Loss Prevention Managers
11-9199.08Median wage $141,900/yr622,190 employed (US)Rank #382 of 923 scored · top 41% by substitution
Plan and direct policies, procedures, or systems to prevent the loss of assets. Determine risk exposure or potential liability, and develop risk control measures.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution30
Exposure27
Augmentation62
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
27 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
7%
Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.
Why this score
The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.
Task automatabilityw 35%28
panel mean rating 2.1/5 → substitution pressure 28/100
Technical feasibility todayw 20%26
panel mean rating 2.1/5 → substitution pressure 26/100
Cost vs. human wagew 15%31
panel mean rating 2.2/5 → substitution pressure 31/100
Adoption barriersw 20%inverted — strong barriers lower the score38
panel mean rating 3.5/5 (barrier strength) → substitution pressure 38/100
Sector adoption velocityw 10%31
panel mean rating 2.2/5 → substitution pressure 31/100
Task breakdown (27 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.
Maintain databases such as bad check logs, reports on multiple offenders, and alarm activation lists.
80CI 67–92 · exposure 83 · augmentation 75 · importance 2.8/5 · click for rater detail
Maintain databases such as bad check logs, reports on multiple offenders, and alarm activation lists.
80| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Loss prevention and retail security are relatively digitized sectors with strong cost-control incentives, and database automation is already commonplace in production deployments across major retailers and financial institutions. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and security sectors have moderate digitization with loss-prevention software increasingly used, but many smaller retailers still rely on manual spreadsheets or legacy systems, keeping adoption uneven. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can assist managers by auto-populating database fields, flagging anomalies, and generating summary reports, substantially raising productivity even if humans retain oversight of flagging and verification decisions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI and automated database tools can significantly speed up data entry, flagging of repeat offenders, and alarm log analysis, letting loss prevention managers focus on investigation and response rather than manual record-keeping. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Database maintenance for structured logs, reports, and lists is a purely mechanical task involving data entry, organization, and updating—ideal for automation. Current AI systems can extract, organize, and maintain these databases with minimal human intervention, easily achieving >50% time savings. |
| Task automatability | claude-sonnet-5 | 4/5 | Database maintenance, log updates, and structured record-keeping (bad checks, repeat offenders, alarm activations) are highly structured data-entry and organization tasks well suited to automation via scripts, RPA, and AI-assisted data tools with human spot-checks. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While data security and compliance (e.g., PCI-DSS, fraud reporting regulations) introduce some oversight requirements, there are no legal mandates requiring a human to personally maintain these databases, creating low adoption barriers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement mandates a human perform database maintenance, though data accuracy, privacy of offender records, and internal audit/compliance processes create some organizational friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automating routine database maintenance costs a fraction of human labor; inference and integration overhead are minimal, making AI orders of magnitude cheaper than paying a manager or clerk to perform these clerical updates. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated data logging and database updates via software/RPA cost far less per record than a manager's time, though initial system setup and integration add some cost before per-task efficiency is realized. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Deployed products (RPA, database management systems, and AI-powered data entry tools) reliably perform this task at scale in production environments across retail, banking, and security organizations today. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed database management, RPA, and CRM/loss-prevention software exist and are used in retail/security operations, but full automation of cross-referencing multiple offender lists and alarm data still requires integration work and some manual oversight in most current deployments. |
Analyze retail data to identify current or emerging trends in theft or fraud.
75CI 75–75 · exposure 75 · augmentation 100 · importance 3.8/5 · click for rater detail
Analyze retail data to identify current or emerging trends in theft or fraud.
75| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large and mid-size retail chains have widely adopted automated fraud and theft detection systems; this is a mature sector with strong digitization and financial ROI pressure driving adoption. Smaller retailers lag, but major players (grocery, big-box, specialty) use these tools in production. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Retail and loss prevention functions have adopted data analytics and fraud detection tools rapidly, especially with mature exception-reporting and POS analytics markets. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments loss prevention managers by flagging patterns humans would miss in large datasets, prioritizing investigation effort, and enabling real-time alerts. Managers retain judgment on severity, investigation direction, and legal action, making this a high-value assistive pairing. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-driven analytics substantially enhance a manager's ability to detect subtle or emerging fraud patterns while the manager retains judgment over investigation and action. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI systems can automatically detect statistical anomalies, flag suspicious patterns, and identify emerging trends in structured retail data (sales, inventory, transaction logs) with minimal human setup. However, contextual interpretation and investigative follow-up typically require human judgment, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 4/5 | Data analysis for anomaly detection and trend identification in transactional/loss data is well-suited to ML/statistical tools that can process large volumes of retail data far faster than manual review.AI can flag patterns humans would take hours to find. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Loss prevention managers are not licensed professionals, and no legal requirement mandates human sign-off on trend analysis itself. However, organizational risk aversion, need for human investigative follow-up, and skepticism of algorithmic recommendations provide modest friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for data analysis itself, though decisions based on flagged fraud may require human review before action, creating minor organizational friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Once implemented, AI inference and alerting on retail data cost a fraction of the human analyst hours needed to manually scan transactions and inventory records. Integration and oversight add cost, but the ratio still strongly favors automation at scale. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated analytics scale across millions of transactions at a fraction of the cost of manual analyst hours, though software licensing and data infrastructure costs remain nontrivial. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed analytics and fraud-detection platforms (e.g., Palantir, Sensormatic, vendor-specific BI tools) reliably perform pattern detection and anomaly flagging in production retail environments. Some false positives and domain-specific tuning remain necessary, but the core task is mature and widely deployed. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Retail loss prevention analytics platforms (e.g., exception-based reporting, POS analytics, fraud detection systems) are widely deployed in production at major retailers today, though they still require human interpretation and tuning. |
Review loss prevention exception reports and cash discrepancies to ensure adherence to guidelines.
63CI 59–67 · exposure 58 · augmentation 88 · importance 4.3/5 · click for rater detail
Review loss prevention exception reports and cash discrepancies to ensure adherence to guidelines.
63| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Retail and hospitality sectors are piloting AI-driven exception detection and anomaly flagging, but most deployments remain in the assisted-review stage rather than autonomous processing. Adoption is advancing faster in large chains with digitized POS systems than in smaller operations, suggesting middling overall velocity. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and loss prevention functions have adopted exception-reporting and POS analytics tools for years, but full automation of judgment calls is still emerging with pilots more common than full deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at surfacing priority exceptions, highlighting patterns, and generating summary reports that accelerate a manager's review cycle. Tools that rank exceptions by risk or auto-populate context from transaction logs measurably improve analyst productivity while keeping human judgment in the loop for final determination. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-driven exception detection and anomaly scoring significantly speeds up a manager's ability to prioritize which discrepancies need investigation, greatly boosting productivity while human judgment remains central. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can reliably parse structured exception reports and cash discrepancy data, flag anomalies against defined guidelines, and produce exception summaries with high accuracy. However, final judgment on policy violations and corrective actions often requires nuanced contextual knowledge, preventing full end-to-end automation without human review, though time savings of 50%+ are achievable in initial triage and report generation. |
| Task automatability | claude-sonnet-5 | 3/5 | Reviewing exception reports and flagging cash discrepancies against rules is a pattern-matching/data-analysis task that AI can largely automate, though final judgment on ambiguous cases and follow-up action still often needs human review.orks. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Loss prevention review is not typically gated by licensing or hard legal requirements for a human signature. However, audit trails, liability concerns (incorrect exception dismissal), and organizational preference for human accountability in fraud/theft investigations create moderate friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for this analytical task, though internal audit/compliance policies and accountability for loss decisions create some organizational friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference costs for report review are minimal (cents per report), and integration with existing exception-reporting systems is straightforward. The loaded cost of a loss prevention manager reviewing and investigating discrepancies is substantially higher, making AI cost per task-equivalent an order of magnitude lower once deployed. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated anomaly detection and rules-based flagging software runs at a fraction of the cost of a manager manually scanning reports, though some human oversight cost remains. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Document processing and anomaly detection products exist and perform well on structured financial data, but real-world loss prevention exception reports vary in format and require interpretation of subjective guideline compliance. Production deployments in retail and hospitality exist, but material error rates and the need for human override keep reliability below the 'mature at scale' threshold. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Retail loss-prevention analytics platforms (e.g., exception-based reporting tools with anomaly detection) exist and are used in production, but they typically flag anomalies for human review rather than fully closing the loop autonomously. |
Maintain documentation of all loss prevention activity.
