Curators
25-4012.00Median wage $63,420/yr12,150 employed (US)Rank #477 of 923 scored · top 52% by substitution
Administer collections, such as artwork, collectibles, historic items, or scientific specimens of museums or other institutions. May conduct instructional, research, or public service activities of institution.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution28
Exposure27
Augmentation58
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
15 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
0%
Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.
Why this score
The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.
Task automatabilityw 35%29
panel mean rating 2.2/5 → substitution pressure 29/100
Technical feasibility todayw 20%23
panel mean rating 1.9/5 → substitution pressure 23/100
Cost vs. human wagew 15%26
panel mean rating 2.0/5 → substitution pressure 26/100
Adoption barriersw 20%inverted — strong barriers lower the score35
panel mean rating 3.6/5 (barrier strength) → substitution pressure 35/100
Sector adoption velocityw 10%21
panel mean rating 1.9/5 → substitution pressure 21/100
Task breakdown (15 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.
Provide information from the institution's holdings to other curators and to the public.
69CI 52–85 · exposure 70 · augmentation 75 · importance 3.9/5 · click for rater detail
Provide information from the institution's holdings to other curators and to the public.
69| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Major museums and cultural institutions are piloting AI-powered collection access and chatbots, but production deployment remains uneven; adoption is accelerating but not yet mainstream across smaller institutions. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Cultural institutions (museums, libraries, archives) are generally slow adopters of AI compared to finance or tech, with digitization and IT modernization often underfunded and uneven. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI dramatically assists curators by auto-drafting responses, cross-referencing holdings, and surfacing contextual metadata, allowing curators to focus on complex interpretation and scholarship rather than routine information retrieval. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI search, digital catalogs, and generative summarization tools significantly speed up curators' ability to locate and communicate information about holdings, even though final vetting remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | AI systems can generate, organize, and deliver information about museum/institution holdings with minimal setup—retrieval-augmented generation over collection databases requires no human intervention and saves >50% curator time on routine inquiries. Modern systems reliably match queries to holdings, provide contextual details, and format responses for public consumption. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can retrieve and summarize catalog/collection metadata and answer common inquiries via chatbots or search tools, but nuanced provenance questions, condition assessments, and specialized scholarly context still require human curatorial expertise. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Minor barriers exist: institutions may prefer human curatorial voice, require fact-checking, or have governance policies about AI use; however, no legal licensing requirement mandates a human curator field public inquiries about holdings. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement exists for sharing collection information, though institutional policies, accuracy/liability concerns, and preference for expert-verified answers create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Inference cost for answering bulk inquiries via AI is orders of magnitude cheaper than allocating curator staff time to respond to routine collection questions, emails, and public requests, especially at scale. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated search/chat systems are cheap to run once built, but the digitization, metadata curation, and integration costs plus need for expert oversight keep total cost roughly comparable to staff time for complex requests. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed museum chatbots and AI-powered collection search systems (used by major institutions and museums) demonstrate this task reliably in production, handling both basic queries and some curatorial depth; occasional errors in attribution or nuance persist but are acceptable for public-facing information dissemination. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Museums and libraries increasingly deploy digital collection search portals and AI-assisted chat interfaces, but these handle only well-documented, digitized holdings reliably and often fail on complex or undigitized items. |
Schedule events and organize details, including refreshment, entertainment, decorations, and the collection of any fees.
61CI 35–87 · exposure 58 · augmentation 63 · importance 2.8/5 · click for rater detail
Schedule events and organize details, including refreshment, entertainment, decorations, and the collection of any fees.
61| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Museums and cultural institutions increasingly adopt event management software, scheduling automation, and digital payment systems; professional and cultural sectors show steady and fairly rapid adoption of these tools. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions (where curators work) are generally slow adopters of AI for operational/event-planning tasks compared to fast-moving tech or finance sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assists curators significantly by auto-generating event timelines, suggesting vendor matches, handling scheduling conflicts, and managing payment flows, freeing curators to focus on curation strategy and attendee experience rather than logistics. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI tools can assist with calendar management, budget tracking, and drafting communications, providing moderate productivity gains while humans manage relationships and physical logistics. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | AI systems can automate the full workflow: scheduling using calendar integration, vendor coordination via email/API, collecting fees through payment processing, and generating event specifications (refreshments, entertainment, decorations) based on templates and preferences. This meets the ≥50% time-saving threshold with off-the-shelf tools. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help with scheduling logistics and drafting checklists but coordinating vendors, negotiating contracts, and physically arranging decorations/entertainment requires human execution and judgment that current AI cannot fully replace. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist for automation of scheduling and fee collection; most are administrative tasks with low liability risk. Minor friction from vendor relationships and client preference for human touch, but nothing prevents substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational trust, vendor relationships, and the need for a human point of contact for logistics create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-driven event management platforms, scheduling bots, and payment automation cost orders of magnitude less per event than hiring administrative staff or event coordinators to handle these logistics tasks. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can reduce time on drafting schedules and communications, but human oversight, vendor negotiation, and on-site coordination still dominate cost, keeping savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (calendar AI, event management platforms, payment processors, CRM systems) reliably handle scheduling, vendor communication, and fee collection in production. Minor gaps remain in real-time entertainment/decoration sourcing and dynamic adjustments, but core components are deployed at scale. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Event-planning software and AI scheduling assistants exist but are narrow-purpose tools; no deployed product manages the full multi-vendor coordination and fee collection reliably end-to-end. |
Write and review grant proposals, journal articles, institutional reports, and publicity materials.
49CI 39–59 · exposure 50 · augmentation 88 · importance 3.8/5 · click for rater detail
Write and review grant proposals, journal articles, institutional reports, and publicity materials.
49| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Many museums, universities, and cultural institutions are piloting AI writing tools for draft generation, but adoption remains experimental. Production deployment is cautious because of quality-control, attribution, and compliance concerns; full workflow automation is not yet common in the sector. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slower adopters of AI tools compared to tech-forward sectors, though AI writing assistants are increasingly used informally for drafting. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI writing assistants demonstrably improve curator productivity by generating first drafts, suggesting structures, and handling routine sections, enabling curators to focus on content strategy, voice, and validation. This assistive role is already valued in academic and institutional publishing workflows. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI writing assistants substantially speed up drafting, editing, and structuring of grants, articles, and publicity materials while curators retain final judgment and domain expertise. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can draft substantial portions of grant proposals, journal articles, and reports with coherent structure and content, potentially saving 40–50% of initial composition time. However, the task requires domain expertise, institutional knowledge, and strategic framing that demand human review and revision, preventing end-to-end automation at equal quality. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft substantial portions of grant proposals, reports, and publicity copy, but curator-specific expertise, institutional context, and accuracy checks still require significant human revision and oversight. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Grant agencies, journals, and institutions have reputational and legal stakes in proposal and publication quality; AI-generated content must be signed off by a credentialed human, and many funding bodies explicitly require human authorship and accountability. Institutional governance and professional liability create strong friction against full substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI-assisted writing, though institutional sign-off, funder trust, and accuracy expectations create moderate friction against fully automated submissions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference and integration costs are low, but curators earn professional wages and the task demands expert oversight to catch errors, maintain institutional credibility, and ensure compliance. The all-in cost of AI assistance plus expert review approaches or matches the curator's time cost. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI drafting tools cost a fraction of a curator's time per document, though human review and fact-checking still add cost, keeping it below the maximum ratio. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | LLM-based writing assistants and generative tools are deployed in academic and institutional settings, but they produce material requiring significant expert curation, fact-checking, and rewriting. No production system reliably handles the full scope—strategic positioning, institutional voice, grant strategy—without human intervention. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | General-purpose LLM writing tools are widely deployed for drafting and editing documents, but no specialized product reliably handles curatorial grant/report writing end-to-end without heavy human review. |
Develop and maintain an institution's registration, cataloging, and basic record-keeping systems, using computer databases.
47CI 44–50 · exposure 50 · augmentation 75 · importance 4.2/5 · click for rater detail
Develop and maintain an institution's registration, cataloging, and basic record-keeping systems, using computer databases.
47| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Museums and archives show uneven adoption: large, well-funded institutions increasingly deploy AI-assisted cataloging (GLAMs), but smaller and mid-size organizations remain in pilot or manual phases. Sector digitization is advancing but not yet dominated by AI-driven systems. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums, libraries, and cultural institutions are generally slow, under-resourced adopters of AI compared to finance or tech sectors, with pilots more common than full production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assistants that provide OCR, metadata suggestions, and deduplication significantly boost curator productivity in data entry and organization tasks, allowing experts to focus on interpretation and policy. This is one of the clearer augmentation wins in the curatorial space. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist with metadata generation, data entry, deduplication, and database querying, improving curator productivity while the curator retains oversight of accuracy and cataloging decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | Substantial portions of registration, cataloging, and record-keeping can be automated through database systems and AI-driven metadata extraction, but curatorial judgment about classification schemes, contextual relationships, and institutional standards typically requires human oversight. Current AI can handle routine data entry and basic categorization at ≥50% time savings, but not the full curatorial decision-making loop reliably. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can help design database schemas, generate cataloging metadata, and automate data entry/tagging, but developing and maintaining an institution-specific system still requires human decisions about structure, standards compliance, and institutional needs.rationale continues |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While no legal requirement mandates a curator for database maintenance, institutional practice and the need for subject-matter expertise in classification create adoption friction. Liability concerns around misattribution and authenticity also introduce oversight requirements that slow full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement mandates a human curator specifically, but institutional standards, provenance accuracy requirements, and accountability for collection records create meaningful friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | Software licensing, infrastructure, and AI integration costs are substantial, but when amortized across large collections and staff, can approach parity with curator labor. Small institutions may find AI more expensive; large ones with high-volume cataloging may see cost advantages. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools can reduce time on metadata generation and data cleanup, but the setup, integration, and ongoing database administration still require significant human labor and oversight, keeping costs comparable rather than dramatically lower. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Museum and archive management software with AI-assisted cataloging modules exist in production (e.g., CollectionTrust, Goobi, vendor solutions with OCR and metadata auto-fill), but these still require significant human correction, especially for complex or rare items. Reliable end-to-end automation remains limited to highly standardized collections. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Collections management software (e.g., PastPerfect, TMS) increasingly incorporates AI-assisted metadata tagging and search, but full system development/maintenance remains largely human-led with AI as a component tool, not a reliable end-to-end product. |
Confer with the board of directors to formulate and interpret policies, to determine budget requirements, and to plan overall operations.
36CI 7–64 · exposure 41 · augmentation 63 · importance 3.5/5 · click for rater detail
Confer with the board of directors to formulate and interpret policies, to determine budget requirements, and to plan overall operations.
36| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Museums and cultural institutions are generally conservative and low-digitization sectors; adoption of AI for core governance functions remains limited, with most organizations still relying on traditional board-curator consultation models. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slow adopters of AI for governance-level functions, though they may use AI tools for data/reporting support. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can significantly assist by generating policy options, budget scenarios, and operational recommendations that curators and boards then deliberate and refine; this augmentation raises human decision-making quality and speed without removing human judgment from the process. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help prepare budget analyses, summarize policy options, or draft materials for board discussions, meaningfully aiding preparation even though the core interaction remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | AI can draft policy frameworks, interpret existing policies, analyze budget requirements, and generate operational plans based on data and precedent; a human curator could validate and refine AI outputs in significantly less time than drafting from scratch. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires live interpersonal negotiation, judgment, and relationship management with a governing board, which cannot be end-to-end automated by current AI. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Museums and cultural institutions typically require curators or directors to bear legal and fiduciary responsibility for policy, budget, and strategic decisions; boards depend on human expertise and accountability that cannot be substituted by AI alone. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Board governance typically requires accountable human leadership (e.g., executive director/curator) with fiduciary and organizational authority, creating strong structural barriers to substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-generated policy drafts, budget modeling, and operational planning carry minimal inference costs compared to the high loaded wages of curators and board consultation time; the marginal cost of AI assistance is substantially lower than the human cost. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute performing this task, so cost comparison favors the human entirely; AI cannot replace the governance function. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | AI tools can produce draft policies, budget analyses, and operational plans in deployed systems, but boards require human judgment, accountability, and contextual decision-making that AI cannot fully replace; deployed products assist but do not reliably perform the full conferential and strategic role. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently confers with a board to set policy or budgets; AI is not a participant in governance decision-making today. |
Plan and conduct special research projects in area of interest or expertise.
