonnet-54/5Intelligence work involves classified systems, security clearances, and strict data-handling regulations, creating strong organizational and legal barriers against full automation or unsupervised AI tool use.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI inference for code generation, database schema design, and GIS automation costs pennies per task instance, while intelligence analysts command $80k–$150k+ annual salaries. The all-in cost per task-equivalent is orders of magnitude cheaper than human labor for routine design and maintenance work.
Cost vs. human wageclaude-sonnet-52/5AI tools reduce some coding and query-writing time cheaply, but the overall task still requires expensive skilled analysts/engineers for architecture, security, and maintenance oversight, keeping costs roughly comparable to human-only workflows.
Technical feasibility todayclaude-haiku-4-5-202510014/5Mature products (GitHub Copilot, Claude, ChatGPT for code generation; ESRI's AI-assisted GIS tools; cloud database automation platforms) demonstrate reliable performance on database design and maintenance in production environments. Performance is high for standard tasks, though complex or highly specialized intelligence requirements may still need human review.
Technical feasibility todayclaude-sonnet-52/5Products like GitHub Copilot and GIS-integrated AI plugins exist and assist coding/mapping tasks, but no deployed system autonomously designs and maintains full intelligence-grade database/GIS/AI infrastructures reliably in production.

Identify gaps in information.

47

CI 25–70 · exposure 45 · augmentation 75 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Intelligence and national security agencies, along with financial crime and competitive intelligence sectors, are actively adopting AI-powered gap analysis and knowledge management systems. Adoption is measurable and accelerating in information-intensive sectors.
Sector adoption velocityclaude-sonnet-52/5Intelligence agencies are cautious, security-conscious adopters of AI, with pilots emerging but production deployment for core analytic judgment tasks still limited and slow relative to commercial sectors.
Augmentation potentialclaude-haiku-4-5-202510014/5AI assistance here is strong: systems can rapidly surface candidate gaps, highlight inconsistencies in source coverage, and prioritize information needs, allowing analysts to focus judgment on strategic gaps and validation rather than manual scanning.
Augmentation potentialclaude-sonnet-54/5AI tools can effectively assist analysts by aggregating data, cross-referencing sources, and surfacing potential inconsistencies or missing information for human review, meaningfully speeding up the gap-identification process.
Task automatabilityclaude-haiku-4-5-202510014/5AI systems can systematically scan structured and unstructured data sources, cross-reference information against knowledge bases, and flag missing data patterns with >50% time savings. However, some human judgment about what constitutes a 'gap' in a specialized intelligence context may require review, preventing a full 5.
Task automatabilityclaude-sonnet-52/5Identifying information gaps requires deep contextual judgment about mission priorities, source reliability, and analytic tradecraft that current AI cannot reliably replicate end-to-end; AI can flag surface-level missing data but not assess strategic significance of gaps.
Adoption barriersclaude-haiku-4-5-202510012/5While some intelligence work operates under security clearances and classification restrictions, the task of identifying gaps itself—flagging missing pieces—faces few legal or regulatory barriers to automation. Organizational oversight and verification requirements exist but are not hard blockers.
Adoption barriersclaude-sonnet-54/5Intelligence work involves classification, security clearances, and accountability structures that require human analysts to own judgments about gaps, creating strong organizational and legal barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510014/5Automated information-gap analysis is substantially cheaper than paying analysts to manually review sources and cross-check completeness; inference and data processing costs are low relative to loaded analyst salaries, though integration overhead applies.
Cost vs. human wageclaude-sonnet-52/5Given the need for extensive human oversight, classified data handling, and validation of AI outputs, the effective cost of using AI for this nuanced judgment task is not dramatically cheaper than skilled analyst time.
Technical feasibility todayclaude-haiku-4-5-202510013/5Multiple commercial tools (e.g., knowledge graph platforms, data integration software, OSINT analysis tools) can identify missing information across datasets, but they require careful configuration and tuning, and error rates on novel intelligence domains remain material. Deployed but not yet fully reliable at scale.
Technical feasibility todayclaude-sonnet-52/5Some analytic support tools and RAG-based systems can highlight missing data points or contradictions in document sets, but no deployed product reliably performs holistic intelligence gap analysis at production scale within intelligence agencies.

Study communication code languages or foreign languages to translate intelligence.

42

CI 36–49 · exposure 42 · augmentation 88 · importance 3.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Government intelligence agencies adopt AI translation tools gradually in pilot and augmentative roles, with security and classification constraints slowing deployment. Adoption is faster in private sector translation services and some open-source monitoring, but production-scale full automation in intelligence remains limited and cautious.
Sector adoption velocityclaude-sonnet-53/5Intelligence agencies are adopting AI-assisted translation and NLP tools steadily, but classified environments and cautious government procurement slow full-scale deployment compared to commercial sectors.
Augmentation potentialclaude-haiku-4-5-202510014/5AI translation systems significantly boost analyst productivity by rapidly processing large volumes of foreign-language material and suggesting translations, allowing analysts to focus on interpretation, context, and code-breaking rather than rote translation. This is a mature augmentation use case across intelligence organizations.
Augmentation potentialclaude-sonnet-55/5AI translation tools dramatically speed up initial processing and triage of foreign-language material, letting analysts focus on interpretation, verification, and contextual judgment.
Task automatabilityclaude-haiku-4-5-202510012/5Translation of foreign languages can be partially automated with modern NMT systems achieving 50%+ time savings on routine texts, but intelligence-grade translation demands cultural context, colloquialisms, and code-breaking expertise that current AI handles unreliably without human oversight. The 'communication code languages' aspect—cryptography and ciphers—remains largely resistant to full automation without human cryptanalytic judgment.
Task automatabilityclaude-sonnet-53/5Modern machine translation and NLP tools can translate large volumes of foreign-language text quickly, but nuanced code language, slang, or coded intelligence communications still require human interpretation and validation for accuracy and context.
Adoption barriersclaude-haiku-4-5-202510014/5Intelligence work is heavily regulated and often classified; government agencies require cleared personnel to handle sensitive translations, and there are statutory oversight mandates on automated intelligence processing. Liability and tradecraft requirements create high organizational friction and legal barriers to full automation without human sign-off.
Adoption barriersclaude-sonnet-54/5Intelligence work requires security clearances, chain-of-custody, and accountability for translation accuracy in national security contexts, creating strong institutional and legal barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510014/5Modern neural translation inference is extremely cheap per token compared to the loaded salary of skilled linguists and intelligence analysts ($80k–150k+ annually). Even accounting for human oversight and integration, AI-assisted translation is typically 5–10x cheaper than pure human translation for high-volume routine work.
Cost vs. human wageclaude-sonnet-53/5AI translation is cheap per word, but oversight, security clearance requirements, and validation of intelligence-grade translations add substantial human cost, making the net savings moderate rather than order-of-magnitude.
Technical feasibility todayclaude-haiku-4-5-202510013/5Commercial translation products (Google Translate, DeepL) perform well on standard text, and some organizations deploy neural machine translation in production pipelines. However, intelligence agencies rarely rely on off-the-shelf systems alone for sensitive communications; they use hybrid human-AI workflows with material residual error rates on classified or coded material that demand expert human review.
Technical feasibility todayclaude-sonnet-53/5Deployed translation products (e.g., government-grade MT systems) exist and are used operationally, but they still require human review for high-stakes intelligence work, especially with ciphers or obscure dialects.

