Natural Sciences Managers
11-9121.00Median wage $167,220/yr108,690 employed (US)Rank #442 of 923 scored · top 48% by substitution
Plan, direct, or coordinate activities in such fields as life sciences, physical sciences, mathematics, statistics, and research and development in these fields.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution29
Exposure27
Augmentation69
Substitution — the headline: capability discounted by cost, barriers and adoption.
Exposure — technical capability alone, regardless of whether anyone deploys it.
Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.
Tasks on the substitution scale
16 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
0%
Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.
Why this score
The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.
Task automatabilityw 35%28
panel mean rating 2.1/5 → substitution pressure 28/100
Technical feasibility todayw 20%26
panel mean rating 2.0/5 → substitution pressure 26/100
Cost vs. human wagew 15%29
panel mean rating 2.2/5 → substitution pressure 29/100
Adoption barriersw 20%inverted — strong barriers lower the score31
panel mean rating 3.8/5 (barrier strength) → substitution pressure 31/100
Sector adoption velocityw 10%31
panel mean rating 2.3/5 → substitution pressure 31/100
Task breakdown (16 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.
Prepare project proposals.
61CI 54–67 · exposure 58 · augmentation 88 · importance 3.9/5 · click for rater detail
Prepare project proposals.
61| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Research universities and government labs are piloting AI-assisted proposal drafting, but adoption remains spotty and largely experimental. Financial and professional services firms are further along; natural sciences academia is slower to integrate AI into proposal workflows. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Research and R&D-adjacent sectors are moderately adopting AI writing tools for grants and proposals, but scientific management remains a mixed-digitization, judgment-heavy field with uneven uptake. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly accelerates literature synthesis, outline generation, and first-draft production, allowing managers to focus on strategic framing and high-stakes revision. This assistant role is already proving valuable in many labs and research centers. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially accelerates drafting, editing, formatting, and budget calculations for proposals while the manager retains responsibility for scientific content and final decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can draft proposal sections, compile background research, and generate initial outlines with significant time savings, but preparing competitive proposals typically requires deep domain expertise, stakeholder alignment, and strategic positioning that managers must provide. Roughly half of the mechanical work can be automated, but the higher-value conceptual and persuasive elements remain manager-dependent. |
| Task automatability | claude-sonnet-5 | 4/5 | Drafting proposal text, budgets, and structure from inputs is well within current LLM capability, though final scientific judgment and strategic framing still need human review, giving substantial but not complete time savings. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Proposals often require manager sign-off and institutional accountability, and funding agencies typically expect human judgment on scientific merit and feasibility. Organizational culture often insists on managerial review, creating moderate friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI-assisted drafting, though funding agencies and institutional review may expect named PI accountability and scientific credibility tied to a human author. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference and integration costs for proposal drafting are minimal (dollars per proposal), while a senior manager's time is worth $100+/hour. Even accounting for oversight and refinement, the cost ratio heavily favors AI assistance. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating drafts, boilerplate sections, and budget justifications via AI is far cheaper than manager hours spent on first drafts, though oversight and domain expertise still add cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed tools like GPT-4 and Claude can generate proposal drafts, and some organizations use specialized proposal software with AI assistance. However, these products struggle with nuanced scientific merit assessment, institution-specific strategic fit, and regulatory compliance requirements, limiting reliable end-to-end deployment. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI writing assistants and grant-drafting tools are used in research settings today, but they require heavy human editing for technical accuracy, funder-specific requirements, and scientific novelty claims. |
Prepare and administer budgets, approve and review expenditures, and prepare financial reports.
60CI 50–70 · exposure 62 · augmentation 75 · importance 3.5/5 · click for rater detail
Prepare and administer budgets, approve and review expenditures, and prepare financial reports.
60| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large and mid-sized organizations in finance, professional services, and information sectors are rapidly adopting AI-driven financial automation and analytics platforms. Smaller firms and public sector entities lag, but overall adoption momentum is strong and measurable in production deployments. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Scientific and R&D-focused organizations have moderate digitization and are adopting AI-assisted financial tools, but adoption for actual budget approval authority remains slower than in finance-native sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI dramatically assists managers by automating data collection, variance analysis, forecasting, and report drafting, allowing the manager to focus on strategic allocation and exception handling. This augmentation substantially raises a manager's throughput and analytical depth while keeping them in the governance loop. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools substantially speed up variance analysis, forecasting, and report drafting, letting managers focus on judgment calls around approvals and strategic allocation. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Budget preparation, expenditure approval workflows, and financial report generation are highly structured, data-driven processes that can be largely automated with modern accounting software and AI-assisted analytics. However, strategic budget allocation decisions and exception-based review typically still require human judgment, preventing a perfect 5. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft budgets, analyze spending patterns, and generate financial reports from structured data with significant time savings, but approval decisions require managerial judgment and accountability that current systems cannot fully replace. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Financial controls, audit trails, and regulatory compliance (SOX, GAAP, institutional policy) require human sign-off and oversight, creating material friction. Most organizations mandate that a qualified human manager review and approve final budgets and reports, though AI can handle the heavy lifting of data aggregation and preliminary analysis. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Expenditure approval typically requires designated managerial authority and accountability within organizational controls, creating moderate friction even though report drafting itself has few barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | The cost of cloud-based accounting software, AI analytics tools, and integration overhead is substantially lower than the fully-loaded salary of a manager performing these routine administrative tasks (typically $80K–$150K+). Once amortized, AI-assisted solutions deliver at least a 5–10× cost advantage for the transactional components. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-augmented financial tools reduce time spent on report generation and analysis, but licensing, integration, and required human oversight for approvals keep costs roughly comparable to a manager's marginal time on this task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature ERP and accounting platforms (SAP, Oracle, NetSuite) combined with AI-powered expense management and financial reporting tools (like Workiva, BlackLine) are widely deployed in production at scale across enterprises. These systems handle budget administration, expenditure tracking, and report generation reliably, though some custom workflows and edge cases may require manual intervention. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Financial software with AI-assisted forecasting, reporting, and anomaly detection is widely deployed in production, but expenditure approval and budget administration still rely on human sign-off integrated with these tools. |
Review project activities and prepare and review research, testing, or operational reports.