28CI 25–30 · exposure 25 · augmentation 75 · importance 4.0/5 · click for rater detail
Plan and conduct special research projects in area of interest or expertise.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Curation and research remain heavily human-driven in museums and galleries; while digital tools are adopted for cataloging and access, autonomous research planning and execution is rare in production across these sectors. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slow, resource-constrained adopters of AI tools relative to fast-moving sectors like finance or tech. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists curators through literature mining, collection analysis, pattern identification, and preliminary report generation, allowing expert curators to focus on interpretation and scholarly judgment while maintaining full human oversight of research direction. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist with literature search, data organization, drafting research questions, and synthesizing background material, boosting curator productivity substantially while they retain control of the research direction. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with literature reviews, data gathering, and preliminary analysis, planning and conducting research projects requires human judgment about research direction, methodology selection, and original scholarly interpretation that AI cannot fully replicate end-to-end with 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Research planning and conduct requires original scholarly judgment, novel synthesis, and expertise-driven decisions that current AI can support but not independently execute at equal quality end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Research projects in museum and academic contexts carry institutional accountability, peer review requirements, and professional licensing expectations that functionally require human curators to author, validate, and sign off on findings. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement, but institutional expectations of subject-matter authority, scholarly credibility, and attribution create real friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI research tools have lower per-query costs but require significant human expert time for setup, validation, and interpretation, making the all-in cost roughly comparable to or potentially higher than human-led research. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools reduce some research time (literature search, summarization) but human oversight, subject expertise, and validation still dominate the cost of a full research project. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably plans and conducts entire research projects independently; AI tools exist for literature search and data processing but the synthesis, validation, and direction-setting remain dependent on human curators in production settings. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI research assistants can help with literature review and drafting, but no deployed product independently plans and conducts specialized curatorial research projects reliably in production. |
Arrange insurance coverage for objects on loan or for special exhibits and recommend changes in coverage for the entire collection.
26CI 18–34 · exposure 28 · augmentation 50 · importance 3.3/5 · click for rater detail
Arrange insurance coverage for objects on loan or for special exhibits and recommend changes in coverage for the entire collection.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Museums and cultural institutions operate in specialized, low-digitization sectors with strong professional norms around human expertise; adoption of AI for insurance decisions is minimal, with most institutions retaining traditional insurance brokers and in-house experts. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slow adopters of AI for administrative and risk-management functions compared to finance or tech sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by summarizing coverage options, flagging gaps in collection records, and drafting coverage recommendations, allowing curators to make faster, more informed decisions while retaining human judgment and negotiation. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help curators draft valuation reports, compare policy terms, and flag coverage gaps, meaningfully aiding the human decision-maker without replacing them. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can handle data aggregation, coverage analysis, and draft recommendations by reviewing collection databases and insurance terms; however, the task requires judgment about risk assessment, stakeholder negotiation, and final decision-making that typically involves human expertise and institutional knowledge, limiting full automation to roughly half the work. |
| Task automatability | claude-sonnet-5 | 2/5 | Involves gathering valuation data and drafting coverage recommendations, which AI can assist with, but requires judgment about risk, negotiation with insurers, and institutional accountability that current AI cannot fully replace. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Significant barriers exist: insurance recommendations typically require licensed insurance professionals in many jurisdictions, institutional liability and fiduciary duty fall on the curator or their institution, and carriers often require direct human sign-off on coverage decisions for high-value collections. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Insurance arrangements involve legal/financial liability and institutional accountability, typically requiring authorized staff or fiduciary sign-off, creating strong organizational and legal barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference and integration costs for analyzing coverage options are modest, but the task requires expert human oversight and negotiation with insurance carriers, making the blended cost comparable to or potentially higher than a human specialist managing it independently. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI could cheaply draft valuation summaries or risk analyses, but the actual arranging of insurance and liability decisions still requires human broker interaction and sign-off, keeping overall cost savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI tools can assist with document analysis and coverage comparisons, no deployed product reliably orchestrates the end-to-end task of arranging and negotiating insurance coverage with carriers and institutional sign-off; most implementations remain in pilot or assisted-analysis stages. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product specifically manages museum insurance arrangements or coverage recommendations for collections; this remains a specialized, human-led administrative function. |
Establish specifications for reproductions and oversee their manufacture or select items from commercially available replica sources.