53CI 39–67 · exposure 58 · augmentation 88 · importance 3.9/5 · click for rater detail
Maintain documentation of all loss prevention activity.
53| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Loss prevention remains a compliance-heavy, human-centric function in most organizations; adoption of AI documentation tools is limited to large enterprises with mature security infrastructure, while smaller retail and facility management firms lag significantly. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and security sectors are moderately digitized with growing use of case management software, but AI-specific automation of documentation is still in early-to-mid adoption stages. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at structuring, tagging, and cross-referencing loss prevention data; generating incident summaries; and flagging patterns—substantially boosting a manager's ability to organize and retrieve documentation while they retain authority over content completeness and legal sufficiency. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI can significantly speed up drafting, organizing, and summarizing loss prevention records while the manager retains responsibility for accuracy and final review. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can capture, categorize, and structure loss prevention logs from incident reports, security footage metadata, and investigation summaries with moderate accuracy. However, the task requires judgment about what constitutes complete documentation, context-dependent entry decisions, and compliance-specific formatting that still demand significant human oversight. |
| Task automatability | claude-sonnet-5 | 4/5 | Documentation and record-keeping of incident logs, reports, and activity summaries is largely text-based and structured, making it highly amenable to AI-assisted drafting, summarization, and organization with substantial time savings. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Loss prevention documentation is heavily regulated by retail, financial, and workplace safety frameworks; liability exposure for incomplete or inaccurate records is asymmetric (omissions create legal risk); and organizational risk tolerance is low for fully automated record-keeping without human sign-off. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement mandates a human to personally maintain these records, though internal audit/compliance policies and accuracy-liability concerns create some friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Integration costs for documentation systems, ongoing human review to ensure compliance accuracy, and liability management overhead mean AI-assisted logging remains moderately more expensive than a human documentation specialist at smaller scale. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated documentation tools and AI drafting assistants are far cheaper per report than manager time spent manually compiling logs, though some oversight cost remains. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Products exist for automated incident logging, document management systems with AI categorization, and security event aggregation, but real-world systems often require manual verification, human review of sensitive content, and integration with legacy compliance systems—limiting fully autonomous performance. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Case management and reporting software with AI-assisted templating exists and is used in retail/security operations, but full autonomous documentation with accurate incident details still requires human input and verification. |
Assess security needs across locations to ensure proper deployment of loss prevention resources, such as staff and technology.
53CI 30–76 · exposure 58 · augmentation 88 · importance 3.9/5 · click for rater detail
Assess security needs across locations to ensure proper deployment of loss prevention resources, such as staff and technology.
53| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Retail and large logistics sectors show pilot deployments and growing adoption of predictive loss tools, but full automation of resource deployment decisions remains limited. Production adoption exists but is not yet industry-standard, with many organizations still in evaluation phases. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail and loss prevention functions are adopting analytics and camera-based AI tools gradually, but resource deployment decisions remain largely human-driven with slow, cautious adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI significantly amplifies human loss prevention managers by rapidly synthesizing multi-location data, flag-ging patterns, and generating ranked deployment scenarios that managers review and refine. This keeps humans in high-level decision-making while dramatically reducing data-gathering and scenario-modeling time. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered analytics, predictive risk scoring, and video analytics can meaningfully inform and speed up the assessment process, helping managers prioritize locations and resources more efficiently. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | AI can analyze historical loss data, facility layouts, inventory patterns, and security metrics across locations to generate comprehensive resource deployment recommendations with significant time savings. Modern AI systems can ingest location-specific data and produce risk-ranked deployment plans that meet the ≥50% time-saving threshold compared to manual assessment. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires site-specific judgment, physical risk assessment, and resource allocation decisions that depend on contextual knowledge AI cannot fully gather or weigh today.for full autonomy is not achievable end-to-end.rating reflects partial support only. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While no licensing explicitly forbids AI assessment, organizational liability concerns, corporate governance review requirements, and stakeholder preference for human verification create meaningful friction. Risk management departments often require human sign-off on resource deployment decisions despite AI analysis. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement, but liability for security failures, physical site variability, and organizational trust in human judgment create meaningful friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven security assessment platforms cost a fraction of hiring full-time loss prevention analysts across multiple locations; inference and ongoing analysis are substantially cheaper than the loaded wages of human staff doing equivalent multi-site assessments. Savings compound with scale across many locations. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Human loss prevention managers combine judgment, negotiation, and on-site evaluation; AI tools require significant human oversight and data integration, keeping costs comparable rather than dramatically lower. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (security analytics platforms, predictive loss modeling, and geospatial risk assessment tools) perform parts of this task reliably in production, though human review of final deployment decisions remains common. The core analytical work is production-ready; integration into full workflow varies by organization. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some analytics tools help flag theft patterns or high-risk locations, but no deployed product independently assesses security needs and allocates staff/technology across sites reliably. |
Monitor and review paperwork procedures and systems to prevent error-related shortages.
51CI 48–55 · exposure 50 · augmentation 75 · importance 3.3/5 · click for rater detail
Monitor and review paperwork procedures and systems to prevent error-related shortages.
51| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Loss prevention and logistics sectors show slow-to-moderate digitization; adoption of AI document review remains in pilot phases across most organizations, with mainstream production deployment lagging fintech and professional services. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and logistics sectors are adopting AI-based auditing and anomaly detection at a moderate pace, with pilots common but full replacement of oversight roles still rare. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at flagging anomalies, organizing paperwork by risk level, and identifying patterns that humans might miss, substantially raising a loss prevention manager's throughput and detection accuracy while they retain final judgment and decision-making. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-driven anomaly detection and document review tools significantly speed up identification of error patterns, letting loss prevention managers focus on investigation and corrective action. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can automate significant portions of paperwork review through document classification, pattern detection, and flagging anomalies, potentially saving 40-60% of time. However, judgment calls on systemic vs. isolated errors and determination of root causes often require human context and institutional knowledge. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can review documents/data for anomalies and flag discrepancies at scale, but designing and adapting monitoring procedures across varied paperwork systems still needs human judgment and contextual investigation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Internal control and audit requirements often mandate human sign-off on error-prevention decisions, and organizational inertia around legacy paperwork systems creates friction, though no strict licensing barrier exists for the task itself. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but internal controls, accountability for shortages, and audit trail integrity create moderate organizational friction against pure automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI document automation and monitoring tools are moderately priced, but integration, training, and oversight costs for error-prone financial/compliance work roughly match the wage of a skilled loss prevention analyst doing the monitoring. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated auditing software reduces some labor cost, but implementation, integration with legacy paperwork systems, and human oversight keep costs roughly comparable to a trained manager's time for judgment calls. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Document processing and anomaly detection products exist and see deployment, but they typically require substantial customization to organizational workflows and generate false positives that demand human review, limiting standalone reliability. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Audit and anomaly-detection tools (e.g., exception reporting, OCR-based document review) are deployed in retail loss prevention, but comprehensive procedure oversight is still largely human-managed. |
Train loss prevention staff, retail managers, or store employees on loss control and prevention measures.
42CI 34–50 · exposure 30 · augmentation 75 · importance 4.2/5 · click for rater detail
Train loss prevention staff, retail managers, or store employees on loss control and prevention measures.
42| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Retail and loss prevention organizations are adopting AI-assisted training tools (content generation, LMS automation) at a moderate pace, with pilots common but full autonomous delivery still rare. Digitization in retail is advancing, but legacy staffing models and preference for in-person training in security-critical roles slows deep penetration. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail is a moderate-to-low digitization sector for this specific HR/training function; AI adoption for compliance and procedural training is emerging but not yet widespread or deep. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments trainers by generating personalized content, automating scheduling and assessment, and creating interactive simulations. Loss prevention managers can delegate content creation and routine delivery while focusing on live demonstrations, credibility, and individual mentorship—raising overall training throughput and consistency. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly help loss prevention managers create training materials, quizzes, scenario-based content, and track completion, meaningfully boosting productivity while the manager still delivers and oversees training. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Training delivery involves significant human interaction, feedback loops, and contextual judgment tailored to staff experience levels. While AI can generate training content and draft materials, end-to-end delivery with 50% time savings at equal quality requires live adaptation to trainee questions and organizational culture that current systems cannot reliably replace. |
| Task automatability | claude-sonnet-5 | 2/5 | Training delivery involves live instruction, hands-on demonstration, and adaptive response to trainee questions and store-specific contexts that current AI cannot fully replicate end-to-end..The content generation portion (materials, scripts) is automatable, but delivery and assessment of practical skills is not. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Organizations often prefer live trainer presence for accountability, certification compliance, and brand assurance; some regulatory frameworks may require documented human instruction for loss prevention competency. However, no hard legal barrier prevents AI-assisted or fully automated training delivery in most jurisdictions, creating moderate but not insurmountable friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement mandates a human trainer, but organizational preference for manager-led training, need for contextual judgment, and liability concerns around theft/safety procedures create some friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-generated training content (video, slides, quizzes) costs far less per employee than live trainers once amortized across cohorts. Inference and integration costs for LMS-based training are orders of magnitude below the loaded cost of hiring and scheduling professional trainers or pulling managers away from floor duties. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI can cheaply produce training materials and modules, but human-led training sessions, especially interactive or hands-on components, still require paid staff time, making overall cost savings moderate rather than dramatic. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI can draft training curricula, generate video scripts, and create assessment materials at scale, with products like learning management system integrations now available. However, live instructor roles and dynamic response to trainee needs remain largely human-dependent; deployed systems lack the real-time adaptability and credibility validation needed for production training at organizational scale. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI-generated training content and e-learning modules exist and are used in retail, but full replacement of manager-led training on loss prevention procedures is not standard in deployed products. |
Perform cash audits and deposit investigations to fully account for store cash.
40CI 25–55 · exposure 38 · augmentation 63 · importance 3.6/5 · click for rater detail
Perform cash audits and deposit investigations to fully account for store cash.
40| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Retail and hospitality are moderately digitized, and some chains use audit software, but adoption of AI-driven loss prevention investigation remains in pilot stages. Physical cash handling and investigation require on-site presence and judgment, limiting adoption velocity. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail sector has moderate digitization with growing use of AI-based loss prevention analytics, but adoption is uneven and often limited to large chains. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by flagging anomalies, auto-matching transactions, and highlighting discrepancies, which reduces the time managers spend on initial data review. However, the core investigative and decision-making work remains firmly with the human manager. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-driven anomaly detection and automated reconciliation significantly speed up identifying discrepancies, letting managers focus on investigating flagged cases rather than manual cross-checking. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Cash audits and deposit investigations require physical counting, reconciliation across multiple ledgers, and judgment about discrepancies. While basic arithmetic and record matching can be automated, the investigative component—identifying theft, human error, or procedural breakdowns—demands human reasoning and context that current AI cannot reliably perform end-to-end. |
| Task automatability | claude-sonnet-5 | 3/5 | AI/software can reconcile transaction logs, POS data, and bank deposits automatically, flagging discrepancies, but final investigation of root cause and interviewing staff still requires human judgment.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Cash handling and audit sign-off typically fall under internal control and compliance frameworks (SOX, PCI-DSS, company policy) that require a licensed manager or supervisor to verify and certify results. Regulatory and liability requirements create strong legal barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but internal controls, audit trail integrity, and potential fraud/legal implications create moderate organizational and compliance friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for audit assistance are moderately priced, but the need for trained loss prevention staff to interpret findings, conduct investigations, and document conclusions means total labor cost savings remain modest compared to the loaded wage of a manager performing the full task. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated reconciliation tools reduce labor hours substantially, but licensing, integration, and human oversight for investigations keep costs roughly comparable to a lean human process at many retailers. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some software exists for transaction matching and variance flagging, but deployed systems typically require human review and decision-making for all substantive findings. No mature product performs complete cash audit investigations autonomously without significant human oversight. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Retail loss-prevention and POS reconciliation software exist and are deployed, but full end-to-end automated deposit investigation with exception resolution is not yet standard practice. |
Recommend improvements in loss prevention programs, staffing, scheduling, or training.
31CI 29–34 · exposure 25 · augmentation 63 · importance 3.5/5 · click for rater detail
Recommend improvements in loss prevention programs, staffing, scheduling, or training.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Loss prevention remains a compliance-focused, risk-averse function in most organizations with limited digital transformation. Adoption of AI-generated recommendations has been slow relative to other business functions, with most firms still relying on traditional consultants or internal experts. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Loss prevention and retail security functions are physical, operationally embedded sectors with slower AI adoption compared to pure information/professional services, though analytics tools are gradually being piloted. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist loss prevention managers by analyzing incident patterns, benchmarking against industry data, and drafting recommendation templates, allowing managers to focus on contextual refinement and stakeholder engagement rather than raw data synthesis. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by analyzing shrinkage data, incident patterns, and scheduling data to surface insights and draft recommendations, significantly speeding up the manager's analysis and reporting work. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can analyze loss data and generate templated recommendations, but crafting contextual improvements for organizational programs requires understanding organizational culture, budget constraints, and human factors that AI struggles to assess end-to-end. The task involves strategic judgment beyond data aggregation. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires synthesizing organizational context, security incident history, staffing constraints, and human judgment about workplace culture to formulate actionable recommendations, which current AI cannot fully replicate end-to-end. AI can draft supporting analysis but not independently produce trustworthy final recommendations. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Recommendations on staffing, scheduling, and training often require accountability for decisions affecting employee safety and company liability, creating organizational pressure for human judgment and sign-off. Regulatory compliance in some sectors further elevates the need for qualified personnel to own these recommendations. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement blocks AI involvement, but organizational trust, liability concerns around staffing/security decisions, and need for managerial accountability create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI could potentially reduce analysis labor and generate initial recommendation drafts at lower cost than hiring consultants, but the need for expert oversight and customization keeps total cost roughly comparable to a loss prevention specialist's time. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted analysis can reduce time spent gathering and summarizing data, but a human manager still must review, contextualize, and finalize recommendations, keeping costs roughly comparable to fully human-driven process today. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI can draft recommendations from historical loss data and industry benchmarks, no production system reliably generates implementable improvement recommendations for complex organizational contexts without substantial human review and domain expertise. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Analytics products exist that surface loss trends and staffing gaps, but no deployed product autonomously generates comprehensive program-improvement recommendations accepted without heavy human vetting. Current tools are decision-support, not decision-making. |
Advise retail managers on compliance with applicable codes, laws, regulations, or standards.
29CI 25–34 · exposure 25 · augmentation 63 · importance 3.9/5 · click for rater detail
Advise retail managers on compliance with applicable codes, laws, regulations, or standards.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Retail is moderately digitized but slow to adopt AI for high-risk functions like compliance guidance. Most adoption remains in data analytics and inventory; compliance advice remains a conservative, human-led function with limited measured displacement of the advisory role itself. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail loss prevention is a moderately digitized function with growing use of AI for research support, but formal advisory workflows around compliance remain slow to adopt AI compared to finance or professional services. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist loss prevention managers by retrieving regulations, flagging regulatory changes, drafting initial compliance summaries, and suggesting risk areas. The human manager then refines and owns the advice, improving productivity on research and drafting phases while maintaining decision authority. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can efficiently pull relevant statutes, summarize regulatory changes, and draft advisory language, substantially speeding up a manager's ability to research and communicate compliance guidance. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can retrieve and summarize regulatory information, advising on compliance requires contextual judgment about specific retail operations, risk profiles, and ambiguous regulatory edges that still need human interpretation. Current systems cannot reliably deliver end-to-end compliance advice without significant human review and modification. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can retrieve and summarize relevant codes and regulations, but tailoring compliance advice to specific retail contexts, store layouts, and risk scenarios still requires human judgment and accountability that current systems cannot fully replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Compliance advice carries legal and liability risk; regulatory bodies and courts typically expect human professionals to be accountable for guidance. Many jurisdictions expect a qualified human (loss prevention manager, attorney) to own and sign compliance recommendations, creating a strong legal barrier to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | While no license is strictly required to advise on compliance in most retail contexts, liability exposure and the expectation of accountable human judgment create moderate friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for compliance assistance (legal research platforms, ChatGPT prompting) are relatively cheap, but they still require loss prevention or legal staff to verify, contextualize, and sign off on advice. The all-in cost remains comparable to or higher than direct human advice when quality assurance is included. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted research can cut time spent gathering regulatory information significantly, but the need for expert validation and liability oversight keeps blended costs closer to comparable rather than order-of-magnitude cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No mature production system reliably advises on compliance across diverse retail codes without human legal oversight. AI can draft summaries and flag issues, but deployed products in this domain show material gaps in accuracy and liability coverage; legal counsel typically remain in the loop. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Legal/compliance research assistants and generic LLM advisory tools exist, but no mature product reliably delivers loss-prevention-specific regulatory advice in production without significant human review. |
Administer systems and programs to reduce loss, maintain inventory control, or increase safety.
