Office Clerks, General
43-9061.00Median wage $45,010/yr2,464,940 employed (US)Rank #31 of 923 scored · top 3% by substitution
Perform duties too varied and diverse to be classified in any specific office clerical occupation, requiring knowledge of office systems and procedures. Clerical duties may be assigned in accordance with the office procedures of individual establishments and may include a combination of answering telephones, bookkeeping, typing or word processing, office machine operation, and filing.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution64
Exposure61
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
20 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
55%
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%61
panel mean rating 3.5/5 → substitution pressure 61/100
Technical feasibility todayw 20%60
panel mean rating 3.4/5 → substitution pressure 60/100
Cost vs. human wagew 15%69
panel mean rating 3.8/5 → substitution pressure 69/100
Adoption barriersw 20%inverted — strong barriers lower the score72
panel mean rating 2.1/5 (barrier strength) → substitution pressure 72/100
Sector adoption velocityw 10%57
panel mean rating 3.3/5 → substitution pressure 57/100
Task breakdown (20 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.
Type, format, proofread, and edit correspondence and other documents, from notes or dictating machines, using computers or typewriters.
86CI 75–97 · exposure 87 · augmentation 100 · importance 3.5/5 · click for rater detail
Type, format, proofread, and edit correspondence and other documents, from notes or dictating machines, using computers or typewriters.
86| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Office environments and information-heavy sectors show strong, measurable adoption of AI writing assistance, transcription, and auto-formatting tools. Grammarly alone has hundreds of millions of users, indicating deep penetration in both individual and organizational workflows. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Office and administrative functions across sectors have rapidly adopted AI-assisted writing, transcription, and editing tools, though full end-to-end automation in production is still uneven across smaller firms. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI writing assistants are already transforming clerical productivity by providing real-time suggestions, grammar checking, tone adjustment, and formatting assistance while the clerk retains editorial control. This is a textbook case of human-AI partnership raising output quality and speed. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI dramatically speeds up drafting, transcription, and proofreading while clerks retain oversight for final accuracy and formatting choices specific to organizational needs. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can handle typing, basic formatting, and spell-checking automatically. Proofreading and editing are substantially automatable via LLM-based tools that catch grammar, style, and consistency errors. However, context-specific judgment about tone and intent may require human review, preventing a full 5. |
| Task automatability | claude-sonnet-5 | 5/5 | Typing, formatting, proofreading, and editing text from notes or dictation is squarely within the capability of current AI tools like transcription software and LLM-based editors, easily exceeding 50% time savings at equal or better quality.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or licensing barriers exist for automating general clerical editing and formatting. Some organizations prefer human review for confidentiality or quality control, but these are soft preferences rather than hard regulatory requirements. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There is no licensing, legal, or human-contact requirement for typing and editing documents; organizations can freely substitute AI tools with minimal friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI transcription, formatting, and editing tools cost pennies per document compared to a clerk's loaded hourly wage. Even accounting for integration and oversight, AI is typically one or two orders of magnitude cheaper per standardized document. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Subscription-based AI transcription and editing tools cost a few dollars per month versus hourly clerical wages, representing well over an order-of-magnitude cost reduction per task-equivalent. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products exist (Grammarly, Microsoft Editor, specialized transcription + editing pipelines) that demonstrably perform these tasks in production at scale across many organizations. Some edge cases and stylistic nuance remain, but the core functionality is reliably deployed. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Mature deployed products (Word's editor, Grammarly, Otter.ai, LLM-based writing assistants) perform transcription, formatting, and proofreading reliably at scale in production today. |
Complete and mail bills, contracts, policies, invoices, or checks.
83CI 79–87 · exposure 83 · augmentation 63 · importance 4.1/5 · click for rater detail
Complete and mail bills, contracts, policies, invoices, or checks.
83| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Financial services, insurance, and large enterprises have already deployed RPA and document automation for billing and invoice processing at scale; adoption is deep in white-collar information sectors and continues to accelerate. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Billing and invoicing automation is widespread across finance, administrative, and back-office functions, a sector showing fast, deep adoption of digital and AI-assisted workflow tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI assists by generating and formatting documents, flagging anomalies, and routing to appropriate mailboxes, meaningfully raising clerk productivity; however, the task is straightforward enough that augmentation adds less value than full automation would. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-driven billing systems significantly speed up document preparation, error-checking, and mailing/dispatch logistics, letting clerks focus on exceptions and oversight rather than manual entry. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Completing and mailing bills, contracts, policies, invoices, or checks is largely a structured data-entry and document-generation task. Current AI systems can extract data, populate templates, generate documents, and integrate with mail systems—readily achieving >50% time savings at equal quality through off-the-shelf workflow automation and RPA tools. |
| Task automatability | claude-sonnet-5 | 4/5 | Generating and sending bills, invoices, and standard documents is largely templated data-entry work that current AI and RPA/document-automation tools can handle end-to-end with substantial time savings, though check issuance and physical mailing require some human/physical steps. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist for AI to complete routine administrative documents; however, some organizations impose internal controls requiring human sign-off on financial documents, and a small subset of contracts may require licensed professional involvement. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for issuing bills or invoices, though check signing may require authorized signatories and some compliance/audit controls create minor friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-driven document automation and mail integration costs (per-transaction fees, licensing) are typically orders of magnitude cheaper than loaded human clerk wages for the same volume of bills, invoices, and checks. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated billing/invoicing software costs a small fraction of a clerk's hourly wage per transaction once set up, offering order-of-magnitude cost savings for high-volume repetitive document generation and mailing. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (document generation APIs, RPA platforms like UiPath, and cloud ERP systems) reliably perform this task in production across organizations. Minor friction remains in handling edge cases or custom formats, but the core task is demonstrably deployed at scale. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature invoicing, billing, and accounts-payable/receivable software (e.g., QuickBooks, SAP, e-invoicing platforms) already automate generation and electronic delivery of these documents in production at scale, though physical mail and check printing still need some manual or semi-automated handling. |
Complete work schedules, manage calendars, and arrange appointments.
81CI 72–89 · exposure 80 · augmentation 100 · importance 3.7/5 · click for rater detail
Complete work schedules, manage calendars, and arrange appointments.
81| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Scheduling automation has achieved broad adoption in information-sector organizations (tech, finance, professional services); calendar assistants are standard in Microsoft 365 and Google Workspace environments, reflecting rapid penetration. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Scheduling automation is common in tech-forward and professional-services firms but adoption in many general clerical settings across smaller organizations remains uneven and often supplementary rather than fully autonomous. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI scheduling assistants meaningfully augment human clerks by automating routine coordination, freeing them to handle complex exceptions, stakeholder communication, and strategic scheduling—transforming productivity while the human remains in control. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI calendar assistants substantially reduce the time and cognitive load of arranging appointments and managing schedules while humans retain oversight for exceptions and judgment calls. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can automate most of this task end-to-end: calendar parsing, conflict detection, appointment scheduling, and reminder management are well-solved with current AI and integrations. Some edge cases (unusual constraints, human negotiation) may require oversight, but the ≥50% time-saving bar is clearly met. |
| Task automatability | claude-sonnet-5 | 4/5 | Scheduling and calendar management are structured, rule-based tasks well within reach of current AI scheduling assistants and agent tools that can negotiate meeting times, send invites, and manage conflicts with minimal human input.deployed. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Minimal legal or regulatory barriers exist; most organizations control their own calendaring systems and can deploy automation. The main friction is organizational preference for human availability and occasional sensitivity around access to executives' schedules, but these are soft, not hard barriers. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There are no licensing, legal, or regulatory requirements tied to scheduling tasks, and no inherent need for human authorization or signoff. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Calendar automation via AI is cheap (near-zero marginal cost per scheduling action) compared to the fully-loaded cost of a human clerk performing these tasks manually. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI scheduling tools operate at subscription costs far below the fully loaded cost of clerical labor for the same volume of scheduling work, though some oversight and exception-handling costs remain. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Multiple mature products (Google Calendar AI, Outlook scheduling assistants, dedicated scheduling bots) already perform this reliably in production at scale, with natural-language understanding and integration into enterprise calendaring systems. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Products like Microsoft Copilot, Google Calendar AI features, and dedicated scheduling assistants (e.g., Clara, Reclaim.ai) reliably perform calendar management and appointment scheduling in production today, though edge cases still need human correction. |
Prepare meeting agendas, attend meetings, and record and transcribe minutes.