37CI 25–50 · exposure 38 · augmentation 75 · importance 4.1/5 · click for rater detail
Review project activities and prepare and review research, testing, or operational reports.
37| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Life sciences, pharmaceutical, and government R&D sectors move cautiously on automating quality review; pilots of AI-assisted report drafting exist, but few organizations have deployed AI to replace or substantially reduce manager review workload at scale. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | R&D-intensive sectors are adopting AI writing and analysis tools at a moderate pace, with pilots common but full integration into managerial reporting workflows still emerging. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at extracting data, summarizing findings, flagging inconsistencies, and producing draft reports that managers can then review and refine—significantly raising their productivity while they retain oversight and accountability. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids drafting, summarizing, and checking reports for consistency and clarity, meaningfully boosting manager productivity while they retain final review and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with basic report generation and editing, this task requires human judgment to evaluate project activities, validate research findings, and make strategic recommendations—core elements that demand contextual understanding and accountability that current systems cannot reliably provide end-to-end. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft report sections, summarize data, and flag inconsistencies, but substantive scientific review of project activities requires domain judgment and accountability that current systems cannot fully replicate end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Scientific integrity and regulatory compliance (e.g., FDA, EPA, internal audit standards) often require a human manager to sign off on reports, and organizational norms expect a qualified human to bear responsibility for project oversight and accuracy. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement for report review itself, but managerial accountability, liability for scientific/operational accuracy, and organizational sign-off norms create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted drafting tools are cheap, but integrating them into a manager's review workflow requires significant setup and human oversight, and the cost of errors in report validation is high, making the all-in cost comparable to or exceeding human review. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI drafting assistance reduces time spent on report preparation, but human oversight for accuracy, compliance, and scientific validity remains necessary, keeping total cost roughly comparable to a partially augmented human workflow. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some AI products exist for report drafting and data summarization, but they lack the domain expertise and evaluative rigor needed to review scientific or operational reports for accuracy, significance, and appropriateness—tasks where errors have high organizational cost. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | LLM-based tools are deployed for drafting and summarizing technical reports, but reliable, autonomous review of research/operational content at managerial accountability level is not yet standard in production. |
Make presentations at professional meetings to further knowledge in the field.
31CI 25–37 · exposure 25 · augmentation 75 · importance 3.2/5 · click for rater detail
Make presentations at professional meetings to further knowledge in the field.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for live presentation delivery in professional scientific settings is slow; most organizations still expect human domain experts to present at conferences and professional meetings. AI remains mostly an assistive tool for preparation, not a replacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Scientific/professional conference culture remains slow to adopt AI-driven presentation delivery, though AI tools for slide creation are increasingly used in prep. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists in preparing presentations by generating slides, organizing research summaries, and suggesting content structure, which can substantially improve manager productivity in presentation design while the human remains the essential presenter. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools substantially help with drafting slides, summarizing research, generating visuals, and rehearsing talking points, meaningfully boosting preparation efficiency while the human still presents. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can generate presentation slides and draft talking points at scale, but delivering a live professional presentation requires real-time audience engagement, questions, and credibility that is difficult to automate end-to-end. A human presenter remains essential for authentic knowledge dissemination. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help draft slides and scripts, but delivering a live presentation at a professional meeting, engaging with audience questions, and representing original scientific contributions requires human presence and judgment that cannot be end-to-end automated today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: professional credibility, organizational reputation, and audience expectations typically require a human expert to be physically present and answerable for the knowledge presented. Legal and reputational liability falls on the human presenter, not an AI system. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement bars AI assistance, but professional norms expect the credentialed scientist/manager to personally present and defend their work, creating moderate social and credibility barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-assisted slide generation and content drafting costs are minimal compared to the manager's time investment in preparation, making AI assistance much cheaper than the human cost of manual presentation creation. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Since the core act of presenting and networking cannot be delegated to AI, there is no meaningful AI cost substitute for the full task, though drafting support is cheap. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Products exist to draft presentations and suggest content, but no deployed system reliably replaces the live presenter role in professional settings. Current AI falls short of the production-grade reliability needed for authentic knowledge sharing at conferences. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed product autonomously represents a manager at a conference; AI presentation tools (slide generators, speech synthesis) exist but are assistive, not substitutive for the full task. |
Develop client relationships and communicate with clients to explain proposals, present research findings, establish specifications, or discuss project status.