25CI 20–30 · exposure 20 · augmentation 50 · importance 2.5/5 · click for rater detail
Establish specifications for reproductions and oversee their manufacture or select items from commercially available replica sources.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Museums and galleries have been slow to digitize curatorial workflows; many still rely on manual inspection and vendor relationships. While some institutions pilot AI-assisted cataloging, production deployment in reproduction specification is minimal. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slow adopters of AI for specialized curatorial judgment tasks, with pilots more common in digitization/cataloging than reproduction oversight. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist curators by auto-generating specification drafts from images, flagging replica candidates against known collections, and comparing material properties. However, the curator must retain final judgment, so augmentation is partial rather than transformative. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help curators research materials, compare commercial replica options, and draft specification documents, providing moderate productivity gains while the human retains final judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in generating specifications and comparing replica options against image databases, the task requires domain expertise in cultural artifacts, material authenticity, and conservation standards. The human curator must make final judgment calls on quality and appropriateness, which AI cannot reliably do end-to-end with the required consistency. |
| Task automatability | claude-sonnet-5 | 2/5 | Selecting replicas and defining reproduction specs relies on expert judgment about historical accuracy, materials, and provenance that AI cannot independently determine, though AI can assist research and comparison steps.dis Only a minority of the workflow is automatable end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Museums and cultural institutions typically require curator sign-off on reproductions for legal, ethical, and reputational reasons. Professional credentials, institutional liability, and authentic preservation standards create strong friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement exists, but institutional standards, authenticity concerns, and reputational risk create real friction against delegating this judgment to automated systems. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI image analysis and specification drafting is low, but human curators remain essential for oversight and approval, making the blended cost comparable to pure human labor. The time savings do not yet justify full replacement. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools may cheaply assist with sourcing research or comparing vendors, but the core specification-setting and oversight still requires paid expert time, keeping overall costs comparable to human-only workflows. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs the full task of establishing reproduction specifications and quality assurance. Existing vision and database systems can flag candidates or draft specs, but curators still manually inspect, negotiate with vendors, and verify fidelity—AI offers partial support only. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | There are no deployed products that autonomously establish reproduction specifications or manage manufacturing oversight for museum artifacts; this remains a specialized human curatorial function. |
Design, organize, or conduct tours, workshops, and instructional or educational sessions to acquaint individuals with an institution's facilities and materials.
21CI 11–30 · exposure 13 · augmentation 63 · importance 3.8/5 · click for rater detail
Design, organize, or conduct tours, workshops, and instructional or educational sessions to acquaint individuals with an institution's facilities and materials.
21| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Museum and cultural sectors are slower to digitize and automate compared to information/finance; AI deployment in educational programming remains sparse and experimental. Adoption is primarily at the pilot stage, with limited production displacement of curator-led tours and workshops. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slow adopters of AI for public-facing programming, with pilots (like AI docent apps) still rare and experimental. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist curators by drafting tour narratives, suggesting interactive elements, personalizing content recommendations for different audience segments, and generating multilingual materials. However, the live facilitation and audience rapport remain human-dependent, limiting augmentation to preparation and planning phases. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist curators in drafting tour scripts, workshop curricula, educational materials, and audience-tailored content, improving efficiency in the design phase. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate tour scripts and suggest workshop content, the task fundamentally requires real-time, adaptive interaction with live audiences—reading engagement, answering impromptu questions, and adjusting pacing dynamically. Current AI cannot reliably conduct in-person educational sessions with the responsiveness and human judgment this demands, though it could assist with planning and material generation. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires in-person facilitation, physical navigation of a venue, live audience management, and real-time adaptation—none of which current AI can perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Museums, libraries, and educational institutions typically have strong organizational preferences for human educators, visitor expectations of personal engagement, and liability concerns around unsupervised AI-led instruction. Regulatory and accreditation standards often implicitly or explicitly require qualified human staff to lead educational programs. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but institutional expectations, visitor preference for human interaction, and liability for group safety during physical tours create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The overhead of setting up, integrating, and supervising AI systems to deliver educational experiences comparable to human curators remains high, and would require human oversight or backup. The all-in cost per session likely exceeds the wage cost of a curator, especially for small-to-medium institutions. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply generate scripts or content outlines, but the actual conducting of tours and workshops still requires paid human staff, so overall cost savings are limited to the planning portion. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product conducts full educational tours or workshops autonomously in production settings. AI can generate educational content and outline structures, but reliable production systems for live facilitation of educational sessions with heterogeneous groups do not exist at scale; this remains largely in research/prototype stages. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product conducts physical museum tours or in-person workshops; AI tour aids exist only as supplementary audio guides or chatbots, not substitutes for the curator's live design and delivery role. |
Plan and organize the acquisition, storage, and exhibition of collections and related materials, including the selection of exhibition themes and designs, and develop or install exhibit materials.
19CI 14–25 · exposure 20 · augmentation 50 · importance 4.3/5 · click for rater detail
Plan and organize the acquisition, storage, and exhibition of collections and related materials, including the selection of exhibition themes and designs, and develop or install exhibit materials.
19| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Museums and cultural institutions are among the slowest sectors to adopt automation; they are small, dispersed, not heavily digitized, and operate under professional norms that value human expertise. Adoption of AI in curatorial roles remains minimal. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are typically small, under-resourced, and slow to adopt AI tools compared to sectors like finance or tech; AI use is mostly experimental. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist curators with collection search, metadata tagging, theme ideation, exhibit layout visualization, and visitor analytics feedback, raising productivity in parts of the task. However, the core curatorial decisions remain human-driven, limiting the scope of augmentation. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help brainstorm exhibition themes, draft interpretive text, research provenance, and generate design mockups, meaningfully aiding curators without replacing core judgment and physical work. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with collection cataloging, theme generation, and exhibit layout suggestions, the task fundamentally requires human curatorial judgment about acquisition decisions, artistic vision, and contextual interpretation. Current AI systems cannot autonomously perform the full end-to-end planning and acquisition decision-making that defines this role. |
| Task automatability | claude-sonnet-5 | 2/5 | This task blends physical logistics, strategic planning, aesthetic judgment, and stakeholder negotiation that current AI cannot execute end-to-end; AI can assist with subtasks like drafting exhibit text but not the full workflow. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong professional and institutional barriers protect curatorship: museums typically require accredited curatorial staff; curatorial decisions carry reputational and legal liability; professional standards and institutional policies mandate human judgment over acquisition and exhibition choices; and stakeholders (boards, donors, public) expect human expertise and accountability. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Curatorial authority often involves institutional accreditation, provenance/legal responsibility for collections, insurance and conservation standards, and donor relationships that require human accountability. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for collection management and exhibit design are supplementary rather than substitutive, and the expert human curator's loaded wage remains significantly lower than the cost of developing, integrating, and overseeing AI systems that could replace curatorial judgment. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | Physical installation, object handling, and negotiation with donors/lenders require human labor and judgment that AI cannot substitute for at any meaningful cost advantage today. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs curatorial planning, collection acquisition, or exhibition design autonomously. Tools exist for inventory management and design mockups, but none handle the integrated curatorial function at production scale in museums or galleries. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product manages museum acquisitions, storage logistics, and physical exhibit installation reliably; this remains far outside current production AI capability. |
Inspect premises to assess the need for repairs and to ensure that climate and pest control issues are addressed.
19CI 7–30 · exposure 13 · augmentation 50 · importance 3.8/5 · click for rater detail
Inspect premises to assess the need for repairs and to ensure that climate and pest control issues are addressed.
19| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Museums and cultural institutions typically adopt automation slowly; they remain conservative stewards of collections and rely on domain-expert judgment, with remote monitoring sensors used piecemeal rather than full autonomous inspection. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slow adopters of automation for physical facility tasks, though environmental monitoring sensors are increasingly common as an aid. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-powered sensors and alerts can assist curators by flagging anomalies (temperature drift, motion detection) and surfacing inspection prioritization, enabling more efficient human rounds without replacing the curator's interpretive judgment. |
| Augmentation potential | claude-sonnet-5 | 3/5 | IoT sensors and AI-based monitoring systems for temperature, humidity, and pest detection can alert curators to problems, meaningfully augmenting but not replacing the physical inspection and judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Visual inspection of premises for structural damage and environmental conditions is partially automatable via cameras and sensors, but identifying pest control issues and assessing repair urgency requires contextual judgment and real-time problem-solving that current AI struggles with at scale. |
| Task automatability | claude-sonnet-5 | 1/5 | Requires physical presence, sensory inspection of a building and collections environment, and judgment about structural/pest/climate risks—no AI system can perform this physical inspection today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Museum accreditation standards and conservation best practices often require credentialed professionals to certify the condition of premises and collections; liability for missed pest infestations or environmental failures creates legal accountability that favors human sign-off. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically for this inspection, but institutional liability, collection-care standards, and insurance requirements create real organizational friction around who assesses risk. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Inspection robots and AI vision systems require substantial upfront capital, integration, and continuous human oversight to validate findings, making the all-in cost comparable to or exceeding a curator's periodic inspection rounds. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the physical inspection itself, so there is no viable AI cost comparison; sensor systems add cost rather than replacing the human walk-through. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Robotic inspection systems and computer vision exist for facility monitoring, but they are narrowly scoped (e.g., image detection of visible damage) and require significant setup and human verification; no mature product reliably performs the full diagnostic task end-to-end in museum/curation environments. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously inspects museum/archive premises for repairs, pest, and climate issues; this remains a human physical task, though sensors can log conditions. |
Study, examine, and test acquisitions to authenticate their origin, composition, history, and to assess their current value.