77CI 75–79 · exposure 75 · augmentation 88 · importance 3.6/5 · click for rater detail
Prepare meeting agendas, attend meetings, and record and transcribe minutes.
77| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Information-sector organizations (tech, finance, professional services) have rapidly adopted AI meeting transcription and minute-generation tools; mainstream enterprise deployment is now common. Smaller or traditional sectors lag, but velocity is clearly strong in high-digitization sectors. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | AI notetaking bots are now commonly deployed in corporate and professional settings via Zoom, Teams, and Google Meet integrations, reflecting fast adoption in white-collar office environments. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI augmentation is significant: humans use AI-generated transcripts and draft minutes to focus on synthesis, action-item tracking, and follow-up communication rather than manual note-taking. This substantially raises clerk productivity while keeping human oversight. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially augments this task by auto-generating agendas, live transcripts, and searchable summarized minutes, letting clerks focus on formatting, distribution, and follow-up actions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI can generate meeting agendas from prior context, auto-generate transcripts from audio, and produce draft minutes with high fidelity. However, selecting what matters most in a meeting and synthesizing action items still benefits from human judgment, preventing a full 5-rating. End-to-end automation easily clears the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 4/5 | AI transcription and summarization tools can now generate accurate meeting minutes and draft agendas from calendars/prior notes, meeting most of the ≥50% time-savings bar; only physical meeting attendance and light editing remain human tasks. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing or legal requirement mandates human secretarial work, and minimal liability attaches to automated transcription or draft minutes. Organizations may prefer human polish or privacy oversight, but these are weak friction points, not hard barriers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement exists for taking minutes, though some organizations prefer a human present for confidentiality, accuracy verification, or to answer real-time questions during meetings. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI transcription and minute-generation services cost pennies to dollars per meeting, while a clerk's loaded labor cost to attend and manually transcribe is $15–40+ per hour. At-scale AI is roughly 50–100× cheaper per task-equivalent. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Subscription-based AI transcription/summarization tools cost a few dollars per user per month versus substantial clerk time spent manually transcribing and formatting minutes, yielding large cost savings for the automatable portion. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (Fireflies, Otter, Microsoft Teams transcription, AI-assisted note-taking in Notion/Confluence) reliably handle transcription and draft minute generation at scale in production. Quality is strong for standard meetings, though edge cases and complex discussions occasionally require correction, keeping this below a 5. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed products like Otter.ai, Microsoft Teams/Zoom AI notetakers, and Google Gemini in Workspace reliably transcribe and summarize meetings in production at scale today, though accuracy varies with audio quality and jargon. |
Compile, copy, sort, and file records of office activities, business transactions, and other activities.
76CI 72–79 · exposure 75 · augmentation 63 · importance 4.0/5 · click for rater detail
Compile, copy, sort, and file records of office activities, business transactions, and other activities.
76| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Corporate and financial sectors have widely deployed RPA and document automation for back-office record management and filing; adoption is measurable and accelerating, particularly in large enterprises and finance/insurance where high-volume processing justifies the investment. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Office administrative functions across many industries have adopted document automation and RPA at a moderate pace, but many smaller and less digitized offices still rely on manual processes. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI and automation tools assist human clerks by handling bulk sorting, auto-filing, and data entry validation, improving productivity on routine portions of the task. However, augmentation is limited because the task itself is heavily automatable, leaving less room for meaningful human-AI collaboration. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered search, auto-tagging, and smart filing significantly speed up a clerk's ability to compile and organize records, even where full automation is not yet total. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Modern document management systems, RPA, and AI can handle copying, sorting, and filing records with high efficiency and minimal human intervention. The repetitive, rule-based nature of compilation and filing allows for substantial (likely >50%) time savings, though initial setup and handling of edge cases may require human oversight. |
| Task automatability | claude-sonnet-5 | 4/5 | Compiling, sorting, and filing records is highly structured and repetitive text/data work that current document-processing and workflow-automation tools can handle largely end-to-end, especially with digital records and templated file structures. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers exist for automating record compilation and filing; most organizations have discretion over their document workflows. Organizational inertia and IT infrastructure constraints present some friction, but no licensing requirement or mandatory human sign-off applies to this routine administrative task. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal sign-off is required for this clerical task, though some organizational inertia and data-handling/privacy policies create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | RPA and document management systems have low per-transaction costs (often fractions of cents) compared to loaded human clerical wages ($30-50k annually), yielding an order-of-magnitude cost advantage for high-volume filing and sorting operations. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated filing and document sorting software costs far less per unit of throughput than a human clerk's loaded wage, though initial integration and occasional oversight add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products including document automation platforms (e.g., UiPath, Automation Anywhere), enterprise content management systems, and cloud filing solutions reliably perform sorting, copying, and filing at scale in production environments. Some variation in document formats and business logic may require configuration. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature products (RPA tools, OCR/document management systems, cloud file organization with AI tagging) are deployed at scale in offices today, though edge cases like ambiguous categorization or physical documents still require human intervention. |
Compute, record, and proofread data and other information, such as records or reports.
76CI 72–79 · exposure 75 · augmentation 88 · importance 3.8/5 · click for rater detail
Compute, record, and proofread data and other information, such as records or reports.
76| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Clerical and administrative sectors have been early and rapid adopters of RPA and intelligent automation; many organizations already use these tools in production for data handling and reconciliation. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Clerical/administrative functions are adopting AI tools steadily but unevenly across industries and firm sizes, with many small offices still using manual processes despite available technology. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can significantly assist human clerks by auto-populating fields, flagging anomalies, and pre-screening records, allowing humans to focus on exception handling and complex cases while productivity increases substantially. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-powered proofreading, data validation, and spreadsheet formula tools substantially boost speed and accuracy for clerks while they remain in the loop for judgment calls and exceptions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Data entry, computation, and proofreading are largely rule-based and automatable; OCR and AI systems can extract, process, and verify information with high accuracy. The task requires minimal judgment or context, making it well-suited for automation that could easily exceed 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 4/5 | Computing, recording, and proofreading structured data is highly amenable to LLMs and automation scripts, especially for text-based or tabular data with clear formats.4/5 as full end-to-end automation still needs occasional human verification for edge cases and system integration. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Minimal regulatory or licensing barriers exist for automating clerical tasks; organizations mainly face internal adoption friction and employee transition concerns, but nothing legally prevents substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal requirement mandates a human for this clerical work; the main friction is organizational inertia and error-correction workflows rather than regulatory or liability barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | API-based data processing, OCR, and verification tools are extremely cheap compared to loaded labor costs for clerical work; automation delivers an order of magnitude or greater cost reduction. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated data entry and proofreading tools cost a small fraction of a clerk's hourly wage per unit of output, though some human oversight is still needed for exceptions, keeping it just below the top tier. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (RPA, intelligent document processing, AI-powered proofreading tools) already perform these functions reliably in production environments at scale, though some oversight and human spot-checking may remain depending on domain. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Products like spreadsheet AI add-ins, OCR/data-entry automation tools, and grammar/proofreading software (e.g., Grammarly, Excel Copilot) are deployed at scale and reliably handle much of this work today. |
Process and prepare documents, such as business or government forms and expense reports.