29CI 25–32 · exposure 25 · augmentation 75 · importance 4.0/5 · click for rater detail
Develop client relationships and communicate with clients to explain proposals, present research findings, establish specifications, or discuss project status.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Research organizations and consulting firms have adopted AI for drafting and organization, but actual client-facing communication and relationship stewardship remain largely human-led. Adoption is in the augmentation phase rather than replacement, even in digitally mature sectors. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Scientific and R&D-adjacent management sectors show moderate AI adoption for communication support tools, but full automation of client-facing relationship management remains uncommon in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at drafting client communications, organizing research data for presentation, and flagging key findings to highlight. These tools measurably increase a manager's ability to prepare, personalize, and iterate on client interactions while the human retains full relationship ownership and decision authority. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly aid preparation of presentations, proposal drafts, and research summaries, improving efficiency while the manager retains direct client interaction and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can draft proposal text and summarize research findings, the task requires building and maintaining relationships, reading contextual social cues, and negotiating specifications—elements that demand human judgment and trust. End-to-end automation would fall short of the ≥50% time-saving threshold at equal quality because relationship continuity and client confidence are core outputs. |
| Task automatability | claude-sonnet-5 | 2/5 | Client relationship-building requires trust, negotiation, and adaptive interpersonal judgment that current AI cannot autonomously execute end-to-end, though AI can draft communications or summarize findings for human delivery. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Clients typically expect to interact with and authorize specifications from a human expert; trust and accountability in research relationships create implicit human-contact requirements. Liability and reputational risk if AI misrepresents findings or loses track of client commitments are substantial organizational barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement exists, but strong organizational and client expectations for human-to-human trust, accountability, and rapport create moderate friction against automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can lower the cost of certain communication scaffolding (draft emails, summaries), but the manager's time remains the binding constraint since client relationship continuity and authority cannot be delegated. Total cost remains dominated by human oversight and the irreplaceability of the human-client interaction itself. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Because a human manager must remain the primary relationship interface, AI mainly supplements rather than replaces labor, so cost savings are limited to partial task components like drafting or summarization. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed products can assist with report generation and talking points, but no mainstream system reliably handles the full spectrum of client relationship management—from reading emotional context to closing specifications on complex research projects. Pilots exist but production deployment remains narrow and heavily supervised. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some products assist with meeting summaries, email drafting, and presentation generation, but no deployed system independently manages client relationships or negotiates specifications reliably. |
Confer with scientists, engineers, regulators, or others to plan or review projects or to provide technical assistance.
26CI 25–28 · exposure 25 · augmentation 63 · importance 4.0/5 · click for rater detail
Confer with scientists, engineers, regulators, or others to plan or review projects or to provide technical assistance.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Research and regulated sectors (pharma, energy, aerospace) adopt AI slowly for management and oversight roles due to compliance requirements. Adoption remains in the pilot stage; routine AI delegation of technical conferencing is not yet widespread in production. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | R&D-intensive sectors (pharma, tech, engineering) are adopting AI tools for meeting prep, document review, and technical assistance, but the core conferring/coordination role remains largely unautomated in practice. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by preparing technical briefing materials, summarizing prior discussions, and highlighting key regulatory constraints before a conference, moderately raising manager productivity. However, the human must still lead the actual interaction and make judgments about scientific and project feasibility. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by summarizing technical documents, drafting talking points, analyzing project data, and flagging regulatory issues in advance, substantially improving preparation quality and efficiency for these meetings. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can help draft summaries and prepare materials for technical discussions, the core task requires real-time judgment about project feasibility, scientific trade-offs, and interpersonal problem-solving that current systems cannot reliably perform end-to-end. Conference planning and stakeholder coordination remain heavily dependent on human discretion and relationship management. |
| Task automatability | claude-sonnet-5 | 2/5 | This task centers on live interpersonal conferring, negotiation, and technical judgment across stakeholders with differing interests, which current AI cannot conduct end-to-end; AI can prep materials but not replace the interaction itself. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Regulatory sign-off, professional liability for technical guidance, and stakeholder trust are substantial barriers. Many regulated industries require a qualified human manager to formally review and approve projects, and external regulators often demand direct communication with authorized personnel. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Regulatory interactions and technical sign-off often require accountable, credentialed human judgment, and organizational/legal expectations mean a manager must personally represent and negotiate on behalf of the organization. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI assistance for background research and document prep is cheap, but the human manager must still conduct the actual conferences and provide technical oversight. The savings are partial and do not offset the manager's loaded salary for the core conferencing work. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | A manager's expertise, credibility, and relationship management are central to this task; AI can cut prep time but cannot substitute the human interaction, so cost savings are limited to partial support functions. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably conducts multi-stakeholder technical conferences, evaluates regulatory alignment, or makes binding project decisions. AI can support document preparation and agenda drafting, but independent execution of the conferencing and advisory task is not demonstrable in production systems. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed product autonomously confers with scientists/regulators to plan or review projects; AI assistants exist for meeting summarization or technical Q&A support but not for conducting the substantive review/negotiation role. |
Provide for stewardship of plant or animal resources or habitats, studying land use, monitoring animal populations, or providing shelter, resources, or medical treatment for animals.