14CI 5–23 · exposure 13 · augmentation 50 · importance 3.8/5 · click for rater detail
Study, examine, and test acquisitions to authenticate their origin, composition, history, and to assess their current value.
14| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Museum curation remains a specialized, small-scale sector with limited digitization adoption and strong preference for human expert judgment. Sector-wide deployment of AI authentication is negligible; pilots are few and adoption is laggard. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Museums and cultural institutions are typically slow-adopting, resource-constrained, and this task involves physical/archival work with minimal AI agent penetration to date. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist curators by automating initial image analysis, cross-referencing historical records, and flagging compositional anomalies, raising the speed of preliminary assessment. However, the human expert remains central to final authentication and valuation decisions. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with provenance research (searching records, databases, image comparison) and stylistic analysis, meaningfully aiding curators though final authentication judgment remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with image analysis, composition assessment via spectroscopy, and historical research lookup, the task requires expert judgment on authenticity and nuanced evaluation that demands human verification. Current systems cannot reliably handle the full chain of authentication without substantial human oversight, falling short of the 50% time-saving bar. |
| Task automatability | claude-sonnet-5 | 1/5 | Physical examination, forensic testing, provenance research involving physical archives, and expert judgment on authenticity require hands-on inspection and specialized expertise that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Museums face strong regulatory and fiduciary requirements to verify provenance and authenticity; many jurisdictions impose legal accountability on curators for acquisitions. Institutional risk tolerance, professional liability concerns, and donor/collector expectations create substantial friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Authentication often carries legal, insurance, and provenance liability implications requiring accountable human expert sign-off, especially for high-value or historically significant acquisitions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tooling for imaging and database lookup is inexpensive, but the high cost of errors in authentication and valuation (liability, reputational damage) means museums typically retain expert staff. The all-in cost of an AI system plus required human oversight approaches or exceeds specialist curator wages. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | Physical testing equipment, chain-of-custody documentation, and expert judgment cannot be replaced by cheaper AI inference; human expert cost is unavoidable for authentication decisions. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI products exist for artifact imaging and basic historical research retrieval, but no deployed system reliably authenticates origin or assesses value end-to-end. Museums rarely rely on AI alone for acquisition authentication; expert curation remains the standard practice. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously authenticates artifacts, tests composition, or assesses provenance and value; this remains a human expert-driven, research-stage-only application for AI. |
Train and supervise curatorial, fiscal, technical, research, and clerical staff, as well as volunteers or interns.
6CI 5–7 · exposure 0 · augmentation 38 · importance 3.6/5 · click for rater detail
Train and supervise curatorial, fiscal, technical, research, and clerical staff, as well as volunteers or interns.
6| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Museum and cultural institutions tend to be laggard sectors in AI adoption, and staff management is among the most human-centric functions; actual displacement is minimal and unlikely near-term. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Museums and cultural institutions are generally slow adopters of AI for management functions, with most current use limited to administrative support tools rather than supervisory replacement. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could marginally assist with scheduling training sessions, documenting feedback, or aggregating performance data, but these are peripheral to the core supervisory and relational work of training and mentoring staff. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help with scheduling, training material creation, performance tracking, and communication drafts, moderately aiding but not transforming the supervisory task itself. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task fundamentally involves managing human personnel—training, supervision, evaluation, and mentorship. Current AI cannot meaningfully replace the interpersonal judgment, accountability, and adaptive guidance required for staff development and oversight. |
| Task automatability | claude-sonnet-5 | 1/5 | Supervising and training a diverse mix of staff and volunteers requires interpersonal leadership, judgment, and institution-specific mentorship that AI cannot perform end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Legal and organizational liability, employment law compliance, and fiduciary duty requirements mean that a human supervisor must legally retain responsibility for personnel decisions, training records, and performance management. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Personnel supervision typically involves institutional policy, HR/legal accountability, and human trust-building that create strong organizational and quasi-regulatory barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | An AI system attempting to handle staff supervision would require significant ongoing human oversight (negating cost savings) and could not substitute for the accountability a human supervisor provides to the organization. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no AI substitute for actual supervisory labor, so no meaningful cost comparison favors AI; a human manager is required regardless. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI systems perform staff training and supervision reliably in production. These tasks require real-time human judgment, conflict resolution, and personalized feedback that remains beyond current technology. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product manages or supervises human staff directly; this remains a human management function with only ancillary AI tools for scheduling or documentation. |
Attend meetings, conventions, and civic events to promote use of institution's services, to seek financing, and to maintain community alliances.