75CI 75–75 · exposure 75 · augmentation 75 · importance 3.8/5 · click for rater detail
Process and prepare documents, such as business or government forms and expense reports.
75| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Enterprise and government organizations have rapidly deployed document automation and RPA for back-office form processing; this is one of the earliest and most mature RPA use cases in production at scale across finance, HR, and administrative departments. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Expense and document processing automation is a mature, fast-adopted category within back-office/financial administration functions across many industries. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI document processing assists clerks by auto-populating fields, flagging anomalies, and organizing documents, materially raising throughput and reducing errors even when humans remain responsible for review and submission. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly speed up data extraction, form-filling, and validation, letting clerks focus on exceptions and verification rather than manual entry. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can extract data from structured and semi-structured forms, classify document types, and populate templates with high accuracy. RPA and document processing AI (e.g., layout-based OCR + form field extraction) routinely achieve >50% time savings on form processing; end-to-end automation of routine expense reports and standard business forms is demonstrably feasible today. |
| Task automatability | claude-sonnet-5 | 4/5 | Processing and preparing standardized forms and expense reports involves structured data extraction, entry, and formatting that current AI/OCR/document automation tools handle well, though exception handling still needs human review. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers exist for automating routine form completion and data entry. Organizations may impose internal approval workflows and accuracy checks, but no licensing requirement mandates human involvement in these clerical tasks. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Minor friction exists around approval sign-offs and compliance checks for financial forms, but no licensing requirement mandates a human clerk perform this specific task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Cloud-based document processing and RPA have fallen to commodity pricing ($0.01–0.10 per document depending on complexity and volume), making them substantially cheaper than a human clerk at loaded wage ($25–35/hour) when processing dozens of forms daily. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated document processing and expense report tools cost a small fraction of clerical wages per document processed once integrated, though initial setup and occasional exception handling add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed products (UiPath, Automation Anywhere, document intelligence APIs from Azure/AWS) reliably handle form processing in production environments, especially for high-volume, standardized documents. Error rates on well-formatted inputs are low, though edge cases and handwritten or scanned documents still present challenges. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Products like expense management systems (Expensify, SAP Concur), OCR-based document processors, and RPA tools are widely deployed in production for exactly this workflow across many organizations. |
Inventory and order materials, supplies, and services.
74CI 72–75 · exposure 75 · augmentation 75 · importance 3.5/5 · click for rater detail
Inventory and order materials, supplies, and services.
74| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Retail, e-commerce, manufacturing, and logistics sectors have rapidly adopted automated inventory and procurement systems over the past decade; this is among the most digitized office functions, with widespread production deployment. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Adoption varies widely by firm size and sector; large organizations have adopted automated procurement systems while small offices still handle this manually, giving middling overall velocity. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assists clerks by automating routine reordering and flag-setting while the human validates exceptions, approves unusual purchases, and manages supplier relationships—substantial productivity gains while maintaining human oversight. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered inventory software significantly boosts clerks' productivity by flagging low stock, suggesting order quantities, and automating routine purchase orders while humans retain oversight for exceptions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can automate most of this task: monitoring inventory levels, flagging reorder points, generating purchase orders, and integrating with supplier systems are all feasible with existing tools. Some human judgment around vendor selection or emergency sourcing may remain, but the core workflow achieves >50% time savings. |
| Task automatability | claude-sonnet-5 | 4/5 | Inventory tracking and reordering follows clear rules and thresholds that can be handled by software with AI-driven demand forecasting and automated purchase order generation, meeting the 50% time-saving bar for most routine cases.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist for automating inventory management. Main friction is organizational inertia and vendor integration complexity, not licensing or human-sign-off requirements that would hard-block substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational friction exists around vendor relationships, approval authority, and error costs from wrong orders requiring some human oversight. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Cloud-based inventory and procurement automation is inexpensive per transaction and scales well; the AI cost per cycle is typically a small fraction of the clerk labor it displaces, especially in bulk or continuous operations. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated inventory/ordering software costs a small fraction of a clerk's wage for the same volume of transactions once implemented, though setup and integration add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature ERP and inventory management systems with AI/ML capabilities are deployed at scale in organizations today (SAP, Oracle, Netsuite, Shopify). These reliably track inventory, predict demand, and automate ordering, though edge cases and manual overrides still occur in production. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature inventory management systems (ERP modules, procurement platforms with AI forecasting like SAP, NetSuite) are widely deployed in production and reliably automate reordering and stock tracking today. |
Maintain and update filing, inventory, mailing, and database systems, either manually or using a computer.