29CI 25–32 · exposure 25 · augmentation 63 · importance 4.3/5 · click for rater detail
Administer systems and programs to reduce loss, maintain inventory control, or increase safety.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Retail and logistics sectors are adopting AI surveillance and inventory tools incrementally, but deployment remains fragmented; true end-to-end administrative automation of loss prevention programs is rare and mostly piloted, not in widespread production. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and logistics sectors are adopting AI-driven loss prevention tools (video analytics, RFID, predictive shrink models) at a moderate pace, with pilots widespread but full programmatic administration still human-led. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist loss prevention managers by automating inventory audits, flagging suspicious patterns in transaction data, and generating real-time alerts from surveillance footage, meaningfully boosting their efficiency on data-heavy tasks while they retain decision-making and program direction. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially augments this task through predictive analytics, automated anomaly flagging, and inventory forecasting, letting loss prevention managers focus oversight and strategic decisions on higher-value judgment calls. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with monitoring systems and flagging anomalies in inventory or surveillance data, administering comprehensive loss prevention programs requires ongoing human judgment, policy decisions, relationship management with staff, and context-specific risk assessment that current AI cannot fully replace. |
| Task automatability | claude-sonnet-5 | 2/5 | This is a broad managerial task involving program design, policy administration, cross-functional coordination, and decision-making that current AI cannot fully replicate end-to-end, though analytics components (inventory tracking, anomaly detection) are automatable. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Loss prevention administration often involves legal liability (security decisions, employee discipline, regulatory compliance with theft/safety laws), executive sign-off, and direct responsibility for organizational risk; human accountability and decision-making authority remain legally and organizationally required. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement typically exists, but liability for safety programs, security decisions, and legal compliance creates meaningful organizational friction against fully automating program administration. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI solutions for loss prevention (surveillance systems, inventory software, monitoring) require significant setup, human oversight, and integration costs that approach or exceed the loaded cost of a loss prevention manager's salary, especially for comprehensive administration. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While individual analytics tools are cheap, the full administrative and managerial scope still requires human oversight, policy-setting, and cross-departmental coordination, keeping all-in AI substitution costs comparable to or above a manager's cost for the whole task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI tools exist for inventory tracking, anomaly detection, and incident logging, but no integrated system reliably administers end-to-end loss prevention programs autonomously; deployed products handle narrow subtasks (e.g., video anomaly detection) with material error rates and limited organizational integration. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products exist for inventory analytics, video-based theft detection, and shrink analysis, but no deployed system administers a comprehensive loss-prevention/safety program autonomously; human managers still integrate and oversee these tools. |
Perform or direct inventory investigations in response to shrink results outside of acceptable ranges.
29CI 25–32 · exposure 25 · augmentation 63 · importance 3.9/5 · click for rater detail
Perform or direct inventory investigations in response to shrink results outside of acceptable ranges.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Retail and warehouse sectors have adopted inventory analytics tools slowly, with most organizations still relying on human-directed investigations. Digitization of shrink analysis is uneven, and regulatory/liability concerns limit aggressive automation in employee-facing investigative contexts. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and loss prevention functions have adopted AI-driven analytics (video analytics, POS exception reporting) at a moderate pace, but full investigative automation is still uncommon. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully augment by flagging shrink anomalies, clustering affected inventory areas, suggesting common root causes, and generating preliminary analysis reports that a loss prevention manager reviews before directing investigation. This assistance can accelerate investigation planning without replacing human direction. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly aid investigators by flagging shrink outliers, analyzing transaction data, and reviewing video footage, substantially speeding up the investigative process while humans direct and conclude the investigation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in analyzing shrink data and flagging anomalies, directing investigations requires judgment about root causes (theft, breakage, administrative error), discretion in questioning employees, and contextual decision-making that current systems cannot reliably perform end-to-end. Some data analysis could be automated, but the investigative direction and human judgment remain essential. |
| Task automatability | claude-sonnet-5 | 2/5 | The task requires physical site investigation, interviewing staff, reviewing footage, and directing corrective action—only data analysis portions (identifying anomalous shrink patterns) are AI-amenable, not the full investigative and managerial process. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: loss prevention investigations often involve employee conduct, potential legal/HR consequences, and liability if investigation conclusions are wrong. Most organizations require a licensed or qualified human manager to direct investigations and make determinations; liability exposure and employment law considerations create strong friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but investigations often carry HR, legal, and evidentiary implications requiring human judgment and accountability, creating moderate organizational and liability barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI systems for investigation support plus required human oversight would likely exceed or match the cost of a loss prevention manager performing the task directly, especially when factoring in integration, false-positive investigation costs, and the need for human judgment to validate findings. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply surface anomalies in POS/inventory data, but the investigation itself (interviews, site review, decision-making, coordination with HR/security) still requires paid human labor, keeping overall cost comparable to a human-led process. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs end-to-end inventory investigations. AI can help analyze shrink data and generate reports, but investigation direction—deciding which inventory areas to examine, interview strategies, and causality assessment—remains human-dependent in practice. Tools exist for data analysis but not for autonomous investigation management. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Retail analytics products can flag shrink anomalies and suspicious transactions, but no deployed system autonomously conducts or directs full loss investigations; that remains human-led. |
Identify potential for loss and develop strategies to eliminate it.
28CI 25–30 · exposure 25 · augmentation 63 · importance 4.0/5 · click for rater detail
Identify potential for loss and develop strategies to eliminate it.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI in loss prevention remains slow outside large retail and financial institutions; most mid-market and smaller organizations rely on manual processes or basic rule-based systems rather than deployed AI agents generating strategic recommendations. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Loss prevention sits within retail/security operations, a sector with moderate digitization but historically slower AI adoption for strategic roles compared to finance or professional services. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist managers by surfacing anomalies, organizing transaction patterns, and flagging risky behaviors, enabling faster investigation and data-driven strategy refinement while the manager retains decision authority and contextual interpretation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-driven analytics, anomaly detection, and predictive modeling substantially help loss prevention managers identify risk patterns and prioritize interventions, meaningfully boosting their productivity while they retain strategic control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Loss identification can be partially automated through data analysis and pattern detection (e.g., anomaly detection in transaction or inventory data), but developing comprehensive strategies requires contextual judgment, stakeholder engagement, and organization-specific knowledge that current AI systems cannot reliably execute end-to-end without substantial human oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires synthesizing organizational context, physical security assessment, and strategic judgment across varied risk domains, which current AI cannot do end-to-end despite being able to analyze data patterns that inform the process.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Loss prevention involves internal security decisions, fraud investigations, and potential legal/compliance actions, which are typically reserved for qualified human managers. Liability and reputational risk if AI-recommended strategies fail or discriminate creates strong organizational and fiduciary barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing mandate strictly requires a human, but liability for security failures, need for on-site judgment, and organizational trust in a manager's accountability create meaningful friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Loss prevention managers earn substantial salaries (often $60k–$100k+), and implementing AI systems with sufficient integration, validation, and human oversight typically costs more than the incremental savings on routine detection tasks, especially given the small scale of most loss-prevention teams. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Analytics tools reduce some data-crunching cost, but the strategic development and organizational buy-in components still require expensive human expertise, keeping overall cost comparable to or only modestly below human-only approaches. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While some AI tools exist for anomaly detection and fraud flagging, no deployed product reliably performs the full cycle of loss identification and strategy development. Most organizations rely on analysts interpreting AI outputs rather than AI-driven systems making independent strategic recommendations. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI analytics products exist for fraud/shrinkage pattern detection, but no deployed product independently identifies loss potential and develops comprehensive prevention strategies without a human strategist driving the process. |
Monitor compliance to operational, safety, or inventory control procedures, including physical security standards.