3CI 0–5 · exposure 0 · augmentation 38 · importance 3.6/5 · click for rater detail
Attend meetings, conventions, and civic events to promote use of institution's services, to seek financing, and to maintain community alliances.
3| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Cultural institutions remain highly traditional and relationship-dependent sectors with low digitization of community-facing roles. Adoption of AI for representing the institution in public and donor settings is negligible. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Museums and cultural institutions are a small, low-digitization sector with minimal evidence of AI displacing in-person networking, fundraising, or civic engagement roles. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist a curator by drafting promotional materials, summarizing community feedback, or identifying funding opportunities before meetings, but it offers limited assistance for the core interpersonal work of attending events and building alliances. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help curators prepare talking points, draft outreach materials, research attendees, and follow up on contacts, meaningfully aiding preparation and follow-through even though the core event participation is human. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires sustained in-person presence, real-time social negotiation, relationship-building, and authentic representation of institutional values—activities that current AI cannot perform end-to-end. AI cannot meaningfully attend events or negotiate alliances on behalf of an institution. |
| Task automatability | claude-sonnet-5 | 1/5 | This is an in-person relationship-building and representational task requiring physical presence, live networking, and real-time persuasion that current AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Institutional representation, community trust, and stakeholder relationship-building are deeply human-contact requirements. Legal, fiduciary, and reputational norms strongly require a human representative with real authority and accountability to fulfill these roles. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Institutional representation, fundraising relationships, and community trust require a recognized human figurehead; stakeholders expect and often require personal presence and accountability from institutional leadership. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The cost of AI systems designed to attend events and build relationships (if such existed) would far exceed the cost of a curator doing this work directly, given the low task complexity and high need for human presence and judgment. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute for physical attendance and relationship management, so cost comparison favors the human by default. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product performs this task in production. While AI can draft communications or suggest talking points, it cannot substitute for the human presence, credibility, and interpersonal dynamics essential to attending meetings and maintaining community alliances. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product attends meetings or civic events on a curator's behalf to build community alliances or secure financing; this remains firmly a human, embodied social activity. |
Negotiate and authorize purchase, sale, exchange, or loan of collections.
0CI 0–0 · exposure 0 · augmentation 38 · importance 3.8/5 · click for rater detail
Negotiate and authorize purchase, sale, exchange, or loan of collections.
0| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Museums and cultural institutions operate in highly regulated, risk-averse sectors with strong traditions of human expert judgment. Adoption of AI for transaction authorization is minimal; negotiation and acquisitions remain entirely human-driven. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Museums and cultural institutions are low-digitization, slow-adopting environments for AI-driven decision authority, especially for high-stakes asset transactions. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI can modestly assist by gathering provenance data, comparing market prices, or drafting research summaries, but the core negotiation and authorization acts remain curator-led. Augmentation exists but does not materially amplify curator productivity in the authorizing decision itself. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with market research, provenance checks, valuation comparables, and drafting loan agreements, but the negotiation and authorization itself remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task fundamentally requires negotiation, judgment calls about institutional value, and authorized decision-making that current AI cannot execute independently. AI can assist with research and documentation but cannot legally authorize financial or contractual transactions. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires trust-based negotiation, valuation judgment, institutional authority, and legal sign-off that current AI cannot execute end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Strong legal and fiduciary barriers exist: curators must be authorized agents of their institution, collections involve legal title transfer, and liability for authenticity and valuation rests with the human decision-maker. Institutional and legal frameworks require a licensed professional signature. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Authorization of acquisitions/loans typically requires institutional governance, board or director approval, fiduciary responsibility, and legal accountability that only an authorized human can hold. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | A curator performing this task commands a high salary, often $50k–$100k+, and the authorization and fiduciary responsibility cannot be fully assumed by AI systems, making human-performed cost lower when accountability is included. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this task, so cost comparison favors the human curator by default; AI cannot yet produce the output at all reliably. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI system can independently negotiate acquisitions or authorize purchases on behalf of institutions. This requires human authority, accountability, and legal standing that no current product provides. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously negotiates or authorizes acquisitions or loans of museum collections; this remains firmly outside current commercial AI offerings. |
Related occupations — Educational Instruction & Library
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.