Prepare plans to intercept foreign communications transmissions.

41

CI 9–73 · exposure 53 · augmentation 63 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5U.S. and allied intelligence agencies have been systematically deploying AI and machine learning for SIGINT automation over the past decade; this is among the fastest-adopting sectors for AI-driven automation due to national security mandate and existing technical infrastructure.
Sector adoption velocityclaude-sonnet-51/5Intelligence agencies are notoriously slow and cautious in adopting AI for core operational planning due to security, reliability, and accountability concerns, keeping actual deployment minimal.
Augmentation potentialclaude-haiku-4-5-202510014/5AI systems significantly amplify human analyst productivity by pre-filtering massive signal volumes, automatically flagging suspicious patterns, and surfacing priority targets for human review and decision-making, allowing analysts to focus on interpretation and actionable intelligence rather than raw monitoring.
Augmentation potentialclaude-sonnet-53/5AI can assist with data analysis, pattern recognition, and organizing intercepted metadata to support human analysts, but the strategic planning itself remains a human-centered judgment task.
Task automatabilityclaude-haiku-4-5-202510015/5Modern AI systems excel at signal detection, pattern recognition, filtering, and prioritization of communications data at scale. Intercepting (capturing and routing) communications transmissions is largely a technical/algorithmic task where AI can autonomously perform signal processing, frequency analysis, and automated sorting with >50% time savings compared to manual monitoring.
Task automatabilityclaude-sonnet-52/5Planning signal interception requires deep domain expertise, classified technical knowledge, and real-time strategic judgment that current AI cannot autonomously replicate end-to-end.a Portions like data collation could be assisted, but the core planning remains human-driven.
Adoption barriersclaude-haiku-4-5-202510015/5This task sits squarely within classified government authority and is subject to strict legal authorization, oversight, and regulatory frameworks (Foreign Intelligence Surveillance Act, executive orders, agency controls). Only licensed/authorized personnel and systems can legally conduct foreign communications interception, creating hard legal and organizational barriers.
Adoption barriersclaude-sonnet-55/5This task involves classified national security operations requiring security clearances, legal authorization (e.g., FISA-type approvals), and strict chain-of-command sign-off, making it one of the most barrier-protected tasks conceivable.
Cost vs. human wageclaude-haiku-4-5-202510015/5Once infrastructure is in place, the marginal cost of AI-driven signal processing and filtering is negligible compared to staffing teams of human analysts 24/7 to monitor and manually sort communications data; the cost advantage is orders of magnitude.
Cost vs. human wageclaude-sonnet-51/5Given the sensitivity, required clearances, and need for exhaustive human validation, any AI assistance adds oversight costs rather than reducing them relative to skilled analyst labor.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed signals intelligence (SIGINT) systems already use AI for automated signal detection, classification, and routing in production environments within government agencies. Commercial and government signal-processing tools incorporate machine learning for frequency identification and traffic filtering, though integration with operational planning workflows remains semi-manual.
Technical feasibility todayclaude-sonnet-51/5No deployed commercial or government AI product autonomously creates interception plans; this remains a highly classified, human-led analytical process with no publicized production system performing it.

Gather, analyze, correlate, or evaluate information from a variety of resources, such as law enforcement databases.

35

CI 25–45 · exposure 38 · augmentation 75 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Intelligence agencies are conservative adopters of AI for analytical tasks due to security, classification, and liability concerns. While some pilot programs exist, production deployment of autonomous AI analysis remains limited and heavily human-supervised.
Sector adoption velocityclaude-sonnet-53/5Government and law enforcement sectors adopt analytic AI tools cautiously due to security, privacy and accountability concerns, with pilots and narrow production tools more common than broad deployment.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can substantially augment intelligence analysts by automating database searches, suggesting correlations, flagging patterns, and summarizing large datasets, allowing human analysts to focus on judgment-heavy interpretation and source evaluation. This productivity boost is already beginning in real intelligence organizations.
Augmentation potentialclaude-sonnet-54/5AI substantially augments analysts by rapidly surfacing correlations, summarizing large volumes of records, and highlighting anomalies, meaningfully speeding up the gathering and initial evaluation phases while humans retain interpretive control.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist with data extraction and correlation from structured databases, but the task inherently requires human judgment to evaluate reliability of sources, resolve contradictions, and draw contextual conclusions about sensitive intelligence matters. Full end-to-end automation would fail the equal-quality bar due to the need for nuanced reasoning about source credibility and geopolitical context.
Task automatabilityclaude-sonnet-53/5AI can rapidly gather and correlate structured data across databases and flag patterns, but the correlation of ambiguous, incomplete, or classified intelligence with contextual judgment on threat significance still requires substantial human analysis and verification.
Adoption barriersclaude-haiku-4-5-202510014/5Intelligence analysis typically requires security clearances, legal authorization to access classified databases, and organizational/legal accountability for conclusions. These licensing and liability requirements create substantial barriers to full automation and delegation to unsupervised AI systems.
Adoption barriersclaude-sonnet-54/5Intelligence analysis involves classified systems, chain-of-custody, legal authorization requirements for law enforcement database access, and accountability structures that generally mandate a cleared human analyst's sign-off.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI inference for data correlation is cheap, but the extensive human oversight, validation, and re-analysis needed to ensure accuracy in intelligence contexts makes the total cost-per-task comparable to or higher than employing trained analysts.
Cost vs. human wageclaude-sonnet-53/5AI-assisted correlation tools reduce analyst hours significantly for data aggregation, but licensing, security-cleared infrastructure, and mandatory human review keep total costs closer to comparable rather than order-of-magnitude cheaper.
Technical feasibility todayclaude-haiku-4-5-202510012/5Tools exist for database querying and pattern detection, but no mature deployed product reliably performs the full 'gather, analyze, correlate, and evaluate' pipeline with the judgment required in intelligence work. Existing systems require substantial human validation and cannot independently assess source credibility.
Technical feasibility todayclaude-sonnet-53/5Deployed analytic platforms (e.g., Palantir, i2, various OSINT/law enforcement fusion tools) perform data correlation and link analysis in production, but final analytic judgments and cross-source evaluation still rely heavily on trained analysts due to error rates and sensitivity.

Evaluate records of communications, such as telephone calls, to plot activity and determine the size and location of criminal groups and members.

