Operations Research Analysts
15-2031.00Median wage $88,940/yr108,510 employed (US)Rank #188 of 923 scored · top 20% by substitution
Formulate and apply mathematical modeling and other optimizing methods to develop and interpret information that assists management with decisionmaking, policy formulation, or other managerial functions. May collect and analyze data and develop decision support software, services, or products. May develop and supply optimal time, cost, or logistics networks for program evaluation, review, or implementation.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution39
Exposure33
Augmentation77
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
17 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
6%
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%34
panel mean rating 2.3/5 → substitution pressure 34/100
Technical feasibility todayw 20%32
panel mean rating 2.3/5 → substitution pressure 32/100
Cost vs. human wagew 15%36
panel mean rating 2.4/5 → substitution pressure 36/100
Adoption barriersw 20%inverted — strong barriers lower the score57
panel mean rating 2.7/5 (barrier strength) → substitution pressure 57/100
Sector adoption velocityw 10%44
panel mean rating 2.8/5 → substitution pressure 44/100
Task breakdown (17 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.
Review research literature.
81CI 79–84 · exposure 75 · augmentation 100 · importance 3.5/5 · click for rater detail
Review research literature.
81| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Academic and professional services sectors are rapidly integrating AI literature-review tools into workflows. ChatGPT, Perplexity, and specialized research platforms are widely adopted by graduate students, consultants, and analysts; adoption is moving quickly from pilots to routine use. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Operations research and adjacent analytics fields sit in the professional services/data-intensive sector where AI research and summarization tools have seen fast uptake. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically assists literature reviewers by automating the most time-consuming phase (scanning and summarizing), freeing researchers to focus on critical appraisal, synthesis, and novel insight generation. The human analyst is more productive and strategic when AI handles the bulk-processing phase. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI dramatically accelerates locating, filtering, and summarizing literature, letting analysts focus on evaluation and application while remaining in the loop. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Large language models can efficiently scan, summarize, and extract key findings from research papers at scale, reducing the time spent on literature review by 60–80%. However, critical appraisal, contextual judgment of relevance to specific research questions, and synthesis into novel frameworks typically still require human oversight and domain expertise. |
| Task automatability | claude-sonnet-5 | 4/5 | LLMs can search, summarize, and synthesize research literature quickly, meeting or exceeding the 50% time-saving threshold for most literature review work, though final judgment on relevance/quality still benefits from human oversight. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers exist; researchers and analysts can freely adopt these tools. The main friction is institutional preference for human verification and organizational reluctance to trust automated summaries without spot-checking, but these are soft, not hard, barriers. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or regulatory requirement mandates a human review literature; this is a research support task with minimal legal or liability constraints. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | The cost of AI inference and integration (negligible for API-based services) is orders of magnitude lower than the loaded wage of a research analyst spending days or weeks on manual literature review. Even with oversight, the cost-per-task ratio favors AI by 10:1 or more. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | AI-based literature search and summarization costs a small fraction of an analyst's hourly wage for equivalent coverage of papers. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (academic AI tools, ChatGPT, Consensus, and specialized research platforms) demonstrably perform literature scanning, summarization, and keyword extraction in production today. Minor limitations include occasional hallucinations in citation details and inability to access behind-paywall papers, but the core task is reliably executed at scale. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed tools (e.g., AI research assistants, summarization tools integrated into reference managers and search engines like Elicit, Consensus, Scite) reliably perform literature discovery and summarization in production today, though not fully autonomous for complex synthesis. |
Educate staff in the use of mathematical models.
61CI 38–85 · exposure 58 · augmentation 75 · importance 3.6/5 · click for rater detail
Educate staff in the use of mathematical models.
61| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Operations research teams and analytics functions show moderate AI adoption for content generation and explanation, with pilots and some production use of AI tutoring systems, but live training and staff education remain partially human-led in most enterprises. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research and analytics fields have moderate AI adoption for documentation and training aids, though live staff education remains a slower-adopting activity. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists human educators by drafting lesson plans, generating examples, producing visualizations of models, and personalizing explanations, thereby multiplying instructor effectiveness and freeing human experts to focus on mentoring, clarification, and complex problem-solving. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly help analysts prepare training materials, generate explanations, create interactive examples, and answer follow-up questions, meaningfully boosting the educator's productivity. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Creating educational content, explanatory materials, and instructional videos about mathematical models can be largely automated using AI systems that generate text, visualizations, and even interactive simulations. Current tools can draft lectures, produce diagrams, and tailor explanations to different skill levels, achieving >50% time savings compared to human instruction preparation. |
| Task automatability | claude-sonnet-5 | 2/5 | Explaining models can be partly supported by AI-generated documentation or tutoring content, but effective staff education requires live interaction, contextual adaptation, and organizational trust-building that current AI cannot fully replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers prevent AI from generating educational materials about mathematical models; however, organizations often prefer human instructors for credibility, complex Q&A, and organizational trust, creating modest adoption friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing barrier exists, but organizational preference for human-led training, trust-building, and interactive Q&A creates moderate friction against pure AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-generated educational content (text, video scripts, interactive modules) costs a fraction of traditional instructor time once developed; the cost per learner scales down dramatically compared to live training, making AI orders of magnitude cheaper at scale. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can generate explanatory materials cheaply, the human analyst still needs to tailor training, answer nuanced questions, and build buy-in, limiting overall cost savings relative to a human trainer. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products exist (LLMs, educational AI platforms, automated documentation tools) that reliably generate training materials and explanations of mathematical concepts; however, live instruction delivery and real-time adaptation to complex learner needs still benefit from human facilitation, preventing a full 5 rating. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tutoring and explanation tools exist and are used for technical content, but no deployed product reliably conducts full staff training on custom mathematical models within specific organizational contexts. |
Break systems into their components, assign numerical values to each component, and examine the mathematical relationships between them.