28CI 25–30 · exposure 25 · augmentation 63 · importance 3.8/5 · click for rater detail
Monitor compliance to operational, safety, or inventory control procedures, including physical security standards.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption is limited to supplementary analytics (alerts on inventory discrepancies, access-control logs) in digitized retail and logistics environments. Production deployment of autonomous compliance monitoring remains rare; most organizations retain human loss prevention managers for primary accountability and judgment. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail and physical security sectors have adopted AI video analytics and RFID-based inventory tools at a measured pace, but broad, deep automation of compliance monitoring functions remains limited to pilots in most organizations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by flagging anomalies in structured data streams (CCTV motion detection, inventory variance alerts, access-log irregularities), freeing managers to focus investigation and judgment on flagged cases. However, the core task of interpreting context and making compliance decisions remains human-led. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered video analytics, anomaly detection, and automated audit checklists meaningfully assist loss prevention managers by flagging issues faster and reducing manual review time, while humans still make final compliance decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Monitoring compliance requires contextual interpretation of complex procedures, judgment about exceptions, and integration of multiple data sources. While AI can flag anomalies in structured data (inventory counts, access logs), the nuanced assessment of procedural compliance and safety standards remains difficult without significant human judgment and domain expertise. |
| Task automatability | claude-sonnet-5 | 2/5 | Parts of compliance monitoring (video review, inventory anomaly detection, checklist tracking) can be AI-assisted, but synthesizing findings across physical security, operations, and inventory into actionable oversight still requires human judgment and site presence. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: liability for missed compliance violations, potential regulatory requirements for human sign-off on safety and security assessments, and organizational preference for human accountability in risk-sensitive environments. Legal and insurance frameworks typically require human loss prevention professionals to own compliance decisions. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement mandates a human specifically, but liability for security incidents, need for on-site judgment calls, and organizational trust in human oversight create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Integration of compliance monitoring AI into existing systems requires substantial setup, oversight, and human review to validate alerts. The all-in cost (infrastructure, false-positive review, liability risk management) remains comparable to or higher than the loaded cost of human monitoring, especially given the need for human judgment on exceptions. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI surveillance and inventory analytics tools reduce some labor but still require licensing, integration, and human oversight, so total cost is only modestly below a loss prevention manager's wage for the full scope of this task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some monitoring tools exist (CCTV analytics, inventory management alerts), but deployed systems typically detect only narrow anomalies rather than comprehensive compliance with operational and safety procedures. Production systems lack the contextual reasoning to evaluate compliance across the full scope of operational, safety, and inventory control standards. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Deployed video analytics and inventory-shrink detection tools exist and are used in retail loss prevention, but they cover narrow slices of the task and still require human managers to interpret and act on results. |
Verify correct use and maintenance of physical security systems, such as closed-circuit television, merchandise tags, and burglar alarms.
28CI 25–30 · exposure 25 · augmentation 50 · importance 3.8/5 · click for rater detail
Verify correct use and maintenance of physical security systems, such as closed-circuit television, merchandise tags, and burglar alarms.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption remains slow and partial; most retail and enterprise environments use humans for periodic physical security audits, with AI-assisted monitoring as a supplement. Few organizations have automated the verification and maintenance assessment function itself. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail and loss prevention sectors are adopting AI-based video analytics and anomaly detection but physical system verification remains largely manual and adoption of full verification automation is slow.' |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can augment a loss prevention manager by automating CCTV review, alerting to potential system faults, and logging system status—useful but limited. The manager still performs the critical verification and maintenance judgment, so productivity gain is moderate. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-enabled video analytics, IoT sensor dashboards, and automated alerts can flag anomalies or system faults, helping the manager prioritize checks and maintenance actions more efficiently.' |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can monitor CCTV feeds and flag anomalies, verifying 'correct use and maintenance' requires hands-on system testing, physical inspection of equipment condition, and judgment about compliance with standards—tasks that demand human presence and expertise. Current AI can assist with video monitoring but cannot autonomously verify maintenance logs or conduct physical checks at scale. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires physical inspection, hands-on testing of hardware, and site walk-throughs that current AI cannot perform end-to-end without robotic or human physical presence.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | High barriers exist: loss prevention verification often requires authorized personnel with security clearances, liability for failed system checks rests with accountable humans, and regulatory frameworks (retail security standards, insurance requirements) typically mandate documented human sign-off on system maintenance and functionality. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically, but organizational reliance on trusted personnel for security oversight and liability for security failures creates moderate friction against full automation.' |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | CCTV analytics and monitoring services are moderately priced, but the total cost of integrated AI oversight for all three systems (CCTV, tags, alarms) plus required human verification remains comparable to or higher than a loss prevention manager's periodic audits. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Physical verification still requires a human presence on-site; AI tools may supplement remote monitoring but do not replace the labor cost of physical checks, so overall cost savings are limited.' |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Video analytics products exist for CCTV monitoring, but no deployed system reliably performs the full verification task (checking alarm functionality, inspecting physical tags, testing system integration, documenting maintenance). Production systems handle detection only, not maintenance verification. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI video analytics products exist for monitoring CCTV footage, but verifying correct maintenance/installation of tags, alarms, and physical systems is not something deployed AI products do autonomously today.' |
Direct loss prevention audit programs including target store audits, maintenance audits, safety audits, or electronic article surveillance (EAS) audits.
28CI 25–30 · exposure 25 · augmentation 63 · importance 3.8/5 · click for rater detail
Direct loss prevention audit programs including target store audits, maintenance audits, safety audits, or electronic article surveillance (EAS) audits.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Retail and loss prevention are moderately digitized but adoption of AI for audit direction remains limited; most organizations use legacy audit management systems with light AI analytics layers. Production deployment of autonomous audit programs is rare, with pilots emerging in large chains but not widespread. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail loss prevention is a moderately digitized but operationally physical sector; AI adoption is growing in fraud/theft detection analytics but program-level audit direction remains largely manual. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist meaningfully by analyzing incident trends, flagging high-risk stores, auto-generating audit templates, and summarizing compliance data, enabling managers to focus on site visits and stakeholder work. This assists productivity without replacing the human director's core judgment and accountability role. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-driven analytics dashboards, anomaly detection, and reporting tools meaningfully help loss prevention managers prioritize audits and identify risk patterns, improving efficiency while humans still direct the program. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with audit scheduling, data analysis, and report generation, the task requires site-specific inspection (walking stores, assessing maintenance), judgment about safety hazards, and stakeholder interviews that demand human presence and expertise. Current AI cannot perform on-site audits or make the nuanced risk assessments that define loss prevention work. |
| Task automatability | claude-sonnet-5 | 2/5 | Directing an audit program requires setting priorities, managing staff, interpreting site-specific risk, and making judgment calls that current AI cannot fully replace end-to-end; AI can support data analysis but not the overall direction and management function. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: retailers face liability for audit findings (missed safety hazards), regulatory compliance (OSHA, fire codes) requires documented human accountability, and audit direction involves stakeholder communication and organizational judgment that creates legal/reputational risk if delegated entirely to automation. Human sign-off is effectively required. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement mandates a human, but safety audits often involve compliance/liability considerations and organizational trust that favor human accountability for sign-off and enforcement decisions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce audit costs via analytics and scheduling optimization, but the human loss prevention manager's salary is modest relative to the value of site-based inspection work; full automation would require expensive robots or field agents, making the all-in cost higher than current human audit direction. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Software can cut some analysis time cheaply, but the managerial oversight, physical site visits, and coordination still require paid human labor, keeping overall cost comparable to human-led programs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product performs full audit direction end-to-end; AI tools exist for audit data management and risk scoring, but oversight, site audits, and program direction remain human-driven. Some retailers use AI for incident flagging and pattern analysis, but not for directing audit programs themselves. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Analytics tools exist for flagging shrink anomalies or EAS alert patterns, but no deployed product manages or directs a full multi-site audit program including maintenance and safety audits reliably today. |
Supervise surveillance, detection, or criminal processing related to theft and criminal cases.