34

CI 25–43 · exposure 42 · augmentation 75 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While law enforcement and intelligence agencies experiment with automated pattern-detection and network-mapping tools, adoption remains measured and cautious due to legal constraints, civil liberties concerns, and the need for human validation. Most organizations still rely heavily on human-led investigation rather than AI-driven identification.
Sector adoption velocityclaude-sonnet-52/5Government intelligence and law enforcement sectors are typically slow adopters of AI due to security clearance requirements, legacy systems, and cautious procurement cycles, despite some high-profile pilot deployments.
Augmentation potentialclaude-haiku-4-5-202510014/5AI excels at rapid visualization of communication networks, anomaly detection, and pattern surfacing (e.g., identifying previously unconnected individuals), substantially accelerating the analyst's investigative reach and enabling them to focus expertise on validation and interpretation rather than manual data compilation.
Augmentation potentialclaude-sonnet-54/5AI-powered link analysis, pattern detection, and natural language processing on communication records substantially speed up an analyst's ability to visualize networks and surface leads, while the analyst retains interpretive and decision-making control.
Task automatabilityclaude-haiku-4-5-202510012/5AI can extract and categorize communication metadata (phone numbers, timestamps, call patterns) and flag suspicious clusters, but determining criminal group membership, hierarchy, and intent requires contextual judgment, legal interpretation, and source evaluation that remains heavily human-dependent. Significant manual review and expert validation would still be required.
Task automatabilityclaude-sonnet-53/5AI can process call records, build link-analysis graphs, and flag patterns, but final determination of group size/location requires human judgment integrating context, corroboration, and ambiguity resolution that current systems cannot fully replace.'
Adoption barriersclaude-haiku-4-5-202510015/5This task is heavily regulated by law enforcement and intelligence frameworks (Fourth Amendment, wiretap laws, FISA, agency oversight); authorized human intelligence analysts must legally collect, interpret, and sign off on findings, especially for sensitive cases. Liability for misidentification creates a hard requirement for human accountability and judgment.
Adoption barriersclaude-sonnet-54/5Intelligence and law enforcement work involves classified data handling, chain-of-custody and evidentiary standards, legal authorization requirements, and accountability structures that generally mandate human analysts to interpret and certify findings.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI infrastructure for communication analysis (data ingestion, pattern detection, visualization) is relatively low-cost, but the requirement for skilled human analysts to validate, contextualize, and act on findings means the per-task cost remains comparable to or only modestly cheaper than human-only workflows.
Cost vs. human wageclaude-sonnet-52/5Specialized intelligence software licensing, secure infrastructure, and mandatory human oversight for legal/operational accuracy keep costs relatively high compared to a human analyst's marginal output, though some efficiency gains exist.
Technical feasibility todayclaude-haiku-4-5-202510013/5Social network analysis tools and communication pattern-detection systems exist in production (used by law enforcement agencies), but they generate high false-positive rates and require substantial human oversight to avoid misidentification. The reliability gap is particularly acute in distinguishing coincidental patterns from genuine criminal coordination.
Technical feasibility todayclaude-sonnet-53/5Deployed link-analysis and social network analysis tools (e.g., Palantir, i2 Analyst's Notebook with AI features) are used by intelligence/law enforcement agencies today, but they still require significant analyst interpretation and have real error rates on ambiguous data.

Analyze intelligence data to identify patterns and trends in criminal activity.

32

CI 25–39 · exposure 38 · augmentation 75 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Intelligence and law enforcement agencies are conservative adopters; pilots exist but production AI-driven analysis remains limited; organizational inertia, compliance requirements, and the high cost of analytical errors slow deployment compared to commercial sectors.
Sector adoption velocityclaude-sonnet-52/5Government and law enforcement intelligence sectors are historically slow adopters of AI due to security, procurement, and legal constraints, with pilots more common than deep production use.
Augmentation potentialclaude-haiku-4-5-202510014/5AI assistants already meaningfully augment analysts by rapidly surfacing patterns, cross-referencing datasets, and highlighting anomalies—transforming the speed of hypothesis generation while humans retain judgment on interpretation, source credibility, and investigative prioritization.
Augmentation potentialclaude-sonnet-54/5AI tools significantly enhance analysts' ability to sift large datasets, flag anomalies, and visualize networks, substantially boosting productivity while humans retain interpretive and decision-making control.
Task automatabilityclaude-haiku-4-5-202510013/5AI can assist with pattern detection, anomaly identification, and trend visualization in structured data, but the task requires contextual judgment, source evaluation, and linking disparate intelligence types—all areas where human expertise remains critical for reliable output. Roughly half the analytical workflow could be automated with significant setup of training data and domain validation.
Task automatabilityclaude-sonnet-52/5AI can process structured data and surface candidate patterns, but synthesizing multi-source intelligence into validated trends requires contextual judgment, source credibility assessment, and adversarial reasoning that current systems cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Strong legal and regulatory frameworks govern intelligence analysis; classified or sensitive data handling requires security clearances and authorized personnel; liability for erroneous analysis affecting enforcement actions; organizational protocols typically mandate senior analyst sign-off rather than autonomous systems.
Adoption barriersclaude-sonnet-54/5Intelligence analysis often involves classified data, chain-of-custody, legal admissibility, and accountability requirements that mandate human analyst sign-off, creating strong institutional and regulatory barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5While AI inference costs are low, the integration burden (data pipeline setup, model tuning for domain-specific patterns, continuous human expert review of outputs) and oversight demands make the all-in cost comparable to or slightly higher than human analyst time.
Cost vs. human wageclaude-sonnet-52/5Specialized intelligence software and cleared personnel oversight are expensive, and the analytical judgment layer still requires costly human analysts, keeping all-in AI costs comparable to or only modestly below human costs.
Technical feasibility todayclaude-haiku-4-5-202510013/5Commercial AI products (graph analytics, anomaly detection, document summarization) operate in intelligence and law enforcement settings, but with material error rates and high false-positive rates that demand expert human review. Production deployments exist but require substantial human oversight rather than autonomous end-to-end analysis.
Technical feasibility todayclaude-sonnet-52/5Deployed analytics and NLP tools exist in law enforcement/intelligence settings for link analysis and anomaly detection, but they are narrow-scope aids with significant false-positive rates, not reliable end-to-end pattern analysts.

Link or chart suspects to criminal organizations or events to determine activities and interrelationships.

26

CI 25–28 · exposure 25 · augmentation 75 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Intelligence agencies and law enforcement have adopted graph databases and visualization tools, but AI-driven autonomous link analysis and relationship inference are still in pilot phases; deployment remains cautious due to accountability concerns.
Sector adoption velocityclaude-sonnet-52/5Law enforcement and intelligence agencies are cautious adopters due to security, procurement cycles, and legal sensitivities, with AI tools used as aids rather than replacing analytic judgment; adoption is slower than in commercial information sectors.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can significantly assist analysts by rapidly suggesting potential links, visualizing network graphs, and flagging statistical anomalies, substantially increasing the speed and scope of manual investigative review while keeping the analyst in control of conclusions.
Augmentation potentialclaude-sonnet-54/5AI-powered link-chart and network-visualization tools significantly speed up data organization, pattern detection, and relationship mapping, meaningfully boosting analyst productivity while the analyst retains interpretive control.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist with pattern matching and link visualization, but determining causal relationships and interpreting complex interrelationships between suspects and organizations requires human judgment informed by context, sources, and investigative expertise that AI systems cannot reliably perform end-to-end.
Task automatabilityclaude-sonnet-52/5AI can assist with entity extraction and link-chart drafting from structured data, but synthesizing disparate intelligence sources, judging reliability, and inferring meaningful relationships requires human tradecraft and contextual judgment that current systems cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Intelligence and law enforcement organizations operate under strict legal, evidentiary, and regulatory constraints; findings must be defensible in court, liability for false accusations is high, and human analysts typically must review and authorize conclusions, creating strong adoption friction.
Adoption barriersclaude-sonnet-54/5Intelligence work involves security clearances, chain-of-custody and evidentiary standards, and legal/institutional accountability for conclusions used in investigations or prosecutions, creating strong barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Integrating AI link-analysis tools into analyst workflows, plus the required human oversight and expert validation, approaches human labor cost; cost savings are modest because human intelligence work remains central.
Cost vs. human wageclaude-sonnet-52/5Licensing, integration, and required human oversight for classified/sensitive intelligence work keep AI costs comparable to or only modestly below analyst costs once accuracy and security requirements are factored in.
Technical feasibility todayclaude-haiku-4-5-202510012/5Graph visualization and entity-linking tools exist, but no deployed AI system reliably determines the investigative meaning of relationships or produces analyst-ready conclusions about criminal networks at scale; most tools require heavy human interpretation and validation.
Technical feasibility todayclaude-sonnet-52/5Link analysis tools with AI-assisted graph generation exist (e.g., Palantir, i2 Analyst's Notebook) but require significant human curation and validation; fully autonomous, reliable production use for this specific analytic task is not demonstrated at scale.