25CI 25–25 · exposure 25 · augmentation 63 · importance 4.2/5 · click for rater detail
Provide for stewardship of plant or animal resources or habitats, studying land use, monitoring animal populations, or providing shelter, resources, or medical treatment for animals.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption is slow in practice; conservation and natural resources management remain traditional sectors with limited digital transformation. While large institutions may pilot monitoring tools, widespread production deployment of AI agents for stewardship is rare. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Environmental and natural resource management sectors are slow adopters of AI compared to information/finance industries, with pilots for monitoring tools but limited broader deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI meaningfully assists managers through automated population monitoring via image analysis, habitat data synthesis, and resource planning tools, improving decision-making without replacing human stewardship. However, augmentation is limited to information tasks rather than the full stewardship workflow. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids sub-tasks like analyzing population data, satellite imagery for land use, and predictive habitat modeling, meaningfully boosting manager productivity even though humans remain essential for stewardship actions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with population monitoring (e.g., camera trap image analysis, data synthesis), the task inherently requires field work, hands-on animal care, medical decisions, and adaptive habitat stewardship that demand human judgment and physical presence. AI cannot replicate the full end-to-end task with 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | This task involves fieldwork, hands-on animal care, land assessment, and managerial decision-making that require physical presence and judgment AI cannot replicate end-to-end; only data analysis sub-components are automatable. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: many stewardship decisions require licensed veterinary or environmental expertise, liability for animal welfare and habitat decisions is substantial, and regulatory frameworks often mandate qualified human professionals conduct or oversee conservation and medical work. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Animal welfare, veterinary care, and land management often require licensed professionals, regulatory compliance (e.g., wildlife protection laws), and direct physical intervention, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted monitoring tools carry infrastructure and integration costs, but they primarily augment rather than replace the skilled labor of biologists and veterinarians who command high wages. Full end-to-end cost savings remain marginal compared to human labor. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools can cheaply process monitoring data, but the physical stewardship, medical treatment, and on-site management components still require costly human labor, keeping overall cost comparable or higher than pure AI substitution. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Computer vision systems can identify animals in images and some monitoring tools exist, but deployed products do not reliably handle the full spectrum of stewardship decisions, medical assessments, and resource allocation that this task encompasses. Production deployment remains limited to narrow sub-tasks. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products exist for wildlife monitoring (camera trap AI, satellite land-use analysis) but no deployed system performs habitat stewardship or animal medical care as an integrated task in production. |
Develop or implement policies, standards, or procedures for the architectural, scientific, or technical work performed to ensure regulatory compliance or operations enhancement.
25CI 25–25 · exposure 25 · augmentation 63 · importance 3.7/5 · click for rater detail
Develop or implement policies, standards, or procedures for the architectural, scientific, or technical work performed to ensure regulatory compliance or operations enhancement.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for policy generation is nascent; most organizations rely on human managers, consultants, and legal teams. Pilots exist in some tech and finance firms, but production deployment of AI-authored compliance policies at scale is extremely rare due to liability and regulatory risk aversion. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Scientific and technical management sectors adopt AI tools unevenly, with policy/compliance functions typically lagging behind more digitized business functions like finance or customer service. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by analyzing regulatory requirements, summarizing existing standards, drafting template language, and identifying gaps—reducing research and initial drafting burden. However, augmentation is bounded by the need for human judgment on scope, stakeholder impact, and final accountability. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by drafting policy templates, summarizing regulations, and flagging compliance gaps, substantially speeding up the manager's initial drafting and research work. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | This task requires deep organizational knowledge, stakeholder negotiation, and judgment about regulatory landscapes and technical trade-offs. While AI can draft policy language or summarize regulations, end-to-end development and implementation of compliance policies with equivalent quality requires sustained human decision-making and accountability. |
| Task automatability | claude-sonnet-5 | 2/5 | Drafting policy language can be AI-assisted, but developing substantive, context-specific standards for scientific/technical operations requires domain judgment, regulatory interpretation, and organizational knowledge that current AI cannot reliably supply end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: regulatory compliance policies often require authorized human sign-off (managers, legal, compliance officers), liability falls on named individuals, and many sectors (healthcare, finance, pharmaceuticals) legally require credentialed professionals to certify policy adequacy. Organizational friction around policy changes also slows substitution. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Regulatory compliance policies often require sign-off by qualified, accountable managers or licensed professionals, and liability for compliance failures creates strong resistance to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Policy development requires domain expertise, legal review, stakeholder alignment, and accountability oversight that remains expensive. AI can reduce some research and drafting costs, but the total cost (tool + human oversight + legal validation) is still comparable to or higher than direct human work. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply generate draft text, but the human cost of review, validation against regulations, and organizational buy-in still dominates the effective cost, keeping overall savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products reliably generate organization-wide policies and standards that meet regulatory and operational requirements without substantial human review and legal/technical refinement. AI tools can assist with research and drafting, but production-ready policy systems remain rare. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Generic AI writing tools exist for policy drafting, but no deployed product autonomously develops and implements regulatory compliance standards for scientific operations in production settings. |
Design or coordinate successive phases of problem analysis, solution proposals, or testing.