58CI 46–70 · exposure 53 · augmentation 88 · importance 3.7/5 · click for rater detail
Break systems into their components, assign numerical values to each component, and examine the mathematical relationships between them.
58| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Operations research, analytics, and data science teams are rapidly adopting AI-assisted modeling, optimization software, and automated data analysis; information, finance, and tech sectors show fast integration of these tools in production. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within analytics-heavy sectors (finance, logistics, consulting) with moderate AI tool adoption, though full modeling automation is still mostly pilot-stage. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI excels at assisting analysts by automating data exploration, suggesting component relationships, and generating candidate models, while analysts apply judgment to validate, refine, and interpret results—this human-AI pairing significantly amplifies productivity. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly speed up formulation of equations, suggest relevant variables, and check mathematical consistency, meaningfully boosting analyst productivity while the analyst retains conceptual control. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI systems can decompose systems, assign quantitative metrics, and identify mathematical relationships at a high level today. However, domain expertise is often required to validate component definitions and ensure relationships capture the true system dynamics, limiting full end-to-end replacement. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can assist in formalizing systems into mathematical models and identifying relationships, especially with well-structured problems, but decomposing novel or ambiguous real-world systems into components still requires substantial human domain judgment and validation.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few hard regulatory or licensing barriers exist for automated system analysis itself, though organizational preference for human validation and business risk aversion create moderate friction to full substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI use, but organizational trust in mathematical modeling correctness and consequences of flawed models create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven analytics and modeling platforms offer significant cost savings compared to manual system decomposition and relationship analysis, particularly for large-scale or data-intensive problems, though oversight and domain expert review still add cost. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Because human oversight and validation of model structure are still essential, the all-in cost of AI assistance plus analyst review is not dramatically cheaper than a skilled analyst working with existing modeling software. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Machine learning and optimization tools exist and are used in production for parts of this task (e.g., data analysis, relationship discovery), but most deployments still require substantial human direction on system framing and validation of mathematical models. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | LLMs and optimization tools can draft formulations for known problem classes, but no deployed product reliably performs open-ended system decomposition and quantification across diverse business contexts without heavy human curation. |
Prepare management reports defining and evaluating problems and recommending solutions.
47CI 37–57 · exposure 42 · augmentation 88 · importance 4.5/5 · click for rater detail
Prepare management reports defining and evaluating problems and recommending solutions.
47| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Professional services and large enterprises are piloting AI-assisted report generation and analytics, but production adoption remains limited. Many organizations still rely on analysts for final report ownership and recommendation authority. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Operations research analysts work within professional/business services and finance-adjacent functions where generative AI adoption for report drafting and analysis summarization is already widespread and growing quickly. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at accelerating data extraction, outlining options, drafting sections, and surfacing patterns that analysts then refine and contextualize. This assistance substantially raises analyst productivity while the human analyst retains judgment on problem framing and solution fit. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially speeds up drafting, summarizing data, formatting, and generating first-pass recommendations, letting analysts focus on validation and strategic judgment while remaining in the loop. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate report drafts and perform data analysis, the task requires problem definition and solution recommendation that depend heavily on organizational context, stakeholder priorities, and judgment about feasibility and impact. Current systems cannot reliably understand unstated constraints or deliver management-ready recommendations without substantial human review and iteration. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft report structure, summarize data, and generate recommendation language from provided analysis, but defining the right problem and validating recommendations against business context typically requires human judgment and domain expertise. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Management reports often carry liability and decision-making weight; organizations typically require analyst sign-off and accountability. Regulatory oversight is limited but internal governance (quality control, stakeholder sign-off) creates friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for producing internal management reports, though organizational risk tolerance and accountability for recommendations create some friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce writing and data compilation time, but the integration overhead (ensuring accuracy, embedding in organizational context, management-level review) and required human oversight offset cost savings. For analysts handling high-stakes problems, total cost remains comparable to or above human-only workflows. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI drafting tools are cheap per report, but the analyst's oversight, data validation, and domain-specific judgment still consume significant time, making net cost savings moderate rather than dramatic. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | LLMs and BI tools can draft report sections, summarize analyses, and outline recommendations in production settings, but output quality varies and often requires domain expertise to validate. No deployed product reliably produces end-to-end management reports without human intervention and refinement. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | LLM-based tools are commonly used today to draft business/management reports and summarize analytical findings, but reliability on complex, novel operations research problems with nuanced recommendations remains inconsistent. |
Develop and apply time and cost networks to plan, control, and review large projects.
47CI 39–55 · exposure 42 · augmentation 75 · importance 3.8/5 · click for rater detail
Develop and apply time and cost networks to plan, control, and review large projects.
47| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI-driven project planning is still in the pilot phase across most sectors; traditional project management tools (MS Project, Jira) dominate, and organizations move slowly on mission-critical planning automation. Financial and large tech firms lead, but uptake remains far from mainstream even in information-heavy industries. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Project management and analytics sectors show moderate AI adoption—pilots and add-on features are common in scheduling tools, but full-scale autonomous network planning remains uncommon in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at augmenting this task: rapid scenario generation, cost and schedule sensitivity analysis, alternative network suggestions, and automated updates as project parameters change. Analysts using AI assistants can explore vastly more options and validate plans more thoroughly while retaining judgment and accountability. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids scenario modeling, critical path recalculations, and cost forecasting, enhancing analyst productivity while humans retain oversight over assumptions and strategic decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can help construct network diagrams, perform calculations, and suggest optimizations for project timelines, the task requires iterative human judgment about project scope, resource constraints, and trade-offs that vary significantly across contexts. Current systems lack the domain-adaptive reasoning needed for end-to-end autonomous control and review of large projects at equal quality. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can build and analyze time/cost network models (CPM/PERT) given structured project data and clear parameters, but scoping the project, gathering accurate inputs, and adapting to organizational context still require human judgment.dividido, so only partial end-to-end automation is achievable today. Roughly half the task—computation, scenario analysis, and reporting—can be automated with proper setup. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | There are moderate friction points: large projects often require sign-off from licensed project managers or senior engineers, organizational risk aversion to AI-driven planning, and liability concerns if automated plans fail. However, no single hard legal or regulatory requirement mandates human authorship, allowing cautious automation in lower-risk contexts. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI use in project scheduling, though large projects often require accountable human sign-off for contractual and risk-management reasons, creating moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI infrastructure (LLMs, optimization engines, project management APIs) is relatively inexpensive per inference, but the total delivered cost includes integration, data preparation, and ongoing human review to ensure plan validity. This approaches the cost of a mid-level analyst hour for routine planning work, but surpasses it for novel or high-stakes projects. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Software licenses and analyst time are still needed to validate and interpret model outputs; while computation is cheap, the overall cost including data preparation and review is roughly comparable to human-led planning for large projects. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI tools for project management and scheduling exist (e.g., Gantt generation, CPM analysis), and some systems can ingest project data and produce network visualizations. However, real production use still requires substantial human validation, iterative refinement, and oversight due to domain complexity and the consequences of planning errors. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Project management and scheduling software (e.g., MS Project, Primavera, some AI-enhanced planning tools) can generate and optimize network diagrams, but reliable end-to-end deployment for large, complex projects still requires significant human oversight and customization. |
Study and analyze information about alternative courses of action to determine which plan will offer the best outcomes.
44CI 32–55 · exposure 38 · augmentation 88 · importance 4.3/5 · click for rater detail
Study and analyze information about alternative courses of action to determine which plan will offer the best outcomes.
44| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Operations research and analytics teams use AI tools (optimization software, simulation platforms) routinely, but primarily for augmentation; full autonomous replacement of analyst judgment remains uncommon in production. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within professional/analytics fields showing moderate-to-fast AI tool adoption, though many organizations still rely on traditional modeling workflows with AI as an add-on. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially assists with data synthesis, scenario generation, constraint modeling, and outcome prediction, allowing analysts to explore more alternatives faster and make more informed recommendations while remaining in the decision loop. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI significantly enhances scenario generation, data analysis, and comparative evaluation of alternatives, greatly boosting analyst productivity while the human retains final decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can support data compilation and scenario modeling, but the core task of evaluating tradeoffs and determining 'best outcomes' requires human judgment about values, constraints, and organizational context that AI cannot reliably perform end-to-end at the quality threshold. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can rapidly generate and evaluate scenarios, run simulations, and summarize tradeoffs, but selecting the 'best' plan requires domain judgment, stakeholder priorities, and validation that current AI cannot fully own end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Organizations often require human accountability for strategic recommendations, and decision-makers prefer human analysts they can interrogate; however, no hard legal barrier prevents AI assistance or partial automation of the analysis phase. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No formal licensing requirement, but organizational risk aversion and need for accountable human judgment in high-stakes operational decisions create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI reduces computational and data-prep costs but cannot fully replace the analyst's judgment role; the combined cost of AI tooling plus required human review and validation remains comparable to hiring an operations research analyst. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted analysis can cut analyst hours significantly for data processing and modeling, but licensing, integration, and validation costs keep overall cost roughly comparable to skilled analyst time for complex decisions. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI excels at data analysis and generating optimization scenarios, no deployed product reliably performs the full task of evaluating competing plans and making a defensible 'best outcome' recommendation without substantial human oversight and domain expertise. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed decision-support and optimization tools (e.g., simulation software, LLM-assisted analysis) exist and are used in production, but reliability varies by domain complexity and often requires expert oversight. |
Specify manipulative or computational methods to be applied to models.
40CI 30–50 · exposure 33 · augmentation 75 · importance 4.0/5 · click for rater detail
Specify manipulative or computational methods to be applied to models.
40| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for method specification in OR is occurring in technology-forward firms but remains slow in broader business and government sectors. Most organizations still rely on analyst expertise; production-scale automation is uncommon despite pilot interest. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | OR sits within analytics-heavy sectors (finance, logistics, consulting) that are adopting AI coding/reasoning tools moderately fast, though full method-selection automation remains uncommon in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at augmenting analysts by rapidly generating and comparing candidate methods, visualizing trade-offs, and surfacing literature-supported approaches. Analysts can use these suggestions to accelerate design cycles and explore wider solution spaces while maintaining human judgment over final selection. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools (e.g., code generation, optimization libraries suggestions, natural-language explanation of methods) meaningfully speed up analysts' ability to draft and compare computational approaches. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Specifying which computational methods to apply to models requires domain expertise, understanding of problem structure, and judgment about trade-offs between accuracy and computational cost. While AI can suggest methods given a problem description, the critical selection and justification typically require human expertise and cannot consistently meet the 50% time-saving bar without close human oversight. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can suggest and even implement optimization/computational methods (e.g., LP, heuristics, simulation approaches) given a well-specified model, but choosing the right method requires judgment about problem structure, tractability, and business context that still needs expert oversight. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | OR analysts' work is embedded in professional practices and organizational decision-making where accountability for method choice matters. While no hard legal license requirement exists, organizational friction around trusting automated recommendations and the need for expert sign-off provide moderate barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational reliance on correct model specification for high-stakes decisions creates moderate friction and need for human validation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI advisory tools are not yet cheap enough to decisively undercut the cost of experienced analysts reviewing and validating method selection. Integration costs and human oversight requirements push the cost ratio closer to parity or favor human expertise, especially for high-stakes problems. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI can cheaply draft candidate methods and code, reducing analyst time, but validation, tuning, and integration still require paid expert time, keeping overall cost roughly comparable rather than dramatically cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some AI tools (e.g., constraint solvers, AutoML platforms) can recommend or apply computational methods to specific model classes, but mature deployed products that reliably specify appropriate manipulative methods across diverse OR problem contexts remain limited. Systems exist but lack broad reliability and generalization. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Coding assistants and AI tools can propose algorithms or generate solver code, but no deployed product reliably specifies full computational methodologies for novel OR models without significant analyst guidance. |
Define data requirements, and gather and validate information, applying judgment and statistical tests.
39CI 28–50 · exposure 33 · augmentation 75 · importance 4.5/5 · click for rater detail
Define data requirements, and gather and validate information, applying judgment and statistical tests.
39| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Financial and tech-heavy operations research teams have adopted automated data pipelines and validation frameworks, but the judgment and requirement-definition steps remain largely manual across the sector. Adoption is uneven and primarily assistive rather than replacement-focused. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within professional/analytics services where AI tool adoption is moderately fast, though data requirement definition remains a less-automated pilot-stage activity. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists this task through automated anomaly detection, missing-data flagging, statistical test suggestions, and data profiling dashboards. A human analyst's productivity and confidence in validation can improve substantially with these tools while retaining final judgment responsibility. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly speeds up data profiling, cleaning scripts, and statistical testing, letting analysts focus more time on judgment-intensive requirement definition. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Defining data requirements and gathering raw data are partially automatable through scripting and ETL tools, but validating information and applying statistical judgment require substantial human expertise. AI can assist with preliminary validation (anomaly detection) but cannot reliably determine whether data adequately answers domain-specific research questions. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can assist with defining schemas, generating validation scripts, and running statistical tests, but the initial judgment about what data is relevant and appropriate for a novel business problem still requires substantial human domain expertise.To fully replace this end-to-end would require reliable independent judgment, which current systems lack. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Operations research projects typically require sign-off on data requirements and validation by qualified practitioners due to decision impact and regulatory context (financial, supply-chain, healthcare). Organizational practice and liability concerns strongly favor human certification of data integrity. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement bars AI use, but organizational risk aversion around data quality and analytic judgment creates moderate friction to fully delegate this to AI. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Automating data pipelines with cloud tools is cheaper than manual collection, but the expert judgment and validation components remain labor-intensive. The all-in cost of AI infrastructure plus human oversight is comparable to or slightly less expensive than a junior analyst, but not dramatically cheaper. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI tools can cheaply automate data validation and statistical testing subroutines, but the requirements-definition and judgment components still require analyst oversight, keeping blended cost roughly comparable to a human doing the full task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed tools exist for data gathering (web scraping, API integration) and basic statistical checks, but no end-to-end product reliably performs the judgment-intensive validation and requirements-definition aspects that distinguish this task in operations research. Human oversight remains necessary. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products like AI coding/data assistants and BI copilots can perform pieces of this (data profiling, anomaly detection, statistical test execution) but no deployed product autonomously defines data requirements and validates data with human-level judgment in production analytics workflows. |
Present the results of mathematical modeling and data analysis to management or other end users.
37CI 32–42 · exposure 30 · augmentation 75 · importance 4.6/5 · click for rater detail
Present the results of mathematical modeling and data analysis to management or other end users.
37| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Professional services and analytics-heavy firms are experimenting with AI-assisted reporting and summarization, but production deployment remains limited to support roles rather than replacement of the human presentation function. Pilots are common, but full displacement is not yet observed at scale. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Analytics and professional services sectors adopt AI tools for report generation and visualization fairly quickly, but the presentation/communication step lags behind more back-office automatable tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at augmenting this task: LLMs can draft multiple narrative angles on findings, visualization tools can auto-generate charts from data, and summarization can help analysts distill complex results into management-friendly language. These tools significantly raise analyst productivity in preparing presentations while the analyst retains full control over message and accuracy. |
| Augmentation potential | claude-sonnet-5 | 4/5 | Generative AI significantly speeds up creation of charts, summaries, and narrative explanations that analysts use when presenting, meaningfully boosting productivity while the human still delivers the presentation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate data summaries and create visualizations automatically, the task requires contextualizing results for specific management audiences, anticipating questions, and tailoring technical findings to business priorities—judgment-heavy work that current AI cannot fully replace. Partial automation of chart generation and writeup drafting is possible, but end-to-end presentation to live stakeholders remains heavily dependent on human interpretation and delivery. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft slides, summaries, and visualizations, but the live presentation to management—reading the room, answering unscripted questions, and building trust in recommendations—still requires a human presenter today..' |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | No hard legal or licensing barrier prevents AI assistance, but organizational practice, stakeholder expectations (preference for human-delivered insights), and quality assurance requirements (results must be vetted by a qualified analyst) create significant friction to full automation. Management typically expects a human expert to stand behind findings. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational norms strongly favor a human analyst presenting to build credibility and answer follow-up questions, creating moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted report generation and visualization can reduce labor on formatting and initial narrative drafting, but the skilled human analyst—whose domain expertise and judgment drive the presentation's credibility—remains essential and cost-offsetting. The integration overhead and required human review keep total AI cost closer to the human cost of the full task. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply generate draft materials, but the human labor of delivering, contextualizing, and defending results to stakeholders remains necessary, keeping overall cost savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Tools like LLMs can draft executive summaries and business-focused narratives from analytical outputs, and visualization software is mature, but no deployed product reliably transforms raw modeling results into polished, audience-specific presentations without human review and editing. Products exist for components (summarization, charting) but not for the integrated task of delivering findings credibly to decision-makers. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Tools like Copilot/ChatGPT-based deck generators exist and are used to prepare content, but no deployed product autonomously delivers and defends findings to a management audience in real time. |
Formulate mathematical or simulation models of problems, relating constants and variables, restrictions, alternatives, conflicting objectives, and their numerical parameters.
37CI 25–50 · exposure 33 · augmentation 75 · importance 4.4/5 · click for rater detail
Formulate mathematical or simulation models of problems, relating constants and variables, restrictions, alternatives, conflicting objectives, and their numerical parameters.
37| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | OR is practiced primarily by specialized teams in larger enterprises and consultancies with existing expertise. These organizations move cautiously on automation given model criticality, and adoption of AI model-formulation assistance remains pilot-stage rather than production-scaled. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within professional/analytics services where AI coding and modeling assistants are being piloted, but production-scale automation of full model formulation is still uncommon. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can meaningfully assist analysts by suggesting mathematical structures, identifying constraint patterns, auto-completing parameter definitions, and checking dimensional consistency—all while the human retains model control and validation. This augmentation productivity gain is substantial for skilled practitioners. |
| Augmentation potential | claude-sonnet-5 | 4/5 | LLMs are quite useful for brainstorming model structures, writing formulation code, and suggesting constraints/variables, meaningfully speeding up an analyst's iterative modeling process. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with mathematical notation and parameter extraction from documents, formulating complete models requires deep understanding of problem context, trade-offs, and domain constraints that typically demand human expert judgment. Current systems struggle to autonomously translate ambiguous real-world requirements into coherent mathematical structures without significant human validation. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can help draft formulations of standard OR problems (LP, scheduling, queuing) from a description, but for novel or complex real-world problems it still needs substantial expert validation and iteration, so it doesn't yet meet the 50% time-saving-at-equal-quality bar broadly. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Operations research models often support high-stakes business and policy decisions; liability concerns and organizational risk aversion create strong pressure for human expert sign-off. Many regulated industries require documented human accountability for model specification. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement exists for this task, but organizational trust, complexity, and the high cost of formulation errors create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted model formulation requires expert human oversight, integration costs, and validation effort that rivals or exceeds the cost of having the analyst formulate the model directly. The overhead of managing AI output quality offsets computational savings. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted drafting is cheap per query, but oversight, validation, and correction by a skilled analyst remain necessary, keeping overall cost roughly comparable to human-only work for nontrivial problems. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some tools exist for symbolic mathematics and constraint generation (e.g., pyomo with LLM assistance, constraint solvers), but they operate in narrow domains and require substantial human setup and validation. No deployed product reliably formulates novel, complex OR models end-to-end from problem statements. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | There are demos and coding-assistant capabilities (e.g., LLMs writing optimization code or formulating constraints) but no mature deployed product reliably performs full problem formulation across diverse business contexts in production. |
Observe the current system in operation, and gather and analyze information about each of the component problems, using a variety of sources.
35CI 32–38 · exposure 25 · augmentation 75 · importance 4.4/5 · click for rater detail
Observe the current system in operation, and gather and analyze information about each of the component problems, using a variety of sources.
35| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Adoption of analytics and BI tools is widespread in information and finance sectors, but AI-driven system observation and root-cause analysis are still primarily pilot-stage or early production. Significant resistance remains in risk-sensitive domains. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within analytics-heavy professional services and consulting sectors that are adopting AI tools moderately fast, though full task automation lags behind simpler analytic tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems demonstrably assist analysts by automating data collection, generating preliminary summaries, flagging anomalies, and suggesting correlations across sources, substantially reducing time spent on routine analysis and allowing humans to focus on interpretation and strategic insight. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially assists by rapidly aggregating and analyzing information from multiple sources, letting analysts focus more time on direct observation and judgment-based problem framing. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can gather data from digital sources and perform routine analysis on structured information, observing systems in operation and synthesizing insights from diverse, often unstructured sources requires contextual understanding and domain expertise that current systems struggle with at scale. Significant human judgment is needed to identify which components matter and how to weight conflicting signals. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help analyze data and synthesize information from documents, but directly observing a live operational system in situ and identifying component problems requires physical/organizational presence and contextual judgment that current AI cannot replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | No hard legal barriers exist, but organizational friction is real: domain experts must validate findings, clients often prefer human analysts for credibility and accountability, and the need for human judgment to interpret system behavior creates practical adoption friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI use, but organizational trust, access to proprietary systems, and the need for contextual judgment create moderate practical friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Data collection and analysis tools are relatively inexpensive, but the cost of ensuring accuracy and comprehensiveness in system observation, plus required human oversight of findings, means total cost approaches or may exceed that of a trained analyst performing the work directly. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Data analysis portions can be cheaply automated, but the observational and cross-source synthesis work still requires substantial human analyst time and oversight, keeping blended costs closer to comparable than dramatically cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Products exist for data collection and basic analytics (BI tools, analytics platforms), but comprehensive system observation and multi-source problem component analysis in real operational contexts remains largely manual. Deployed systems handle narrow, well-defined analysis tasks but not the full scope of observation and synthesis this task demands. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Deployed analytics and NLP tools can process logs, reports, and interview transcripts, but no product reliably performs full system observation and problem decomposition autonomously in production today. |
Perform validation and testing of models to ensure adequacy, and reformulate models, as necessary.
34CI 30–37 · exposure 30 · augmentation 75 · importance 4.5/5 · click for rater detail
Perform validation and testing of models to ensure adequacy, and reformulate models, as necessary.
34| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Operations research remains concentrated in larger organizations and specialized consultancies with slower digital transformation; adoption of AI for validation tasks is still in pilot phase rather than widespread production deployment. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within analytics-heavy sectors (finance, logistics, consulting) with moderate AI tool adoption for diagnostics, but full validation workflows remain human-led with pilots more common than production automation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can significantly assist analysts by automating diagnostic test execution, generating validation reports, suggesting statistical anomalies, and recommending reformulation directions, allowing humans to focus on judgment and strategic modeling decisions rather than routine computational checking. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly speed up sensitivity analysis, scenario testing, and code-based validation checks, meaningfully boosting analyst productivity while the analyst retains final judgment on model adequacy. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Model validation typically requires domain expertise, judgment about statistical adequacy, and context-specific decision-making about when reformulation is needed. While AI can assist with running tests and generating diagnostic outputs, the critical validation judgment and reformulation strategy decisions remain human-dependent, limiting time savings to <50%. |
| Task automatability | claude-sonnet-5 | 2/5 | Model validation requires designing test cases, interpreting domain-specific outputs, and judging adequacy against business context, which current AI can assist but not fully execute end-to-end reliably.support statistical tests but cannot autonomously judge model adequacy or reformulate models without heavy human oversight.rationale trimmed. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Organizational standards and domain expertise requirements create moderate friction; stakeholders often prefer human analysts to sign off on model adequacy given high-stakes consequences. However, no hard legal barrier requires a human to perform all validation steps. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but organizational reliance on analyst judgment and accountability for model correctness in high-stakes decisions creates moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce some computational overhead and generate test reports, but the loaded cost of an operations research analyst performing expert validation and model redesign remains lower than AI infrastructure, oversight, and quality-assurance costs for the same task end-to-end. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply run diagnostic computations, but the human oversight, judgment, and iterative reformulation required keep overall costs comparable to or only modestly below human-only effort. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Tools like automated testing frameworks and statistical software can perform some validation checks, but no mature end-to-end system reliably validates complex models or decides when and how to reformulate them without human expert review. Production systems exist for narrow validation tasks but not for the full scope. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some AI/statistical tools automate parts of testing (cross-validation, sensitivity analysis) but no deployed product autonomously validates and reformulates OR models in production without analyst supervision. |
Analyze information obtained from management to conceptualize and define operational problems.
33CI 25–41 · exposure 25 · augmentation 75 · importance 4.3/5 · click for rater detail
Analyze information obtained from management to conceptualize and define operational problems.
33| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Operations research remains concentrated in large corporations, finance, and defense contractors with slower digital transformation and high conservatism around automated problem framing in strategy-adjacent work; most adoption is tooling (better data access) rather than automation of problem conceptualization itself. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within professional services/analytics sectors with moderate AI adoption, but this specific problem-definition step lags behind more automatable quantitative modeling tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can significantly augment this task by rapidly synthesizing management input, identifying patterns in operational data, suggesting alternative problem frames, and generating structured problem statements for analyst review—leaving the analyst to validate, refine, and own the final definition. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can help synthesize documents, summarize stakeholder input, and draft problem statements, significantly aiding analysts even though humans retain the core conceptualization role. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist in structuring and summarizing management input, but conceptualizing and defining operational problems requires understanding business context, strategic intent, and stakeholder priorities that typically need human judgment and domain expertise. Current systems lack the interactive discovery and framing capability needed end-to-end. |
| Task automatability | claude-sonnet-5 | 2/5 | Defining and conceptualizing ambiguous operational problems requires deep contextual understanding of organizational goals, politics, and constraints that AI cannot yet independently extract from management discussions.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong organizational barriers exist: operations research findings inform strategic decisions, making stakeholder trust in the problem definition critical; management typically expects direct engagement with a qualified analyst; and liability for mischaracterizing problems falls on the organization, creating incentive to retain human judgment. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational trust, need for direct stakeholder engagement, and accountability for correctly framing high-stakes problems create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI-powered analysis tools (document processing, summarization) can reduce time spent organizing and synthesizing management input, approaching cost parity with human analysts for the information-gathering phase, but the conceptualization step still requires human effort. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Given current limitations, AI assistance still requires substantial human oversight and iteration, so cost savings versus a skilled analyst are modest rather than order-of-magnitude. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While LLMs can parse management briefs and suggest problem framings, no deployed product reliably translates unstructured management conversations into validated operational problem definitions without material human iteration and verification. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed product autonomously conducts stakeholder analysis and problem framing from raw management input; existing tools assist with structuring information but do not replace this judgment-heavy step. |
Design, conduct, and evaluate experimental operational models in cases where models cannot be developed from existing data.
31CI 25–38 · exposure 25 · augmentation 75 · importance 4.0/5 · click for rater detail
Design, conduct, and evaluate experimental operational models in cases where models cannot be developed from existing data.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While professional services and finance sectors are adopting AI, experimental design automation remains in pilot stage. Organizations still rely on human analysts for novel model development, and adoption of AI-driven experimental design in production systems is slow and sector-dependent. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research sits within business/finance/logistics sectors with moderate-to-fast digital tool adoption, but the specific sub-task of building experimental models from scratch is still mostly pilot-level with AI assistance. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI tools demonstrably assist operations research analysts by automating simulation runs, generating sensitivity analysis, producing statistical summaries, and suggesting alternative model structures—all while the analyst retains design authority and judgment. This augmentation materially raises analyst productivity on exploratory modeling work. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully help analysts by generating simulation code, suggesting model structures, exploring parameter spaces, and analyzing results, significantly speeding up parts of the experimental design workflow. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Designing and evaluating experimental models requires domain expertise, creative problem-framing, and judgment about what to test—capabilities that current AI struggles with. While AI can assist with statistical analysis and simulation execution of a pre-designed model, the core task of conceiving novel experimental designs from scratch remains heavily human-dependent. |
| Task automatability | claude-sonnet-5 | 2/5 | Designing novel experimental models for situations lacking existing data requires original judgment, domain expertise, and creative hypothesis formation that current AI cannot reliably perform end-to-end.pynb; AI can assist with pieces but cannot autonomously design and validate a full experimental operational model. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Operations research models often inform high-stakes business or policy decisions where liability and error costs are asymmetric and significant. Organizations typically require a credentialed analyst to validate, sign off on, and take responsibility for experimental models, creating organizational and professional barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No formal licensing requirement for this specific task, but organizational risk aversion and the need for domain-expert sign-off on novel models create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The integration, domain configuration, and expert validation overhead for using AI tools to design and run experimental models is substantial relative to the labor cost saved. An operations research analyst's judgment in model design adds significant value that AI cannot replace, making the all-in cost unfavorable. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Because heavy human oversight, domain validation, and iterative model-building are required, AI cost savings are modest once integration and expert review are factored in, though some drafting/coding tasks are cheaper via AI. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs end-to-end experimental design and evaluation at the level required for operations research. AI can support components (running simulations, analyzing results) but cannot independently decide what experiments to run, validate assumptions, or interpret findings in context without expert oversight. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed products autonomously design and evaluate novel operational experiments without data; existing tools (e.g., simulation software, LLM copilots) assist analysts but don't independently perform this task in production. |
Develop business methods and procedures, including accounting systems, file systems, office systems, logistics systems, and production schedules.
31CI 25–38 · exposure 25 · augmentation 75 · importance 3.6/5 · click for rater detail
Develop business methods and procedures, including accounting systems, file systems, office systems, logistics systems, and production schedules.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Operations research and business process design remain concentrated in mid-to-large organizations with slower digitization of core decision-making. Pilot projects exist, but production-scale AI-driven procedure development is rare; most adoption remains advisory rather than autonomous. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Operations research and business analytics sectors show moderate AI adoption with pilots for process optimization and RPA, but full-scale AI-driven system design remains uncommon in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at drafting templates, suggesting process improvements based on data, generating documentation options, and accelerating literature review for best practices. These augment analyst productivity significantly while the human retains judgment over strategic fit and organizational feasibility. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly speed up drafting of procedures, generate scenario analyses, model logistics flows, and suggest system architectures, substantially boosting analyst productivity while humans retain final design authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can draft procedures and suggest system structures, these tasks require domain expertise, stakeholder consultation, and judgment about organizational context that current systems struggle with end-to-end. Automation of the full task—from requirements gathering through implementation and validation—falls short of the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | Designing business methods and procedures requires deep contextual understanding of organizational constraints, stakeholder negotiation, and judgment calls that current AI cannot fully replace end-to-end, though it can assist with drafting and analysis of sub-components. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Organizational adoption of new procedures typically requires sign-off from management, compliance review, and legal/audit validation. Liability for flawed accounting or logistics systems creates asymmetric error costs, and stakeholder trust in human-designed procedures remains high, creating friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No strict licensing requirement mandates a human perform this task, but organizational trust, accountability for system failures, and the need for tacit knowledge of internal politics create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | LLM-based assistance is inexpensive per inference, but the integration cost, validation overhead, and need for human expert review and refinement make the all-in cost comparable to or higher than direct human analysis and design. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can cheaply produce drafts of procedures, the human cost of validating, customizing, and integrating these into a specific organizational context remains substantial, keeping the overall cost comparable to human-led work. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No mature product reliably develops complete business procedures and systems from scratch in production. AI can assist with templates and document generation, but real-world procedure design demands iterative stakeholder engagement, testing, and customization that existing tools handle only partially. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools can generate templates, flowcharts, and draft procedures, but no deployed product reliably designs complete, validated organizational systems (accounting, logistics, production scheduling) without heavy human oversight and customization. |
Collaborate with senior managers and decision makers to identify and solve a variety of problems and to clarify management objectives.