Study the assets of criminal suspects to determine the flow of money from or to targeted groups.

26

CI 25–28 · exposure 25 · augmentation 63 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of autonomous financial analysis in intelligence agencies remains cautious and slow; most deployments are still pilots or assistive tools rather than replacement systems, reflecting both cultural conservatism and the high stakes of error in national security contexts.
Sector adoption velocityclaude-sonnet-53/5Law enforcement and intelligence agencies have adopted AI-assisted financial analysis tools at a moderate pace, with pilots and specialized vendor products in use, but broad deployment remains slower than in commercial finance sectors.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully augment analysts by automating data collection, visualizing transaction networks, flagging statistical anomalies, and cross-referencing public records, allowing analysts to focus on judgment-intensive pattern interpretation and investigation strategy.
Augmentation potentialclaude-sonnet-54/5AI significantly speeds up data aggregation, network/link analysis, and anomaly detection across large financial datasets, meaningfully augmenting analyst productivity even though final determinations remain human-led.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with data aggregation and pattern detection in financial networks, the task requires nuanced judgment about intent, contextual understanding of criminal schemes, and integration of disparate intelligence sources that current systems struggle with reliably. Significant human oversight remains essential for accuracy and legal defensibility.
Task automatabilityclaude-sonnet-52/5AI can assist with pattern detection and data aggregation in financial flows, but determining actual asset ownership, tracing money flows to criminal groups, and drawing investigative conclusions requires contextual judgment, corroboration, and legal interpretation that current systems cannot fully replace.
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: intelligence analysis is typically conducted by licensed government or security personnel, legal standards for evidentiary adequacy are strict, and liability for false allegations is high. Regulatory oversight and the requirement for human certification of findings protect this task from automation.
Adoption barriersclaude-sonnet-54/5This task often involves classified or legally sensitive information, chain-of-custody requirements for prosecutions, and analyst certifications, creating strong institutional and legal barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI-assisted financial analysis tools cost significant amounts to license, integrate, and operate with necessary human oversight, often approaching or exceeding the cost of a trained analyst's time for this specialized, high-stakes work.
Cost vs. human wageclaude-sonnet-52/5Specialized financial intelligence software and data licensing costs are substantial, and human analyst oversight remains mandatory, so total cost savings versus a trained analyst are modest rather than order-of-magnitude.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs this end-to-end financial intelligence analysis task independently. Although financial monitoring tools and graph-analysis software exist, they require heavy human interpretation and struggle with the complex, adversarial reasoning needed to determine money flow intent in criminal investigations.
Technical feasibility todayclaude-sonnet-52/5Financial crime analytics tools (e.g., transaction monitoring, link analysis software) are deployed in production, but they primarily flag anomalies rather than autonomously conducting full asset investigations tying suspects to targeted groups.

Conduct presentations of analytic findings.

26

CI 25–28 · exposure 25 · augmentation 75 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Intelligence agencies and security-conscious sectors are beginning to pilot AI-assisted presentation tools (drafting, summarization), but actual adoption of autonomous or heavily AI-driven briefings remains limited by classification, oversight, and the need for human accountability in high-stakes contexts.
Sector adoption velocityclaude-sonnet-52/5Intelligence and government sectors are typically slower adopters of AI tools for sensitive analytic and briefing functions due to security, classification, and trust concerns.
Augmentation potentialclaude-haiku-4-5-202510014/5AI excels at augmenting this task: automatically generating slide layouts, summarizing complex data, drafting narratives, and organizing findings allows analysts to focus on message refinement, audience adaptation, and confident delivery—significantly raising their productivity while keeping them in control.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist analysts in drafting presentation content, summarizing findings, and creating visualizations, significantly speeding up preparation while the human still delivers and defends the analysis.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist with generating slides, organizing data visuals, and drafting talking points, but delivering a presentation requires real-time audience engagement, adaptive communication, fielding unexpected questions, and establishing credibility—capabilities that current AI systems cannot reliably handle end-to-end while maintaining the persuasive, contextual quality intelligence analysts must achieve.
Task automatabilityclaude-sonnet-52/5AI can help draft slides and summaries, but delivering a presentation involves live verbal communication, audience interaction, and answering probing questions that require human judgment and credibility, especially in classified/sensitive intelligence contexts.
Adoption barriersclaude-haiku-4-5-202510014/5Intelligence presentations often require security clearance, authorized access to classified material, and organizational sign-off before briefing; the analyst must personally attest to findings and answer adversarial questions, creating legal and institutional barriers to full automation.
Adoption barriersclaude-sonnet-54/5Intelligence briefings often require cleared personnel, accountability for judgments, and direct interaction with policymakers, creating strong organizational and security-based barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI-assisted presentation prep (content drafting, visualization) reduces some overhead, but the analyst must still customize findings, verify accuracy, rehearse, and deliver in person; total cost savings are modest compared to the analyst's loaded wage for this task.
Cost vs. human wageclaude-sonnet-52/5While AI can cheaply assist with content prep, the actual delivery task still requires a human analyst's time, judgment, and security clearance, limiting cost savings for the full task.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI tools can draft presentation content and generate visuals, but no deployed system reliably conducts a full intelligence briefing autonomously; existing products are narrow (slide generation, data visualization) and require substantial human refinement, review, and delivery.
Technical feasibility todayclaude-sonnet-52/5Products exist for generating slide decks and summarizing analytic content, but no deployed system autonomously conducts intelligence briefings to stakeholders or decision-makers in production settings.

Gather and evaluate information, using tools such as aerial photographs, radar equipment, or sensitive radio equipment.