25CI 20–30 · exposure 20 · augmentation 63 · importance 4.2/5 · click for rater detail
Design or coordinate successive phases of problem analysis, solution proposals, or testing.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While information sectors are adopting AI assistants, the core management and coordination function in R&D and scientific organizations remains human-dominated; pilot projects exist but production replacement is rare and limited to junior planning tasks. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | R&D and scientific management sectors adopt AI tools for specific subtasks (literature review, data analysis) but broad managerial coordination workflows show slow, cautious adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by generating alternative solution proposals, summarizing problem statements, or flagging phase dependencies, raising manager productivity in research planning and documentation. However, the creative and judgment-intensive aspects of designing scientific workflows remain human-centric. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can significantly assist by drafting research plans, synthesizing literature, flagging risks, and organizing testing phases, meaningfully boosting manager productivity while they retain decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in generating solution proposals and organizing data, the task requires continuous judgment about problem framing, phase dependencies, risk assessment, and adaptation—activities that demand human oversight. AI cannot reliably design or coordinate multi-phase workflows end-to-end without substantial human intervention. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires senior scientific judgment, cross-team coordination, and iterative decision-making informed by organizational context that current AI cannot autonomously perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Leadership and design roles typically carry implicit or explicit accountability requirements; liability and organizational authority structures strongly favor human managers signing off on phase transitions and problem scope. Regulatory and accountability expectations in scientific/technical work create friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically bars AI, but organizational accountability, scientific credibility, and managerial authority create substantial friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI tools for workflow design and coordination require significant setup, validation, and human oversight, making them cost-comparable to or more expensive than direct human management for complex, high-stakes problem sequences. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply assist with drafting plans or summarizing data, but the core coordination and judgment work still requires expensive human oversight, making all-in cost comparable to or higher than a manager's time saved. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably orchestrates full problem-analysis-to-testing cycles in production. Project management and workflow coordination tools exist, but AI contributions remain limited to proposal generation or documentation; humans retain design and coordination authority in actual scientific and technical programs. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product manages multi-phase scientific problem-solving and testing coordination for an organization; this remains a human managerial function. |
Develop innovative technology or train staff for its implementation.
25CI 20–30 · exposure 20 · augmentation 75 · importance 3.3/5 · click for rater detail
Develop innovative technology or train staff for its implementation.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While information-sector firms use AI for research support, actual displacement of natural sciences managers by AI in innovation and training roles is minimal. Most organizations maintain human leadership for R&D strategy and staff development; pilot AI tools remain assistive rather than decision-making. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | R&D and scientific management sectors are cautious adopters of AI for strategic and people-management functions, with pilots more common in narrow technical subtasks than in holistic management roles. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI strongly augments these tasks: systems can rapidly synthesize literature, generate training module drafts, flag implementation risks, and prototype curricula—freeing managers to focus on strategic innovation direction and personalized coaching. Managers' productivity and decision quality improve substantially when AI handles information aggregation and content scaffolding. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist in generating training content, summarizing new technologies, drafting implementation plans, and supporting brainstorming, significantly boosting a manager's productivity while they retain decision-making control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with technology research and identify implementation strategies, the core task requires creative innovation, strategic vision, and deep contextual judgment about organizational fit and staff capabilities. Current AI cannot reliably generate fundamentally novel technological directions or customize training programs at the depth required without substantial human oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | This task combines high-level innovation, R&D strategy, and hands-on staff training, none of which current AI can execute end-to-end; AI can assist ideation and training material creation but cannot independently develop novel technology or manage its organizational rollout. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Developing innovative technology and training staff requires licensed expertise, regulatory sign-off in some fields, and organizational accountability for outcomes. Stakeholders heavily prefer human leadership on strategic R&D and training decisions; legal and reputational liability for failed innovation or poor training falls on human managers who must retain decision authority. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing mandate requires a human specifically, but organizational trust, accountability for technical innovation, and the need for domain expertise and leadership create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce labor on research aggregation and training content drafting, but the high-value work—innovation conception and program customization—still requires expensive expert managers. Integration, validation, and human oversight costs are substantial relative to the wage savings on routine research tasks. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools can lower costs for drafting training materials or literature synthesis, but the core innovation and leadership work still requires expensive expert oversight, keeping overall cost comparable to or only slightly cheaper than human-led effort. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI tools exist for research synthesis, training content generation, and implementation planning, but no deployed product reliably handles the full task of developing innovative technology or designing staff training programs end-to-end. These tasks demand understanding of cutting-edge science, organizational context, and pedagogy—areas where current systems show material limitations. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously develops innovative scientific/technical solutions or manages staff training programs at a managerial level; this remains a human-led, judgment-intensive activity. |
Recruit personnel or oversee the development or maintenance of staff competence.