16CI 7–25 · exposure 13 · augmentation 75 · importance 4.3/5 · click for rater detail
Collaborate with senior managers and decision makers to identify and solve a variety of problems and to clarify management objectives.
16| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Despite high digitization in finance and professional services, this task remains heavily dependent on interpersonal trust and executive preference for human analysts. Adoption of AI agents in strategic problem-solving roles remains minimal; most organizations use AI only as a subordinate analytical tool. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | While analytics tools are adopted quickly in professional services, the specific interpersonal collaboration with senior management remains largely untouched by AI agents in practice. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can substantially assist by rapidly synthesizing data, modeling scenarios, generating problem frameworks, and drafting recommendations for the analyst to present—enabling faster, more data-driven engagement with decision-makers while the human analyst retains ownership of stakeholder collaboration. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can help analysts prepare briefing materials, summarize data, and draft framing questions ahead of and during these management conversations, meaningfully boosting preparation productivity. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | This task requires real-time dialogue with senior stakeholders, understanding nuanced organizational context, and building consensus—activities that demand human judgment and executive presence. AI can assist in analyzing problems and outlining options but cannot authentically replace the collaborative relationship-building and decision-maker alignment that is core to this task. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires real-time interpersonal negotiation, trust-building, and organizational political sensitivity with senior stakeholders that current AI cannot autonomously conduct end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong organizational and relational barriers exist: stakeholders expect direct human-to-human interaction with analysts they trust, liability falls on human decision-makers not AI systems, and senior executives will not delegate strategic problem identification and objective clarification to autonomous agents. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Executive trust, accountability for decisions, and organizational norms strongly favor human-to-human interaction for sensitive objective-setting conversations. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The loaded cost of a senior manager's time spent in these collaborative sessions far exceeds current AI inference and integration costs, making AI cost-ineffective for replacing human presence in high-level executive meetings and consensus-building activities. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI product that replaces this collaborative human function, so cost comparison favors humans by default since AI cannot deliver the output alone. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI systems can generate analytical frameworks and draft communication, no deployed product reliably executes end-to-end stakeholder engagement, objective clarification, and problem identification at the level required for senior management decisions. Existing tools are narrow and typically used as support, not autonomous performers. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product substitutes for the human relationship-building and objective-clarification process with executives; AI at best supports background analysis. |
Collaborate with others in the organization to ensure successful implementation of chosen problem solutions.
9CI 5–13 · exposure 0 · augmentation 50 · importance 4.4/5 · click for rater detail
Collaborate with others in the organization to ensure successful implementation of chosen problem solutions.
9| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Organizations have shown minimal adoption of AI for core change management and stakeholder collaboration; these tasks remain deeply human-centric even in digitally mature sectors, with AI limited to supporting tools rather than replacing the analyst. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | While operations research and analytics functions increasingly use AI tools, the collaborative implementation aspect of the role sees minimal direct AI displacement or adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by generating communication templates, summarizing stakeholder feedback, tracking implementation metrics, and managing project documentation, improving an analyst's productivity on the administrative and analytical components of implementation management. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help by drafting communications, summarizing project status, or tracking implementation tasks, but does not fundamentally transform the collaborative core of the task. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Collaboration and change management require real-time human negotiation, relationship-building, and organizational politics that current AI cannot reliably conduct end-to-end. While AI can assist with communication drafts or documentation, ensuring successful adoption across stakeholders fundamentally depends on human judgment and persuasion. |
| Task automatability | claude-sonnet-5 | 1/5 | This task is inherently interpersonal and organizational, requiring relationship-building, negotiation, and stakeholder management that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Organizational implementation inherently requires human accountability, trust-building with stakeholders, and decision-making authority that cannot be delegated to machines. Regulatory and cultural norms strongly expect humans to own change management and stakeholder communication. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement, but organizational trust, accountability, and need for human buy-in create substantial practical friction against automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Deploying AI agents for organizational collaboration would require significant oversight, human validation of stakeholder management decisions, and error recovery, making the total cost exceed hiring an analyst to perform the task directly. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the human relational work needed for successful implementation, so there is no meaningful AI cost basis to compare against human labor for this task itself. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system can independently manage organizational change, navigate stakeholder resistance, or drive solution implementation across departments. AI tools exist for meeting transcription or project tracking, but none demonstrate autonomous successful implementation coordination in production. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product manages cross-functional collaboration and implementation coordination autonomously; this remains a human-driven organizational process. |
Related occupations — Computer & Mathematical
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.