26

CI 20–32 · exposure 30 · augmentation 75 · importance 3.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5U.S. intelligence agencies have invested heavily in AI for imagery and signals analysis (GEOINT, SIGINT programs), with production deployments emerging, but adoption remains concentrated in large government organizations. Broader commercial adoption lags due to regulatory and classification constraints.
Sector adoption velocityclaude-sonnet-52/5Defense and intelligence sectors are historically slow to adopt AI at scale for core analytic tasks due to security, verification, and bureaucratic constraints, though pilots in image/signal processing are increasing.
Augmentation potentialclaude-haiku-4-5-202510014/5AI augmentation is already substantial: automated flagging of anomalies in imagery, pattern detection in signals, and rapid processing of large sensor datasets significantly enhance analyst productivity. The human remains central for judgment, but AI transforms the speed and scale of information processing available to them.
Augmentation potentialclaude-sonnet-54/5AI tools substantially assist by pre-processing imagery, flagging anomalies in radar/sensor data, and filtering large volumes of intercepted communications, letting analysts focus on higher-level evaluation.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can process and analyze aerial photographs and radar data through computer vision and signal processing, the task requires human judgment to evaluate relevance, assess source reliability, and integrate disparate information streams into actionable intelligence. Current systems can automate feature detection but not the full evaluative loop.
Task automatabilityclaude-sonnet-52/5Physical operation of aerial photography, radar, and sensitive radio equipment plus real-time evaluation of ambiguous, adversarial data requires human judgment and classified context that current AI cannot fully replace end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Intelligence analysis operates under strict security clearance, classification, and regulatory frameworks (NSA, DOD, intelligence community authorities). Legal liability for intelligence errors, need for human accountability in classified contexts, and authorization requirements to handle sensitive sources create substantial barriers to full automation.
Adoption barriersclaude-sonnet-55/5Intelligence work involves security clearances, legal authorization for surveillance, chain-of-custody requirements, and strict human accountability for classified judgments, creating hard regulatory and institutional barriers.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for image/signal analysis are expensive (licensing, compute, specialized hardware like radar systems), and integration into intelligence workflows requires significant oversight infrastructure. The loaded cost of trained intelligence analysts is high, but the AI infrastructure cost per decision remains substantial.
Cost vs. human wageclaude-sonnet-52/5Specialized sensor equipment, secure data pipelines, and human oversight for classified intelligence work keep AI costs comparable to or above human analyst costs, despite some efficiency gains in raw data processing.
Technical feasibility todayclaude-haiku-4-5-202510013/5Deployed products exist for image analysis (satellite/aerial imagery classification) and signal processing, but they perform narrow subtasks reliably rather than the full gather-and-evaluate workflow. Human analysts still verify outputs and handle contextual interpretation that varies by intelligence requirements.
Technical feasibility todayclaude-sonnet-52/5Deployed products exist for image classification and signal processing (e.g., object detection in satellite imagery), but integrated multi-source gathering and evaluation with reliable accuracy at scale remains largely research or narrowly deployed within classified systems.

Validate known intelligence with data from other sources.

23

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

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Intelligence agencies operate in conservative, regulated environments with slow technology adoption cycles. While some agencies pilot AI tools for data integration, actual production displacement of validation tasks remains limited and measured.
Sector adoption velocityclaude-sonnet-52/5Intelligence agencies are historically slow, security-constrained adopters of AI tools, with pilots and narrow deployments rather than widespread production use.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist analysts by rapidly cross-referencing multiple data sources, highlighting contradictions, and surfacing relevant historical patterns, which meaningfully speeds human validation workflows. The human analyst retains judgment over credibility assessment and final sign-off.
Augmentation potentialclaude-sonnet-54/5AI can significantly speed up cross-referencing, pattern detection, and initial flagging of corroborating or conflicting data, meaningfully augmenting analyst throughput while humans retain final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can cross-reference data sources and flag inconsistencies, validating intelligence requires nuanced judgment about source credibility, geopolitical context, and subtle disinformation patterns that current systems handle poorly. Meaningful time savings would require human oversight on most outputs, limiting automation to specific data-matching subtasks.
Task automatabilityclaude-sonnet-52/5Cross-referencing data can be partly assisted by AI (search, comparison, flagging discrepancies), but true validation requires judgment about source reliability, context, and tradecraft that current systems cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Intelligence validation is heavily regulated with strict authorization requirements; humans with security clearances must sign off on validated findings. Liability for erroneous intelligence and national security implications create strong legal and organizational barriers to full automation.
Adoption barriersclaude-sonnet-55/5This task sits within classified, high-stakes intelligence work requiring cleared personnel, legal authority, and accountability for judgments—hard institutional and legal barriers prevent full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Integration costs for connecting classified and unclassified sources, combined with required expert oversight, make AI-assisted validation comparable to or more expensive than direct human analysis for sensitive intelligence work. Overhead per validated item remains high.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply process large volumes of data, but the human oversight, security clearance, and verification loop required keeps overall cost comparable to or only modestly cheaper than human-only validation.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably performs end-to-end intelligence validation today. Tools exist for data integration and anomaly detection, but production systems in intelligence agencies still require trained human analysts to assess source reliability and validate findings against classified context.
Technical feasibility todayclaude-sonnet-52/5Some deployed intelligence-community tools use AI to correlate and flag data across sources, but reliable, trusted validation at production scale remains narrow and heavily human-reviewed.

Prepare comprehensive written reports, presentations, maps, or charts, based on research, collection, and analysis of intelligence data.

23

CI 20–25 · exposure 25 · augmentation 75 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Intelligence agencies have been slow to adopt AI for core analytic functions beyond narrow, well-scoped tasks. Security constraints, organizational conservatism, and the need for human accountability limit production deployment of AI-generated intelligence reports.
Sector adoption velocityclaude-sonnet-52/5Intelligence agencies are cautious adopters due to security, classification, and reliability concerns, resulting in slower and shallower AI deployment compared to fast-moving private-sector information services.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist intelligence analysts by rapidly generating draft text, visualizing complex datasets, summarizing open-source materials, and creating presentation templates, allowing analysts to focus on judgment, source evaluation, and tradecraft. These assistive uses are increasingly common in intelligence organizations.
Augmentation potentialclaude-sonnet-54/5AI substantially aids drafting, summarization, translation, data visualization, and pattern detection, meaningfully speeding up report preparation while analysts retain judgment and final responsibility.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can draft text, generate charts, and create maps from structured data, intelligence reporting requires synthesis of sensitive, classified information with contextual judgment, source evaluation, and compliance with strict dissemination protocols. Current systems cannot autonomously perform the full end-to-end task at equal quality without substantial human review and validation.
Task automatabilityclaude-sonnet-52/5AI can draft portions of reports and generate visualizations from structured data, but synthesizing classified/sensitive intelligence into judgment-laden analytic products requires human tradecraft, sourcing evaluation, and accountability that current systems cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Intelligence report preparation is heavily regulated by classification standards, legal review requirements, and organizational protocols mandating human analyst sign-off. Security clearances, need-to-know restrictions, and liability for misclassification create hard barriers to full automation or unsupervised AI use.
Adoption barriersclaude-sonnet-54/5Intelligence reports often require security clearances, chain-of-custody, accountability for analytic judgments, and formal sign-off, creating strong organizational and regulatory barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI text and visualization tools have low marginal cost, but intelligence reporting requires significant human oversight to verify accuracy, check classification levels, and ensure analytic rigor. The total cost including required human validation remains comparable to or exceeds the cost of a skilled analyst producing the initial report.
Cost vs. human wageclaude-sonnet-52/5While AI drafting tools are cheap per query, the required human verification, security clearance overhead, and integration into classified systems keep effective all-in cost close to or above human analyst cost for this task.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI tools exist for text generation, visualization, and data summarization, but no deployed product reliably performs comprehensive intelligence report preparation at scale. Existing systems lack the ability to handle classified data securely, validate source credibility, and ensure analytic tradecraft standards required in intelligence agencies.
Technical feasibility todayclaude-sonnet-52/5Commercial LLMs and BI tools reliably produce drafts, summaries, and charts from clean data, but no deployed product performs the full intelligence analysis-to-report pipeline reliably in classified/high-stakes production environments without heavy human oversight.

Establish criminal profiles to aid in connecting criminal organizations with their members.