24CI 16–32 · exposure 22 · augmentation 75 · importance 3.6/5 · click for rater detail
Recruit personnel or oversee the development or maintenance of staff competence.
24| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Mid-market and large tech/finance firms are adopting AI recruiting tools and learning analytics, but adoption remains pilot-heavy in many sectors and remains conservative in regulated industries and smaller firms. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | HR tech adoption in scientific/technical management sectors is moderate, with AI used for screening but broader staff development oversight remains largely manual and slow to change. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially augments recruiting managers through candidate matching, bias detection, skill gap analysis, and personalized learning recommendations, meaningfully raising productivity while humans retain final authority on hiring and development strategy. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can meaningfully assist with candidate sourcing, resume parsing, training material generation, and competency tracking, boosting manager productivity while decisions remain human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist with candidate screening, resume parsing, and identifying competence gaps, but the core task—recruiting and staff development—requires human judgment on cultural fit, nuanced interviewing, and personalized development plans. End-to-end automation would need to replace final hiring decisions and ongoing mentorship, which remain predominantly human-driven. |
| Task automatability | claude-sonnet-5 | 1/5 | Recruiting and developing staff competence requires nuanced human judgment about fit, motivation, and interpersonal dynamics that current AI cannot perform end-to-end.','not applicable' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant legal and organizational barriers exist: employment law requires human accountability in hiring, discrimination risk demands human review, and organizational culture heavily privileges human-led recruiting and mentorship. Liability asymmetry favors human decision-makers. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Hiring decisions carry legal, ethical, and liability implications (e.g., anti-discrimination law) that generally require human accountability and sign-off, creating strong barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for recruitment and competence tracking have moderate upfront costs and require substantial human oversight (legal review, relationship management, final decisions). The loaded cost of a recruiting manager or L&D specialist remains lower than the integrated cost of AI systems plus required human oversight. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI screening tools reduce some costs, the managerial judgment, interviewing, and mentoring components still require substantial human time, keeping overall costs comparable to human-led processes. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | HR platforms with AI-assisted resume screening and learning analytics exist in production, but they support rather than replace recruiting and staff development decisions. Error rates in bias, fit assessment, and development prioritization remain material, limiting full autonomy. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI products exist for resume screening and skills-gap analysis, but no deployed system autonomously recruits or manages staff development at scale in production. |
Determine scientific or technical goals within broad outlines provided by top management and make detailed plans to accomplish these goals.
23CI 20–25 · exposure 20 · augmentation 63 · importance 3.9/5 · click for rater detail
Determine scientific or technical goals within broad outlines provided by top management and make detailed plans to accomplish these goals.
23| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption in scientific and technical organizations remains low; while some firms pilot AI-assisted planning tools, goal determination remains a core human leadership function with slow displacement in practice. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | R&D and scientific management sectors are adopting AI tools for literature review and data analysis, but strategic goal-setting and planning functions show minimal actual AI-driven displacement. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by generating scenario analyses, consolidating technical constraints, or drafting plan details, enabling managers to focus on judgment and stakeholder alignment, though the core goal-setting remains human-led. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by synthesizing research trends, generating draft project plans, modeling resource scenarios, and summarizing technical literature to inform the manager's decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can draft detailed plans and analyze goals, determining scientific/technical goals requires strategic judgment, stakeholder alignment, and contextual understanding that current systems struggle with at equal quality. AI cannot reliably handle the feedback loops and refinement needed to meet top management's broad outlines in a complex organization. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires synthesizing organizational strategy, technical feasibility, personnel capacity, and stakeholder priorities into concrete goals and plans—AI can assist drafting but cannot autonomously set legitimate technical direction for an organization today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong organizational and accountability barriers exist: goal-setting is typically a leadership prerogative, requires sign-off from top management, and carries reputational and financial risk if misaligned. Liability for failed strategic direction discourages full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Setting technical direction and resource allocation typically requires organizational authority, accountability, and credentialed expertise; management responsibility cannot be legally or practically delegated to an AI system. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI systems (planning/reasoning tools, consulting agents) still require significant human oversight, validation, and iteration, making the total cost comparable to or exceeding a senior manager's time for this complex strategic task. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Human natural sciences managers command high salaries, but the task's judgment-intensive, high-stakes nature still requires substantial human oversight, so AI cannot yet substitute for most of the cost while maintaining quality. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products reliably perform goal-setting and strategic planning for R&D organizations end-to-end; some tools assist with plan generation or visualization, but the core task of determining goals aligned with organizational strategy and technical constraints lacks production-level solutions. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently determines scientific/technical goals and detailed execution plans for a research organization; this remains a human managerial judgment task with only AI-assisted planning tools available. |
Advise or assist in obtaining patents or meeting other legal requirements.
23CI 20–25 · exposure 25 · augmentation 75 · importance 3.0/5 · click for rater detail
Advise or assist in obtaining patents or meeting other legal requirements.
23| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Legal and IP functions adopt AI tools slowly relative to other professional services; firms use AI-assisted search and document review but retain human attorneys for strategy and sign-off. Production-level autonomous patent advising remains rare even in digitized professional sectors. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Legal and IP services are moderately digitized but adoption of AI for actual patent prosecution and legal advice remains cautious due to liability concerns, with pilots more common than full production use. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI tools meaningfully assist with prior-art searching, claim analysis, regulatory checklist generation, and document drafting, significantly raising a patent attorney's or manager's productivity on routine compliance and search tasks while the human retains decision authority. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly assist with prior art searches, drafting technical descriptions, and identifying compliance requirements, meaningfully speeding up the work of managers and legal teams. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Patent advice and legal requirement compliance require nuanced judgment about prior art, claim scope, and jurisdiction-specific regulations. While AI can assist with prior-art searches and document organization, the core advisory and strategic decisions that define this task require human legal expertise and cannot be fully automated to meet the 50%-time-saving threshold without significant attorney oversight. |
| Task automatability | claude-sonnet-5 | 2/5 | Patent drafting and legal compliance require specialized judgment, novelty assessment, and legal accountability that current AI cannot fully replicate end-to-end, though AI can assist with drafting and prior-art searches. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Patents and legal compliance are heavily regulated; only licensed attorneys can provide binding legal advice and sign-off on patent filings in most jurisdictions. Liability for incorrect IP strategy and regulatory requirements for authorized representation create hard barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Patent applications typically require licensed patent attorneys/agents to prosecute before patent offices, and legal sign-off creates strong regulatory and liability barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Patent counsel and Natural Sciences Managers commanding professional salaries (often $100k+) perform this work; AI tools reduce some search and drafting costs but cannot eliminate the need for skilled human review and strategy. The all-in cost of AI plus required attorney supervision remains comparable to or higher than the human cost. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can reduce some research and drafting costs, but patent attorneys and technical experts still must review and certify filings, so overall cost savings are modest given liability and accuracy requirements. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Patent management AI tools exist for prior-art searching and document assembly, but no deployed system reliably performs independent patent strategy advice or full legal compliance assessment. Current products operate at research or narrow-scope levels; real-world patent practice requires attorney judgment that hasn't been reliably automated. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI legal-tech tools (e.g., patent drafting assistants, prior art search tools) exist and are used in production but require substantial attorney review and are narrow in scope, not fully reliable standalone solutions. |
Plan or direct research, development, or production activities.
19CI 16–21 · exposure 20 · augmentation 63 · importance 4.2/5 · click for rater detail
Plan or direct research, development, or production activities.
19| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While information-sector firms experiment with AI-assisted planning tools, actual autonomous or semi-autonomous AI direction of R&D is rare in production. Adoption remains in pilot phases; established sector culture prioritizes senior scientists and managers in strategic roles, limiting velocity of displacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | R&D-intensive science and production sectors adopt AI tools for specific technical tasks but strategic management functions see much slower and shallower AI integration. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist with data-driven parts of planning: generating resource allocation scenarios, identifying bottlenecks in timelines, drafting project schedules, and summarizing research progress. These augmentation benefits are real but incremental; the core direction and strategic decision-making remain human-led. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by synthesizing research data, forecasting timelines, drafting reports, and modeling scenarios, boosting a manager's planning and decision efficiency substantially. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Planning and directing research/development activities requires synthesizing complex information, making strategic trade-offs, and adapting to unforeseen scientific and team dynamics. While AI can assist in scheduling, resource allocation modeling, and draft plans, end-to-end direction with 50% time savings at equal quality requires human judgment, stakeholder negotiation, and contextual decision-making that current systems cannot reliably replace. |
| Task automatability | claude-sonnet-5 | 2/5 | This task requires strategic judgment, resource allocation, stakeholder negotiation, and accountability for scientific outcomes that current AI cannot autonomously perform end-to-end.dummy |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant organizational and accountability barriers exist: research and development direction is typically a licensed or senior management role carrying legal liability for budget, safety, IP, and team performance. Stakeholders (executives, investors, scientists) expect human accountability and judgment; regulatory frameworks in pharma, biotech, and engineering require human sign-off on research direction. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Senior technical/managerial roles carry significant organizational, legal, and accountability requirements (budget authority, safety compliance, personnel decisions) that require human sign-off and expertise. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI systems capable of even partial direction (scheduling, resource modeling) still require substantial human oversight, validation, and final decision-making. The all-in cost of current AI assistance (tool development, integration, human review) typically exceeds the cost savings of automating fragments of planning work. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | Given AI cannot perform the managerial planning/directing function itself, there is no viable AI-only cost comparison; human managers remain necessary and costlier alternatives don't exist. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products today can autonomously plan and direct R&D or production activities at scale with reliable quality. Limited tools exist for scheduling or resource optimization, but real-world research direction involves human accountability, team leadership, and pivoting based on results—functions that remain research-stage or narrowly scoped in practice. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product plans or directs R&D or production activities for natural sciences organizations; AI is used only for narrow sub-tasks like literature review or data analysis. |
Conduct own research in field of expertise.
15CI 0–30 · exposure 13 · augmentation 63 · importance 3.5/5 · click for rater detail
Conduct own research in field of expertise.
15| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Despite high digitization in research institutions, adoption of AI to autonomously conduct novel research remains negligible. Sector adoption favors human researchers supported by computational tools rather than AI-driven replacement, reflecting fundamental barriers to autonomy in knowledge creation. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Scientific research organizations are adopting AI tools for specific research tasks (data analysis, drafting) but full autonomous research replacement remains rare and experimental, lagging faster-adopting sectors like finance or software. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI tools meaningfully assist researchers by automating literature searches, summarizing papers, analyzing data, and drafting sections, raising productivity on execution tasks. However, augmentation is limited to supporting work; the core task of conceiving and designing original research remains fundamentally human. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially assists literature review, data analysis, hypothesis brainstorming, coding, and manuscript drafting, meaningfully boosting researcher productivity while the scientist retains overall direction and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Conducting novel research in a field of expertise fundamentally requires creative hypothesis generation, experimental design, and original intellectual contribution that current AI systems cannot perform end-to-end. While AI can assist with literature review, data analysis, and manuscript preparation, the core act of conducting novel research—formulating research questions, designing experiments, and interpreting results in context—requires human domain expertise and originality. |
| Task automatability | claude-sonnet-5 | 2/5 | Original scientific research requires hypothesis generation, experimental design, physical experimentation, and novel insight that current AI cannot perform end-to-end reliably; AI can accelerate sub-steps like literature review or data analysis but not replace the full research process. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Research publication, funding, and institutional advancement require human authorship, institutional affiliation, and professional accountability. Regulatory and ethical frameworks (IRB approval, research ethics) explicitly require human researchers as responsible parties, creating hard legal and professional barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement to 'do research' per se, but institutional credibility, peer review, funding accountability, and domain expertise create substantial organizational friction against full AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | A natural sciences manager conducting research draws a significant salary ($150k+), and the intellectual value of novel research output far exceeds the cost of AI inference tools. AI cannot generate the equivalent research output, making any cost comparison misleading; this is a non-substitutable task. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools reduce some costs (literature search, analysis) but the manager's expertise, lab work, and judgment remain human-driven, so overall cost savings are modest rather than order-of-magnitude. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product reliably conducts independent research end-to-end in a scientific field. AI tools exist for supporting research components (literature retrieval, statistical analysis), but no system can autonomously design, execute, and report novel research findings at publication quality in specialized scientific domains. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI research assistants and copilots exist for literature synthesis, coding, and data analysis, but no deployed product autonomously conducts original scientific research in a specialized field at production reliability. |
Hire, supervise, or evaluate engineers, technicians, researchers, or other staff.
14CI 3–25 · exposure 13 · augmentation 63 · importance 4.3/5 · click for rater detail
Hire, supervise, or evaluate engineers, technicians, researchers, or other staff.
14| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Most organizations use AI narrowly in recruitment (job posting, resume screening) but retain human managers for interviewing, evaluation, and supervision. Adoption of AI-led hiring or automated performance management remains cautious and fragmented, far from the deep displacement seen in information-intensive professional roles. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | While HR tech adoption is growing for sourcing and screening, actual hiring/supervisory decisions remain human-led with slow, cautious adoption of AI in this specific function due to liability concerns. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially assists managers by automating resume screening, summarizing performance metrics, flagging scheduling conflicts, and generating candidate comparison reports, allowing managers to focus on interviews and relationship-building. These tools clearly raise manager productivity on administrative parts of the task while the human remains in control of final decisions. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can meaningfully assist with resume screening, performance data aggregation, and drafting evaluation documentation, but the manager retains full control over judgment-based decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can assist with candidate screening, resume ranking, and performance data aggregation, but hiring and supervision require nuanced human judgment about cultural fit, team dynamics, and complex interpersonal evaluations that current systems cannot reliably perform end-to-end. Evaluation and feedback also demand contextual understanding and accountability that AI cannot substitute for at 50%+ time savings while maintaining equal quality. |
| Task automatability | claude-sonnet-5 | 1/5 | Hiring, supervising, and evaluating staff requires nuanced judgment, interpersonal relationship management, and legal accountability that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Legal and regulatory requirements (employment law, anti-discrimination, union agreements, duty of care) mandate that a human manager retain accountability and decision authority over hiring, firing, and formal evaluation. Liability exposure for automated or AI-primary hiring decisions is substantial, creating hard barriers to full substitution. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Hiring and performance evaluation carry significant legal liability (employment law, discrimination law) and organizational requirements that mandate human decision-makers with accountability. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted recruitment tools reduce some administrative overhead, but the integration, training, oversight, and fallback human review needed for hiring and personnel decisions mean total cost remains comparable to or exceeds the value of partial automation for this complex task. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | Since AI cannot perform the core managerial functions independently, there is no viable AI substitute cost to compare against the human manager's wage for this task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI tools exist for resume filtering and basic performance metrics dashboards, but no production systems reliably handle the full hiring workflow (interviewing, reference checks, decision-making) or the interpersonal supervision and evaluation tasks at acceptable quality. Products are narrowly scoped and require extensive human oversight. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously hires, supervises, or evaluates staff; AI tools only assist with narrow sub-tasks like resume screening or scheduling. |
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.