28CI 28–28 · exposure 25 · augmentation 75 · importance 3.6/5 · click for rater detail
Supervise surveillance, detection, or criminal processing related to theft and criminal cases.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Retail and some enterprise security teams have adopted AI-assisted video surveillance and anomaly detection at pilot or early production stages, but criminal case processing automation lags due to regulatory and liability concerns. Overall adoption remains in the middling range with pockets of integration. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and loss prevention sectors have moderately adopted AI-based video analytics and anomaly detection, but broader case management and supervisory functions still see limited AI penetration. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI surveillance detection systems significantly assist loss prevention managers by automatically flagging suspicious activity, reducing manual video review burden and accelerating incident identification. This assistive capability materially improves productivity while the manager retains judgment on criminal processing and case decisions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI video analytics, pattern detection, and automated alerts significantly enhance a loss prevention manager's ability to identify theft and prioritize cases, even though the manager retains ultimate oversight. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with video surveillance detection and flagging anomalies, the task fundamentally requires human judgment in criminal processing, evidence evaluation, and decision-making on law enforcement coordination. Supervision and criminal case handling involve legal and ethical complexities that prevent autonomous end-to-end automation meeting the 50% time-savings threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | Supervision of surveillance and criminal processing requires human judgment, coordination with law enforcement, and accountability that current AI cannot fully replace, though AI can assist with detection alerts and video analytics.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Criminal case processing, evidence handling, and police liaison typically require a licensed or authorized human with legal responsibility. Chain-of-custody documentation, testimony preparation, and legal liability create substantial barriers to full automation of the supervisory and criminal case components. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Criminal processing involves legal chain-of-custody, evidentiary standards, and liability concerns that generally require a human manager's authorization and accountability, creating strong organizational and legal barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI video surveillance tools have deployment costs, but loss prevention managers earn professional wages and their criminal processing, investigation coordination, and legal responsibilities cannot be fully offloaded. The all-in cost of AI plus required human oversight approaches or exceeds the loaded wage of the manager. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI surveillance tools reduce some monitoring costs, but the managerial oversight, legal coordination, and case handling still require paid human labor, keeping overall costs comparable rather than dramatically lower. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Video analysis tools exist for detecting theft in surveillance footage, but real-world criminal processing requires human judgment, legal compliance, and chain-of-custody procedures. No deployed system reliably handles the full task—criminal processing especially remains a human-driven function with regulatory and liability constraints. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI-powered video analytics and anomaly detection products exist and are deployed in retail loss prevention, but managing the full supervisory and case-processing workflow remains human-led with narrow AI tool support. |
Advise retail establishments on development of loss-investigation procedures.
28CI 25–30 · exposure 25 · augmentation 63 · importance 3.1/5 · click for rater detail
Advise retail establishments on development of loss-investigation procedures.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Loss prevention is concentrated in mid-to-large retail chains and remains a traditionally human-staffed function with deep organizational ties. While digital tools are used, the advisory and procedural design work has not seen deep AI-agent adoption in production; most sectors in this space are still in pilot or early-adoption phases. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail loss prevention is a moderately digitized but operationally traditional sector where AI adoption for advisory/strategic tasks remains in early pilot stages rather than widespread production use. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist a loss-prevention manager by drafting procedure templates, summarizing regulatory requirements, and identifying best-practice patterns from case data. However, the human must review, contextualize, and sign off on recommendations, so augmentation is useful but not transformative to the core advisory task. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by drafting procedure templates, summarizing best practices, analyzing loss data patterns, and generating investigation checklists, significantly speeding up the manager's advisory work. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Developing loss-investigation procedures requires contextual judgment about specific retail environments, legal compliance, and nuanced risk assessment. While AI could draft template procedures or summarize best practices, the core advisory work—tailoring recommendations to a client's unique circumstances, liability exposure, and operational constraints—depends on human expertise and client interaction that AI cannot reliably replicate end-to-end. |
| Task automatability | claude-sonnet-5 | 2/5 | This is a consultative, judgment-heavy task requiring synthesis of organizational context, legal risk, and operational realities that current AI cannot fully replicate end-to-end without significant human oversight.provisions.customization.integration.beyond drafting. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Loss-investigation procedures carry meaningful legal and liability exposure; retail clients typically expect a qualified human professional to take responsibility for advice that affects investigations and compliance. Professional judgment, potential insurance/licensing considerations, and client expectation of human accountability create substantial organizational and reputational barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement exists for this advisory role, but liability concerns, need for contextual trust, and reliance on a manager's real-world investigative experience create moderate friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | A loss-prevention manager's loaded cost reflects specialized expertise and professional judgment. While AI assistance with research and drafting may reduce per-task labor, end-to-end automation is not achievable, so the cost savings do not reach parity; AI remains a partial tool rather than a replacement. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI-assisted drafting is cheap, the human expertise needed to validate, customize, and stand behind advice for a specific business context keeps overall cost comparable to or only modestly less than a human consultant. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed AI product reliably delivers full loss-investigation advisory services in production. Generic compliance drafting tools exist, but they cannot substitute for the contextual, liability-aware counsel that retail clients require, nor do they handle the interactive feedback and refinement typical of this advisory relationship. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools can draft generic policy templates or summarize best practices, but no deployed product reliably advises specific retail clients on tailored loss-investigation procedures at production scale. |
Coordinate or conduct internal investigations of problems such as employee theft and violations of corporate loss prevention policies.
25CI 25–25 · exposure 25 · augmentation 63 · importance 4.4/5 · click for rater detail
Coordinate or conduct internal investigations of problems such as employee theft and violations of corporate loss prevention policies.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI in loss prevention is nascent, with most organizations using basic analytics rather than investigation automation. Enterprise adoption remains cautious due to legal risk and the fact that investigation volume does not typically justify high-touch AI systems across most organizations. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail and loss prevention functions use AI-driven analytics for fraud/theft detection but the investigative and interviewing components remain human-led with slow adoption of full automation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully augment investigators by flagging suspicious patterns, organizing evidence, analyzing transactions, and suggesting follow-up lines of inquiry. However, the augmentation is bounded by the need for human judgment on credibility, legal procedures, and case direction. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids by flagging anomalies, analyzing transaction/video data, and organizing case evidence, substantially speeding up the investigator's workflow while humans still lead the investigation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with data analysis, document review, and pattern detection in loss prevention cases, conducting internal investigations requires human judgment on credibility assessment, witness interviews, and discretionary decision-making that AI cannot reliably perform end-to-end. The task involves complex interpersonal dynamics and legal liability that necessitate human oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | Investigations require judgment, interviewing suspects, interpreting ambiguous evidence, and making accusatory decisions with legal/HR implications that current AI cannot reliably handle end-to-end.can only assist with data analysis portions. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Internal investigations carry significant legal and HR liability; regulatory requirements and company governance often mandate that qualified human managers oversee investigations. Employment law, documentation standards, and potential litigation create strong barriers to full automation without human sign-off. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Internal investigations often carry legal liability, chain-of-custody, HR policy, and potential litigation risks requiring a qualified human to conduct interviews and make employment decisions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI analysis tools have reasonable per-case costs, but the need for human investigators to conduct interviews, make judgment calls, and manage legal/HR aspects means total AI cost savings are modest. Most of the investigative labor remains human-driven, limiting cost advantage to perhaps 20-30% savings rather than order-of-magnitude reduction. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools can cheaply flag anomalies, but the actual investigation still requires a skilled human investigator, so overall cost savings are limited to the detection phase, not the whole task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products reliably conduct full internal investigations autonomously. AI tools exist for fraud detection and document analysis, but production systems typically support rather than replace investigators. The sensitive nature and legal implications mean current AI falls short of reliable deployment for investigation coordination. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products exist for anomaly detection in POS/inventory data and video analytics flagging suspicious behavior, but no deployed system conducts full investigations including interviews and case resolution. |
Hire or supervise loss prevention staff.
21CI 7–34 · exposure 20 · augmentation 50 · importance 4.0/5 · click for rater detail
Hire or supervise loss prevention staff.
21| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Loss prevention and security roles are in conservative sectors where hiring remains heavily manual and compliance-driven; while larger enterprises use some AI screening, the pace of adoption in this domain is slower than in finance or tech, with production deployment of full hiring automation rare. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail/loss prevention management is a moderately digitized but people-centric field; AI adoption for HR-adjacent supervisory tasks is still nascent and mostly limited to screening tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI tools can assist managers by automating resume screening, suggesting interview questions, and flagging candidate inconsistencies, meaningfully reducing administrative burden; however, the core judgment work of evaluating cultural fit and supervisory capability remains human-centric. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with tasks like resume screening, scheduling optimization, or performance analytics, providing moderate productivity support to the manager without replacing judgment-based supervision. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | Substantial parts of staff recruitment (CV screening, scheduling interviews, initial candidate evaluation) can be automated or semi-automated with current tools, but final hiring decisions and ongoing supervision require human judgment on cultural fit, leadership assessment, and interpersonal dynamics that AI cannot reliably evaluate end-to-end. |
| Task automatability | claude-sonnet-5 | 1/5 | Hiring and supervising staff involves interpersonal judgment, interviewing, performance management, and interpersonal accountability that AI cannot execute end-to-end today.5 |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: employment law requires human accountability for hiring decisions, Equal Employment Opportunity compliance mandates human review, and organizational risk (wrongful termination liability, discrimination claims) pushes companies toward human sign-off on all substantive personnel decisions. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Hiring and supervisory decisions carry legal liability (employment law, discrimination, labor regulations) and typically require human accountability and signoff, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI recruitment platforms reduce screening costs but integration, training, and required human review keep overall savings modest; the loaded cost of a loss prevention manager performing hiring remains substantially higher than AI-assisted tools can offset. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this task, so cost comparison favors the human manager entirely; AI tools add cost as adjuncts rather than replacements. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI recruitment tools exist and are deployed, they are typically narrow (resume screening only) and often require significant human oversight to avoid bias and ensure fair evaluation; production systems handling full hiring and supervision workflows remain immature and limited in scope. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously hires or supervises employees; AI tools at best assist with resume screening or scheduling, not the core managerial task. |
Coordinate theft and fraud investigations involving career criminals or organized group activities.
18CI 7–28 · exposure 13 · augmentation 63 · importance 3.7/5 · click for rater detail
Coordinate theft and fraud investigations involving career criminals or organized group activities.
18| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Loss prevention is concentrated in retail and corporate security—sectors with moderate digitization. While organizations use AI tools for anomaly detection and data analytics, core investigation coordination remains heavily human-driven, and adoption of AI for autonomous investigation is low. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Retail and loss-prevention sectors are adopting AI-driven fraud analytics and video analytics at a moderate pace, though full investigative coordination remains largely human-led. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist loss prevention managers by analyzing transaction data, identifying fraud patterns, linking suspects across cases, and flagging anomalies for review. However, the inherent requirement for human judgment, evidence evaluation, and criminal expertise limits augmentation to supporting rather than transforming the full task. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can significantly aid pattern detection, link analysis, and case documentation, boosting investigator productivity while humans retain control over coordination and judgment calls. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires complex judgment in investigating organized criminal behavior, coordination across multiple parties, negotiation with law enforcement, and dynamic decision-making based on evidence that is unpredictable and context-dependent. Current AI cannot autonomously conduct investigations or make decisions about criminal cases. |
| Task automatability | claude-sonnet-5 | 2/5 | Coordinating multi-agency, multi-source investigations against organized criminal actors requires judgment, negotiation, and adaptive strategy that current AI cannot perform end-to-end; only sub-tasks like data pattern analysis can be offloaded.of |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Investigation of organized crime and fraud typically requires licensed security professionals, potential coordination with law enforcement, and legal accountability for investigative decisions and evidence handling. Liability and regulatory requirements create substantial barriers to full automation or delegation to AI systems. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Investigations often involve law enforcement liaison, chain-of-custody evidence handling, and legal/regulatory constraints requiring accountable human decision-makers, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Investigation and case coordination require skilled human loss prevention managers with domain expertise, legal knowledge, and accountability. AI tool costs for investigation support would be low, but they cannot replace the full task, making the cost ratio unfavorable when comparing AI-alone to skilled human output. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Analytics tools reduce some analyst hours but the coordination, interviewing, and legal liaison work still requires paid human investigators, keeping overall cost comparable to human-led investigation. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product reliably conducts independent theft or fraud investigations involving organized crime. While AI can assist with data analysis or pattern detection, investigation coordination—interviewing, case strategy, inter-agency liaison—requires human judgment and accountability that AI systems do not provide in production. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Fraud-detection and case-management analytics products exist and are used in loss prevention, but no deployed product autonomously coordinates full investigations into organized theft rings. |
Visit stores to ensure compliance with company policies and procedures.
6CI 0–13 · exposure 0 · augmentation 38 · importance 3.8/5 · click for rater detail
Visit stores to ensure compliance with company policies and procedures.
6| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Retail loss prevention remains heavily reliant on in-person human managers visiting stores; adoption of autonomous systems for this task is negligible because of physical, legal, and organizational requirements. Retail has not moved toward automated compliance inspection of this kind. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail loss prevention is a moderately digitized but physically-grounded sector; AI adoption for remote monitoring (cameras, analytics) is growing but on-site visits by managers remain largely unchanged. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI tools (e.g., computer vision for inventory or till checks, scheduling software) can assist loss prevention managers in preparation or data analysis, but they offer limited augmentation during the core in-person compliance visit itself, which remains primarily a human judgment and presence task. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist by analyzing store data, flagging anomalies, or pre-generating compliance checklists before a visit, but the core physical inspection still relies on human presence and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Visiting stores in person and observing compliance with nuanced company policies requires physical presence, real-time judgment about context, and adaptive interaction with staff. Current AI systems cannot autonomously traverse physical retail environments or make situated compliance assessments that involve subjective policy interpretation. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires physically traveling to store locations, walking the premises, and making contextual judgments about compliance—current AI cannot perform physical site visits or in-person observation end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Loss prevention compliance visits often involve legal liability, require human judgment about staff interactions and customer safety, and typically demand that a licensed or authorized human manager sign off on findings. Many retailers require documented human presence for audit and liability reasons. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement, but organizational expectation of physical presence, liability for missed violations, and need for judgment on-site create real friction against remote/AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The infrastructure cost to deploy mobile robotics or autonomous agents for in-store compliance visits, combined with the need for human oversight and liability management, far exceeds the loaded wage of a human loss prevention manager conducting the same task. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute for physical presence, so the comparison defaults to the human being the only viable option, making AI effectively unusable and thus not cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product reliably performs in-person store compliance inspections today. While computer vision can flag certain violations, loss prevention requires holistic evaluation of policies, staff behavior, procedures, and risk context that existing AI systems do not do in production retail settings. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs physical store visits and compliance verification autonomously; this remains a human, in-person task. |
Direct installation of covert surveillance equipment, such as security cameras.
6CI 5–7 · exposure 0 · augmentation 25 · importance 3.4/5 · click for rater detail
Direct installation of covert surveillance equipment, such as security cameras.
6| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Loss prevention and security are traditionally non-digitized, hands-on sectors with low overall automation penetration; physical surveillance installation remains a human-performed activity across nearly all organizations. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Loss prevention and security functions are adopting AI for monitoring/analytics but physical installation direction remains largely untouched by AI adoption trends. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with planning camera coverage zones or detecting optimal placement locations, but the actual installation task itself offers limited opportunity for real-time AI augmentation of the human worker. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI can help plan camera placement via analytics or floor-plan modeling, but it offers limited direct assistance to the physical/legal task of directing covert installation. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Physical installation of surveillance equipment requires hands-on placement, wiring, and integration into existing infrastructure—tasks that current AI systems cannot perform robotically today. While AI can plan camera placement or manage inventory, end-to-end installation remains a purely human-executed task. |
| Task automatability | claude-sonnet-5 | 1/5 | This is a physical, hands-on directive and installation task requiring site assessment, judgment about legal placement, and physical setup that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Installation of covert surveillance equipment is subject to privacy laws, wiretapping statutes, and workplace notification requirements that vary by jurisdiction. Many jurisdictions require explicit authorization and compliance oversight, creating legal barriers that slow or prevent automation-style substitution. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Covert surveillance installation often involves legal compliance (privacy law, workplace surveillance regulations) and requires accountable human decision-making and authorization, creating strong barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI has no role in the physical installation itself, so cost comparison is not meaningful; the human technician cost remains unchanged by any current AI system. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this physical directive task, so comparison to human labor cost favors the human by default. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product performs physical installation of surveillance hardware reliably. This remains entirely within the domain of human technicians and installers. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product directs or performs covert camera installation; this remains a human physical/managerial task with no production automation. |
Investigate or interview individuals suspected of shoplifting or internal theft.
5CI 3–7 · exposure 0 · augmentation 50 · importance 4.2/5 · click for rater detail
Investigate or interview individuals suspected of shoplifting or internal theft.
5| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Retail and loss prevention remain relatively low-digitization sectors with fragmented investment in automation. While some firms deploy AI-assisted transaction monitoring, actual interview and investigation functions remain human-centric, with adoption primarily in larger enterprise environments. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail loss prevention is a moderately digitized but people-heavy sector where interview/investigation work sees little AI agent deployment despite broader retail AI adoption in surveillance analytics. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by summarizing transaction anomalies, flagging video evidence, or organizing case materials before an interview, raising investigator productivity in preparation phases. However, the core interview task itself remains human-led, limiting the transformative potential of AI augmentation. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist by analyzing surveillance footage, flagging suspicious patterns, or helping draft interview questions/reports, but the core interview and judgment remain human-driven. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires real-time human judgment, interpersonal sensitivity, legal authority, and the ability to read non-verbal cues in complex social situations. Current AI systems cannot conduct legally defensible interviews or interrogations that would hold up in criminal proceedings. |
| Task automatability | claude-sonnet-5 | 1/5 | Investigative interviews require reading human behavior, adapting questioning strategy, legal judgment, and building rapport/trust to elicit confessions—capabilities current AI cannot perform end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Conducting interviews of suspects involves legal liability, potential false accusation, and often requires documented chain of custody and admissible evidence. Many jurisdictions impose liability on employers for improper interrogation or detention, and union agreements or corporate policy may restrict third-party or automated questioning. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Interviewing suspects involves legal liability, potential coercion claims, evidentiary standards, and often requires trained/certified personnel, creating strong barriers to non-human execution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI tools for loss prevention (transaction monitoring, video analysis) are supplementary costs; they do not replace the labor of a trained loss prevention manager who must conduct interviews, gather statements, and prepare cases for legal action. The full-task cost per output would exceed human wage. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this task, so cost comparison favors the human entirely; any AI attempt would require extensive human oversight negating savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product reliably performs suspect interviews or investigations independently. While AI can flag unusual transaction patterns or assist in evidence review, the actual investigative interview—which requires legal standing, evidence presentation, and witness interaction—remains outside current system capabilities. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product conducts suspect interviews or theft investigations autonomously; this remains a human-led interpersonal and legal process. |
Develop and maintain partnerships with federal, state, or local law enforcement agencies or members of the retail loss prevention community.
4CI 0–7 · exposure 0 · augmentation 38 · importance 3.9/5 · click for rater detail
Develop and maintain partnerships with federal, state, or local law enforcement agencies or members of the retail loss prevention community.
4| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | This task will see minimal AI displacement because it is inherently relationship-driven and requires human authority; organizational structures and regulatory frameworks mandate human executives in partnership roles. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Retail security/loss prevention is a moderately digitized but relationship- and physical-presence-driven field, with slow adoption of AI for external liaison functions specifically. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could provide marginal assistance by drafting communications, organizing contact databases, or analyzing law enforcement agency bulletins, but the core relationship-building and negotiation work must remain human-led with limited productivity gains. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help managers track contacts, draft communications, summarize incident reports for law enforcement, or analyze crime data trends to inform relationship-building conversations. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task fundamentally requires building and maintaining human relationships, trust, and negotiation with government and industry partners—activities that demand authentic interpersonal connection, institutional credibility, and real-time adaptive dialogue that current AI cannot perform end-to-end. |
| Task automatability | claude-sonnet-5 | 1/5 | Building and sustaining interpersonal relationships with law enforcement and industry peers relies on trust, in-person networking, and social judgment that current AI cannot execute end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Heavy barriers protect this task: law enforcement and institutional partners require authorized human representatives with legal standing, organizational credibility, and accountability; liability and compliance considerations ensure human leadership is legally necessary. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Liaising with law enforcement often requires vetted, authorized personnel, background trust, and organizational accountability, creating strong practical and sometimes formal barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI has no meaningful cost advantage here because the task itself cannot be automated—it requires a human representative with authority and standing to engage with partners, making any AI involvement purely supplementary at best. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this task, so cost comparison favors the human entirely since AI cannot produce the output at all. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product can autonomously establish, negotiate, or maintain institutional partnerships with law enforcement or industry bodies; these require human authority, legal standing, and relationship continuity that AI systems cannot provide. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product manages external relationship-building with law enforcement or professional communities; this remains a purely human relational task. |
Collaborate with law enforcement to investigate and solve external theft or fraud cases.
4CI 0–7 · exposure 0 · augmentation 38 · importance 3.7/5 · click for rater detail
Collaborate with law enforcement to investigate and solve external theft or fraud cases.
4| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Loss prevention remains a human-centric function with limited digital-first adoption patterns. Law enforcement collaboration by definition cannot be fully automated, constraining adoption velocity in the sector. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Loss prevention and security functions are adopting analytics tools slowly for detection, but the human-facing law enforcement collaboration itself sees little AI adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist with case data analysis, document review, or evidence organization, but the collaborative investigation and law enforcement interface fundamentally requires human judgment and legal standing. Augmentation potential is limited. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist by analyzing surveillance data, flagging fraud patterns, and drafting reports to support investigations, aiding the manager without replacing the collaborative and legal aspects of the role. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires real-time coordination with external law enforcement, adversarial investigation strategy, and negotiation with third parties—capabilities well beyond current AI systems. The collaborative, legal, and interpersonal complexity cannot be automated end-to-end. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires interpersonal liaison, judgment calls, negotiation with law enforcement, and case-specific investigative reasoning that AI cannot perform end-to-end today.atability is minimal. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Law enforcement collaboration and fraud investigation are heavily regulated and typically require licensed investigators, legal authority to act, and chain-of-custody obligations. Liability and evidentiary standards create hard barriers to AI substitution. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Interfacing with law enforcement, providing testimony, and handling legal evidence chains typically requires an authorized human representative, creating strong organizational and legal barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The task is highly specialized and requires licensed investigation expertise, legal sign-off, and law enforcement coordination. AI has no cost advantage over the skilled human labor required for this sensitive work. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this function, so cost comparison favors the human entirely; AI cannot replace the liaison role. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product performs law enforcement collaboration or fraud investigation at scale in production. This requires human judgment, legal authority, and cross-organizational relationships that current systems cannot execute independently. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product manages law enforcement collaboration and case-solving for theft/fraud; this remains a human relationship-driven activity. |
Provide recommendations and solutions in crisis situations such as workplace violence, protests, and demonstrations.
4CI 4–4 · exposure 0 · augmentation 38 · importance 4.0/5 · click for rater detail
Provide recommendations and solutions in crisis situations such as workplace violence, protests, and demonstrations.
4| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Organizations cannot substitute AI for crisis decision-making due to legal exposure and the need for qualified human judgment; adoption remains minimal and limited to advisory or monitoring tools rather than autonomous recommendation systems. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Security and loss prevention functions are low-digitization, physically grounded roles where AI adoption for crisis decision-making is minimal to nonexistent in practice. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist by analyzing historical incident data, flagging potential escalation patterns, or organizing threat assessment frameworks, but the core task of crisis recommendation requires experienced human judgment and accountability. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist by aggregating situational data, monitoring social media for protest indicators, or suggesting protocol checklists, providing moderate support while the human retains full decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Crisis response requires real-time judgment about volatile human behavior, stakeholder safety, legal liability, and dynamic escalation patterns that current AI cannot reliably assess or act upon without human expertise and authority. |
| Task automatability | claude-sonnet-5 | 1/5 | Crisis response demands real-time judgment, physical situational awareness, and accountability that current AI cannot exercise independently; no off-the-shelf system can end-to-end manage such a task. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Strong legal, regulatory, and liability barriers exist: crisis decisions involving safety, potential violence, and property protection require certified loss prevention professionals with legal accountability and organizational sign-off authority. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Crisis management involving violence and public safety typically requires trained, authorized personnel with legal accountability, liability exposure, and organizational protocols mandating human decision-makers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Crisis management tools may assist human managers, but the core recommendation function must remain human-driven; the all-in cost of AI systems for advisory support is comparable to or exceeds the value relative to human expert fees. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Given AI cannot reliably perform the core judgment task, any cost comparison favors humans since AI failures in crisis situations carry very high liability and safety costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product can autonomously provide crisis recommendations in high-stakes situations involving workplace violence or protests; these require qualified human professionals with accountability and contextual authority. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously provides crisis recommendations for workplace violence or protests; at best AI is used for background monitoring or information retrieval, not decision-making in these scenarios. |
Related occupations — Management
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.