23

CI 20–25 · exposure 25 · augmentation 63 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Law enforcement and intelligence agencies adopt AI tools slowly due to regulatory scrutiny, liability concerns, and workforce culture. Pilots exist but operational deployment of autonomous profiling remains limited; agencies typically treat AI as a search and flagging tool rather than a decision-making system.
Sector adoption velocityclaude-sonnet-52/5Intelligence and law enforcement agencies are historically slow adopters of AI due to security, legal, and procurement constraints, though some pilot programs for data fusion exist.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist analysts by rapidly surfacing network connections, flagging historical patterns, and cross-referencing databases, helping analysts focus investigation effort. However, augmentation is limited to data retrieval and pattern suggestion; the critical judgment of affiliation strength and organizational role still falls to the human analyst.
Augmentation potentialclaude-sonnet-54/5AI can significantly speed up data aggregation, network analysis, and pattern detection across large datasets, meaningfully augmenting analyst productivity while humans retain final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with pattern recognition across criminal databases and structured data correlation, establishing comprehensive criminal profiles requires deep contextual judgment, source credibility assessment, and understanding of organizational hierarchies that current systems struggle with at production quality. The task involves synthesizing disparate intelligence sources and making probabilistic judgments about affiliations—AI can support components but cannot reliably replace the analyst's integrative reasoning end-to-end.
Task automatabilityclaude-sonnet-52/5Requires synthesizing sensitive, ambiguous, often incomplete multi-source intelligence with judgment about intent and reliability that current AI cannot fully replicate end-to-end, though it can assist with pattern-matching and data linkage.
Adoption barriersclaude-haiku-4-5-202510014/5Criminal profiling informing arrests and prosecutions operates under strict legal discovery and due-process requirements; law enforcement agencies face liability for AI-generated profiles used in investigative decisions. Additionally, many jurisdictions impose human authorization requirements for intelligence products that feed into investigative leads, and institutional caution around AI bias in criminal justice remains high.
Adoption barriersclaude-sonnet-55/5This involves classified/law-enforcement-sensitive work requiring security clearances, legal authorization, chain-of-custody standards, and accountability for decisions affecting investigations and civil liberties.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI-driven profiling tools require significant infrastructure (data integration, validation workflows, security), and the necessary human oversight remains intensive because errors carry high operational and legal costs. The all-in cost per profile remains comparable to or exceeds a trained analyst's time, especially accounting for rework.
Cost vs. human wageclaude-sonnet-52/5Specialized intelligence software plus required human vetting, security clearances, and oversight keep costs relatively high compared to fully autonomous AI use, though it may reduce analyst hours somewhat.
Technical feasibility todayclaude-haiku-4-5-202510012/5No mature product reliably performs full criminal profiling at scale in operational intelligence settings. Some vendors offer graph databases and pattern-matching tools for criminal networks, but these require substantial human curation, validation, and legal review; production deployments show material gaps in accuracy and false-positive rates when autonomous.
Technical feasibility todayclaude-sonnet-52/5Some link-analysis and entity-resolution tools (Palantir, i2 Analyst's Notebook with AI features) are deployed in production, but fully automated profile generation with reliable accuracy is not standard practice.

Predict future gang, organized crime, or terrorist activity, using analyses of intelligence data.

23

CI 20–25 · exposure 25 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Intelligence agencies have experimented with AI tools for data triage and pattern matching, but adoption of AI systems for autonomous or semi-autonomous predictive intelligence remains slow due to regulatory constraints, classification barriers, and institutional conservatism around national security automation.
Sector adoption velocityclaude-sonnet-52/5Intelligence and law enforcement sectors are cautious adopters of AI for predictive judgments due to civil liberties concerns, classification barriers, and accountability requirements, keeping deployment mostly at pilot stage.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist intelligence analysts by rapidly sifting large datasets, flagging statistical anomalies, identifying network patterns, and surfacing relevant historical cases for human review, materially accelerating the human analyst's ability to generate and test hypotheses while they retain judgment over actionable conclusions.
Augmentation potentialclaude-sonnet-54/5AI tools significantly help analysts by processing large volumes of data, flagging patterns, and surfacing links, substantially boosting productivity even though final predictive judgment remains human-driven.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with pattern detection and data correlation in large intelligence datasets, predicting future criminal or terrorist activity requires nuanced geopolitical context, source credibility assessment, and human judgment about intent and capability that current systems struggle with reliably. End-to-end automation would need to match human analysts' ability to synthesize disparate signals into actionable forecasts at 50% time savings, which is not yet demonstrated.
Task automatabilityclaude-sonnet-52/5Predicting future criminal/terrorist activity requires synthesizing ambiguous, incomplete, adversarial data and exercising judgment about intent and capability that current AI cannot reliably replicate end-to-end.','A
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: intelligence work is tightly regulated by law enforcement and national security frameworks, classification requirements restrict AI system inputs, and human analysts with security clearances and organizational accountability remain legally and operationally necessary to act on predictions. Liability for missed threats or false positives creates high error-cost asymmetry favoring human oversight.
Adoption barriersclaude-sonnet-55/5This is classified, security-cleared work with legal accountability, chain-of-custody, and oversight requirements that mandate human analysts sign off on threat assessments.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI infrastructure (data ingestion, model development, continuous retraining, human oversight) to support predictive intelligence analysis is expensive and still requires skilled human analysts to validate and act on outputs, making all-in costs comparable to or higher than traditional human-only analysis.
Cost vs. human wageclaude-sonnet-52/5Given high error costs and the need for extensive human oversight and validation of any AI-generated forecast, all-in costs remain comparable to or higher than skilled analyst labor.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed AI product reliably predicts future organized crime or terrorist activity at scale with acceptable error rates; the stakes and complexity mean intelligence agencies remain heavily human-dependent. Research prototypes and experimental systems exist, but production deployment for this high-consequence task is minimal and success metrics remain elusive.
Technical feasibility todayclaude-sonnet-52/5Deployed intelligence-fusion and pattern-detection tools assist analysts but no production system autonomously generates trusted predictive threat assessments without heavy human validation.

Study activities relating to narcotics, money laundering, gangs, auto theft rings, terrorism, or other national security threats.

23

CI 20–25 · exposure 25 · augmentation 75 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption in intelligence agencies is slower than in commercial sectors because of classification restrictions, risk aversion around errors, legacy systems integration challenges, and the requirement for human accountability in national security contexts. Pilots exist but replacement deployment remains limited.
Sector adoption velocityclaude-sonnet-52/5Intelligence and law enforcement sectors adopt AI cautiously due to security, legal, and error-cost concerns, with pilots more common than deep production deployment.
Augmentation potentialclaude-haiku-4-5-202510014/5AI tools that surface patterns in large data volumes, flag anomalies, and organize information substantially augment human analysts' productivity, allowing them to focus reasoning on higher-level threat assessment. Current deployed tools in intelligence agencies show strong assistive value while analysts remain decision-makers.
Augmentation potentialclaude-sonnet-54/5AI can significantly aid analysts by rapidly processing large datasets, flagging patterns, and summarizing reports, meaningfully boosting productivity while humans retain judgment and accountability.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist with data processing, pattern recognition, and flagging anomalies in large datasets (financial records, communications), but the task requires deep contextual judgment about threat severity, geopolitical implications, and nuanced interpretation of intelligence that depends on human expertise and real-time situational understanding. Current AI cannot reliably replace the investigative reasoning and sense-making required end-to-end.
Task automatabilityclaude-sonnet-52/5This involves synthesizing classified, ambiguous, and adversarial information requiring judgment, tradecraft, and accountability that current AI cannot fully replicate end-to-end, though parts like data aggregation and pattern flagging can be assisted.
Adoption barriersclaude-haiku-4-5-202510014/5Strong regulatory and organizational barriers exist: intelligence work typically requires security clearances, is bound by classified information protocols, involves legal liability for false leads or missed threats, and is embedded in government and law enforcement agencies with strict authorization requirements. Human experts remain legally and operationally central to the function.
Adoption barriersclaude-sonnet-55/5This work involves classified national security matters, legal chain-of-custody, and analyst certification/clearance requirements, creating hard institutional and legal barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI infrastructure for intelligence analysis (specialized security tools, oversight systems, integration with classified networks) is expensive and requires ongoing human expert oversight. The loaded cost of a skilled intelligence analyst is offset by minimal savings, as AI cannot reduce headcount meaningfully given the need for human judgment and validation throughout the workflow.
Cost vs. human wageclaude-sonnet-52/5AI tools reduce some research time but require expensive secure infrastructure, human vetting, and analyst review, so total cost savings versus a trained analyst are modest.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI tools exist for data mining, network analysis, and document processing in security contexts, no deployed system reliably performs the full analytic task autonomously. Prototype systems may identify patterns, but they require substantial human validation and remain prone to false positives in real intelligence operations where accuracy is critical.
Technical feasibility todayclaude-sonnet-52/5Deployed intelligence platforms use AI for link analysis and anomaly detection, but no product independently conducts full investigative analysis of security threats reliably in production without heavy analyst oversight.

Develop defense plans or tactics, using intelligence and other information.

23

CI 20–25 · exposure 25 · augmentation 75 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While intelligence agencies pilot AI for data processing, actual defense planning automation remains limited by classification requirements, inter-agency governance, and conservatism in military/intelligence organizations—adoption is much slower than in commercial sectors.
Sector adoption velocityclaude-sonnet-52/5Defense and intelligence sectors are cautious adopters of AI due to security, accuracy, and accountability concerns; pilots exist but widespread production deployment for planning tasks is limited.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can substantially assist analysts by rapidly processing multi-source intelligence, generating scenario alternatives, identifying pattern correlations, and summarizing threat assessments—productivity gains are significant when human analysts remain in the loop to evaluate and decide.
Augmentation potentialclaude-sonnet-54/5AI tools significantly aid analysts by rapidly processing large datasets, flagging patterns, and drafting summaries, meaningfully speeding up plan development while humans retain final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can synthesize intelligence data and suggest tactical options at scale, developing coherent defense plans requires judgment about geopolitical context, adversary psychology, and organizational constraints that current systems cannot reliably handle end-to-end without substantial human oversight.
Task automatabilityclaude-sonnet-52/5This task requires synthesizing classified, ambiguous, and rapidly evolving information into strategic judgments and tactical plans with accountability for outcomes; current AI can support pieces (summarizing, pattern-finding) but cannot autonomously produce reliable defense plans end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Defense planning is typically subject to security clearance requirements, regulatory oversight, and legal/liability frameworks where human experts must take accountability; organizational culture and statutory requirements strongly favor human sign-off on strategic plans.
Adoption barriersclaude-sonnet-55/5Defense and intelligence work is subject to strict security clearance requirements, chain-of-command accountability, and legal/military authorization, making full automation highly restricted.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI inference costs for large-scale data synthesis are dropping, but integration with classified systems, validation workflows, and human expert oversight still exceed the loaded cost of experienced intelligence analysts for this task.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply process large data volumes, but the overall task still requires expensive human oversight, security clearances, and judgment, so total cost savings versus a human analyst are limited.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some AI products assist with intelligence synthesis and scenario modeling, but no deployed system reliably produces operationally sound defense plans without human experts reviewing, correcting, and taking responsibility for the output.
Technical feasibility todayclaude-sonnet-52/5Deployed AI tools exist for intelligence data fusion, geospatial analysis, and NLP summarization, but no production system autonomously generates defense plans or tactics; human analysts remain central to synthesis and decision-making.

Gather intelligence information by field observation, confidential information sources, or public records.

11

CI 3–20 · exposure 13 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While government agencies use AI for data processing and analysis, actual intelligence gathering in field and human-source domains remains largely manual and conservative. Adoption of automation in sensitive intelligence work is slow due to regulatory constraints, security protocols, and institutional risk aversion.
Sector adoption velocityclaude-sonnet-52/5Intelligence agencies are adopting AI for analysis and data processing, but the human-source and field-collection components remain largely untouched by automation trends.
Augmentation potentialclaude-haiku-4-5-202510013/5AI tools augment intelligence analysts by automating literature review, cross-referencing public records, and organizing large datasets, improving search and recall of information. However, augmentation is limited to preparatory and analytical phases; the critical gathering phase from human sources remains primarily human-driven.
Augmentation potentialclaude-sonnet-53/5AI can help analysts search public records, cross-reference open-source intelligence, and organize information gathered from other channels, improving efficiency around the human-led collection process.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with parsing public records and open-source data collection, field observation and cultivation of confidential human sources fundamentally require human presence, judgment, and trust-building that current AI cannot replicate. End-to-end automation fails at the core requirement of extracting sensitive intelligence from human sources.
Task automatabilityclaude-sonnet-51/5Field observation and cultivating confidential human sources are inherently physical, interpersonal, and trust-based activities that no AI system can perform end-to-end today.ract
Adoption barriersclaude-haiku-4-5-202510015/5Intelligence gathering is heavily regulated by government agencies (CIA, NSA, etc.) and requires personnel with security clearances and legal authorization to access classified sources. Legal liability, national security requirements, and statutory restrictions on who may conduct intelligence operations create hard barriers to substitution.
Adoption barriersclaude-sonnet-55/5Handling confidential sources and field intelligence gathering typically requires security clearances, legal authorization, and human judgment/trust that cannot be delegated to automated systems.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for data aggregation are relatively cheap, but they supplement rather than replace human intelligence analysts. The cost of trained personnel with security clearances and field capabilities far exceeds what AI automation can provide, and integration costs for intelligence workflows are substantial.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute for physical field presence or source handling, so the comparison to human cost is not meaningful in AI's favor.
Technical feasibility todayclaude-haiku-4-5-202510012/5Products exist for parsing public records and automating research workflows, but no deployed system reliably performs the full task of gathering intelligence from field observation or confidential sources at scale. The human-dependent elements of source cultivation and field validation remain beyond current AI capability in production.
Technical feasibility todayclaude-sonnet-51/5No deployed product conducts field surveillance or manages confidential human intelligence sources; AI is limited to processing records after collection, not the collection act itself.

Collaborate with representatives from other government and intelligence organizations to share information or coordinate intelligence activities.

7

CI 0–15 · exposure 8 · augmentation 50 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Intelligence agencies operate in highly regulated, classified environments with deeply entrenched human-centric workflows and legal requirements. Adoption of AI for intelligence collaboration is minimal; the sector is dominated by legal and security constraints rather than automation incentives.
Sector adoption velocityclaude-sonnet-52/5Intelligence and government sectors are slow, security-constrained adopters of AI, especially for tasks involving classified inter-agency liaison.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist with document organization, metadata tagging, summarization of intelligence reports, and scheduling—useful support functions. However, augmentation is limited to preparation and housekeeping; the actual collaboration and coordination remain the analyst's responsibility.
Augmentation potentialclaude-sonnet-53/5AI can help prepare briefing materials, summarize shared intelligence, or draft coordination communications, but the collaboration itself remains human-driven.
Task automatabilityclaude-haiku-4-5-202510012/5This task requires relationship-building, diplomatic negotiation, and real-time judgment calls across sensitive organizational boundaries. While AI could help draft communications or organize shared data, the core collaboration and coordination—especially around classified or sensitive intelligence—fundamentally requires human decision-making, trust, and accountability that current AI cannot authentically provide end-to-end.
Task automatabilityclaude-sonnet-51/5This task is fundamentally interpersonal coordination requiring trust-building, negotiation, and real-time judgment between agencies; AI cannot substitute for the relationship and authority dynamics involved.ed
Adoption barriersclaude-haiku-4-5-202510015/5Hard legal and regulatory barriers apply: intelligence sharing is governed by classification levels, inter-agency agreements, and statutory authority. A human intelligence officer must legally review, authorize, and be accountable for information shared across organizational boundaries. Liability and chain-of-custody requirements prevent full automation.
Adoption barriersclaude-sonnet-55/5Security clearances, legal authority to represent an agency, classified information handling, and inter-governmental protocols make this a hard human-only, credentialed function.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI assistance (e.g., summarization, document tagging) is cheaper than human overhead, but the irreplaceable human analyst who conducts the actual negotiation and coordination dominates the cost structure. AI cannot meaningfully reduce the loaded wage for the human who must attend meetings and sign off on agreements.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this coordination role, so cost comparison is moot; human cost is the only option.
Technical feasibility todayclaude-haiku-4-5-202510011/5No production system today reliably handles inter-agency intelligence collaboration at scale. The task involves classified information, legal authority, and organizational relationships that exceed the scope of deployed AI tools. Research projects may exist, but they are not in operational use by intelligence agencies for this purpose.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs inter-agency liaison or coordination; this remains a human diplomatic and organizational function.

Operate cameras, radios, or other surveillance equipment to intercept communications or document activities.

4

CI 0–7 · exposure 0 · augmentation 38 · importance 2.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Intelligence and law enforcement agencies operate in heavily regulated, security-controlled environments where automation of surveillance operations is severely constrained by legal authorization requirements and classified operational procedures.
Sector adoption velocityclaude-sonnet-52/5Intelligence and defense sectors are cautious adopters of full automation for sensitive collection activities, though AI tools are increasingly used for analysis rather than physical operation.
Augmentation potentialclaude-haiku-4-5-202510012/5AI might assist with post-hoc analysis of intercepted communications (e.g., transcription, pattern detection, translation) or alert analysts to flag activities meeting specified criteria, but the actual operation and real-time decision-making of surveillance equipment remains human-centric.
Augmentation potentialclaude-sonnet-53/5AI can assist with signal processing, transcription, and pattern detection from collected data, improving analyst efficiency, but the physical operation of equipment itself sees little AI augmentation.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time remote operation of physical equipment and situational awareness to identify and intercept communications or activities. Current AI systems cannot independently operate equipment, position cameras, adjust frequencies, or make judgment calls about what to intercept without explicit human direction.
Task automatabilityclaude-sonnet-51/5This is a hands-on physical operation task requiring positioning, judgment, and real-time adaptation to field conditions that current AI cannot perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Operating surveillance equipment and intercepting communications is highly regulated and legally restricted; only authorized personnel with security clearances can perform these tasks. Regulatory and legal barriers fundamentally protect this work from substitution.
Adoption barriersclaude-sonnet-54/5Intelligence collection is heavily regulated (legal authorization, chain of custody, classification), and operating interception equipment typically requires security clearance and legal authority, creating strong institutional and legal barriers.
Cost vs. human wageclaude-haiku-4-5-202510011/5The task demands specialized surveillance equipment, licensed operators, and continuous human monitoring. AI cannot replace the capital and labor costs of operating classified or restricted surveillance systems that intelligence analysts employ.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the physical equipment operation itself, so there is no meaningful cost comparison; a human operator is still required.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product autonomously operates surveillance equipment to intercept communications or monitor activities. This requires legal authorization, physical equipment control, and judgment about operational relevance that remains entirely human-directed in practice.
Technical feasibility todayclaude-sonnet-51/5No deployed product autonomously operates physical surveillance equipment (cameras, radios) in the field to intercept communications; this remains a human-operated activity with AI only assisting downstream analysis.

Interview, interrogate, or interact with witnesses or crime suspects to collect human intelligence.

3

CI 0–5 · exposure 5 · augmentation 50 · importance 3.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Law enforcement and intelligence agencies remain heavily human-dependent for interview and interrogation work; adoption of AI for autonomous interrogation is essentially zero in production. Regulatory constraints and institutional risk aversion keep this task firmly in human hands.
Sector adoption velocityclaude-sonnet-51/5Law enforcement and intelligence agencies are slow, highly regulated adopters of AI for direct human-source interaction, with essentially no production deployment of AI-led interrogation.
Augmentation potentialclaude-haiku-4-5-202510013/5AI tools can assist human interrogators by suggesting follow-up questions, flagging statement inconsistencies in real time, or analyzing linguistic patterns in transcripts post-interview. However, the core task of building rapport, assessing veracity, and conducting the live interaction remains human-led.
Augmentation potentialclaude-sonnet-53/5AI can assist with prep (background research, question generation, translation, sentiment/deception cue analysis on recordings) but does not transform the core interactive task itself.
Task automatabilityclaude-haiku-4-5-202510011/5Interviewing and interrogating witnesses or suspects requires real-time adaptive interaction, reading subtle social cues, building rapport, and assessing credibility—dimensions that current AI cannot reliably perform end-to-end. While AI can draft questions or analyze interview transcripts post-hoc, it cannot conduct the live interaction itself with sufficient reliability.
Task automatabilityclaude-sonnet-51/5Interviewing and interrogating suspects requires real-time human judgment, rapport-building, legal authority, and adaptive psychological tactics that current AI cannot perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Hard legal and procedural barriers protect this task: interrogations are heavily regulated under law (Miranda rights, admissibility rules, cruel and unusual punishment); suspects have the right to counsel; evidentiary standards require human chain-of-custody and credible testimony; many jurisdictions require licensed or trained human interrogators. An AI system cannot legally sign off on the reliability of human intelligence gathered through interrogation.
Adoption barriersclaude-sonnet-55/5Interrogations involve legal authority, chain-of-custody, admissibility rules, and constitutional protections (e.g., Miranda rights) that require a credentialed human officer or agent to conduct and be accountable for.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of an AI system to attempt this task (development, integration, oversight) far exceeds the loaded wage of trained human intelligence analysts, particularly when error costs (false confessions, missed deception, damaged cases) are factored in.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this task, so cost comparison favors humans entirely; any attempted AI system would require extensive human oversight negating savings.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably conducts live interviews or interrogations with human subjects. Research prototypes exist for question generation and statement analysis, but these are not in production use for actual intelligence gathering. Significant human oversight and judgment remain mandatory.
Technical feasibility todayclaude-sonnet-51/5No deployed product conducts autonomous interrogations or witness interviews for intelligence collection; this remains firmly research-stage or nonexistent for this use case.

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.