73CI 67–79 · exposure 70 · augmentation 63 · importance 4.0/5 · click for rater detail
Maintain and update filing, inventory, mailing, and database systems, either manually or using a computer.
73| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Information and administrative sectors have rapidly adopted RPA and database automation; adoption is now common in medium and large organizations, with cloud-based systems accelerating deployment across industries. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Office/clerical functions have moderate AI and automation adoption, with pilots and partial deployments common in administrative software but full replacement still uneven across firms. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI tools assist with data entry accuracy checks, smart filing suggestions, and database query assistance, meaningfully improving clerk productivity without replacing the role, though the augmentation is steady rather than transformative. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI and automation tools significantly speed up data entry, filing, and database maintenance tasks, providing strong productivity gains while humans retain oversight and manage exceptions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Modern AI systems and RPA tools can automate a large majority of filing, inventory tracking, mailing list maintenance, and database updates with minimal manual intervention, easily exceeding 50% time savings. However, edge cases requiring human judgment (ambiguous categorization, data validation exceptions) may still demand oversight. |
| Task automatability | claude-sonnet-5 | 4/5 | Much of this task (digital filing, database updates, mailing list management) can be handled by software automation, RPA, and AI agents integrated with common office systems, though physical filing and inventory still require human action. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist for automating administrative filing and database maintenance; most concerns are organizational inertia and data security practices rather than hard legal requirements. Customer preference for human contact is minimal for back-office tasks. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement exists for this task, but some organizational friction around data accuracy, legacy systems, and physical inventory handling creates moderate resistance. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-driven automation (RPA, cloud database services, OCR) costs a fraction of a full-time clerk's loaded wage ($35k–$50k annually) to operate and maintain, easily achieving an order-of-magnitude cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated database and mailing system updates via software are far cheaper per transaction than manual clerical labor, though initial integration and oversight costs remain. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (RPA platforms, document management systems with OCR, database automation tools, email/mail automation) reliably perform these tasks in production across many organizations. Some variability in data quality handling remains, but core functionality is mature and widely used. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like database management tools, RPA platforms, and AI-assisted CRM/ERP systems exist and are used in production, but end-to-end autonomous handling of mixed physical/digital systems is not yet fully mature. |
Review files, records, and other documents to obtain information to respond to requests.
71CI 67–75 · exposure 70 · augmentation 100 · importance 3.9/5 · click for rater detail
Review files, records, and other documents to obtain information to respond to requests.
71| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Information-intensive sectors (finance, legal support, healthcare administration, government) have rapidly adopted AI-driven document processing and retrieval in production. Enterprise adoption is accelerating as RPA and document-understanding platforms mature and integrate with existing workflows. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Clerical/admin functions across many industries are adopting AI search and copilot tools at a moderate pace, with pilots more common than full-scale deployment in smaller or less digitized organizations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI assistants meaningfully augment clerical workers by instantly surfacing relevant documents, highlighting key information, and auto-populating responses—transforming speed and accuracy while the clerk remains in control of final responses and quality assurance. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-powered search and summarization tools significantly speed up a clerk's ability to locate and synthesize information from files, even when human judgment on the final response remains necessary. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can reliably extract information from structured and semi-structured documents, perform document categorization, and retrieve relevant records with high accuracy, achieving significant time savings. However, complex ambiguous requests or documents with poor quality/unusual formats may still require human judgment, preventing a full 5-rating. |
| Task automatability | claude-sonnet-5 | 4/5 | Reviewing files/records to answer requests is largely information retrieval and synthesis, which current LLM+RAG systems can do well when documents are digitized and structured, meeting the 50% time-savings bar in many cases. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers prevent automating document review for routine information retrieval; most barriers are organizational (preference for human verification, change management) rather than compliance-based. Some sectors with strict audit trails may add oversight requirements but not hard licensing restrictions. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Some organizational friction and data privacy/access controls exist, but there's typically no licensing requirement or legal mandate that a human perform this specific lookup task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-powered document review and retrieval costs (cloud APIs, minimal integration overhead) are substantially cheaper than clerical labor per query or batch, often an order of magnitude lower when deployed at scale. Setup costs amortize quickly across high-volume requests. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Once documents are indexed, AI query/retrieval costs are pennies per request versus a clerk's loaded wage, though initial integration and data cleanup add cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (document processing APIs, RAG systems, enterprise search tools with ML) routinely perform document review and information retrieval in production environments at scale. Some edge cases and complex multi-document synthesis still require human oversight, but core functionality is mature and widely implemented. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Enterprise search, RAG chatbots, and document QA tools are deployed in production, but reliability drops with messy legacy records, scanned paper, or ambiguous requests, so scope is still narrower than full generality. |
Open, sort, and route incoming mail, answer correspondence, and prepare outgoing mail.
71CI 67–75 · exposure 70 · augmentation 75 · importance 3.9/5 · click for rater detail
Open, sort, and route incoming mail, answer correspondence, and prepare outgoing mail.
71| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large enterprises, logistics, and financial services already deploy mail sorting, routing, and automated response systems at scale; digitization of correspondence is rapid in information-sector organizations. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Office/administrative functions are adopting AI email and document tools at a moderate pace, with pilots and partial deployments common but full mailroom automation still uneven across firms. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI assists by pre-sorting mail, flagging urgent items, and drafting templates, raising human productivity on the portions requiring judgment or personalization, though augmentation is secondary to replacement opportunity here. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI drafting assistants, auto-categorization, and smart reply tools substantially boost productivity for the correspondence and routing decision-making portions of this task while a human remains in the loop for physical handling. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI can automate sorting and routing via OCR and document classification, and can draft responses to routine correspondence. Physical mail opening requires robotics, but digital mail workflows are largely automatable, yielding substantial time savings on a high-volume task. |
| Task automatability | claude-sonnet-5 | 4/5 | Sorting/routing physical mail requires manipulation, but answering correspondence and drafting outgoing mail is highly automatable with current AI text tools; the digital-heavy portion easily clears the 50% time-saving bar though physical mail handling remains manual. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing requirement or legal mandate for human involvement in mail handling. Main friction comes from customer preference for personalized responses and organizational inertia, but these are soft barriers with declining force. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal requirement mandates a human for this task, though organizational preference for personal touch in certain correspondence and physical handling creates mild friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Email classification, spam filtering, and correspondence drafting are near-zero marginal cost at scale; physical mail handling requires some infrastructure but remains far cheaper than human labor for high-volume sorting and routing. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | For the correspondence/drafting component, AI inference costs are far below clerk wages; however, physical mail handling still requires human labor, keeping the blended ratio below the top rating. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mail sorting and routing via machine learning, and basic correspondence drafting via LLMs, are deployed in production by courier services, logistics firms, and some enterprise email systems. Error rates on complex routing are non-trivial but the core functions are operationally proven. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Email triage, auto-categorization, and AI-drafted responses are deployed in many organizations, but full end-to-end handling of physical mail sorting and routing still requires human/robotic physical infrastructure not commonly deployed for general office clerks. |
Answer telephones, direct calls, and take messages.
67CI 50–84 · exposure 55 · augmentation 63 · importance 4.3/5 · click for rater detail
Answer telephones, direct calls, and take messages.
67| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large enterprises and service-heavy sectors (finance, healthcare, tech support) have already deployed IVR and call-routing automation at scale; smaller organizations are slower but the trend is clear and accelerating, particularly in high-volume inbound call centers. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Adoption of AI phone/receptionist tools is growing steadily across small and mid-size businesses but is not yet universal, with many offices still using humans or basic IVR. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-powered tools (call transcription, real-time call classification, suggested responses, automated routing) substantially amplify human operator productivity, allowing one person to manage more calls and focus on complex or sensitive interactions while the system handles triage and documentation. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist clerks by pre-screening calls, transcribing messages, and drafting call logs, improving efficiency even when a human remains involved for complex interactions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Current AI can transcribe voicemail and classify call intent, but end-to-end automation requires reliable call routing decisions, context-aware message taking, and real-time human judgment on call priority and sensitivity—capabilities that remain error-prone and require human oversight for complex or sensitive calls. |
| Task automatability | claude-sonnet-5 | 4/5 | AI voice agents and IVR/virtual receptionist systems can answer calls, route them appropriately, and log messages with substantial time savings over a human doing this full-time.:contentReference[oaicite:0]{index=0} |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal barriers exist to automating phone answering; however, customer preference for human contact, expectations of personalized service, and organizational resistance to perceived depersonalization create moderate friction to full adoption. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or legal requirement mandates a human answer phones or take messages; adoption is purely a business/customer-preference choice with minimal regulatory friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | IVR and AI call routing systems are relatively cheap to operate per call compared to full-time human telephone operators; however, integration, training, and error correction overhead partially offset savings, making AI moderately cheaper than hiring dedicated staff. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated phone answering/routing services cost a small fraction (often cents per call) compared to a human clerk's loaded wage for equivalent call volume. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | IVR systems and basic AI-driven call routing exist in production (e.g., phone trees, voicemail-to-text), but they handle routine cases only; human-equivalent performance on varied, dynamic call scenarios with accurate message capture and smart routing remains immature in commercial deployments. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Products like AI receptionists, call routing systems, and voicemail-to-text/message-taking tools are widely deployed in production for businesses today, though edge cases (ambiguous requests, complex routing) still cause errors. |
Collect, count, and disburse money, do basic bookkeeping, and complete banking transactions.
66CI 50–81 · exposure 67 · augmentation 75 · importance 4.1/5 · click for rater detail
Collect, count, and disburse money, do basic bookkeeping, and complete banking transactions.
66| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Information and finance sectors have deeply adopted automated accounting, payment processing, and banking transaction systems. Small businesses increasingly adopt cloud accounting platforms, and digital payment processing is near-universal. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Clerical/admin functions in many industries have adopted accounting automation tools at a middling pace, with pilots and partial deployment common but full replacement rare. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-powered accounting assistants, automated receipt scanning, and transaction categorization dramatically enhance productivity for human clerks managing complex accounts. The human remains essential for judgment and oversight, but AI transforms the speed and accuracy of the underlying work. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered bookkeeping software significantly speeds up data entry, categorization, and reconciliation, meaningfully boosting clerk productivity while humans still manage physical transactions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Money counting, basic bookkeeping, and banking transactions are highly structured, rule-based activities that AI and automated systems handle well. Current OCR, accounting software, and API integrations with banking systems can handle the vast majority of these tasks end-to-end, though collection and cash handling still require human involvement in some contexts. |
| Task automatability | claude-sonnet-5 | 3/5 | Basic bookkeeping and transaction recording can be substantially automated via accounting software and AI-assisted reconciliation, but physical cash handling and disbursement still require human presence and judgment for exceptions.: |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While there are no hard legal licensing barriers (unlike regulated accounting or audit roles), adoption is slowed by organizational inertia, need for reconciliation oversight, audit trails, and internal control requirements that typically keep a human in the loop for approval and verification. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Handling money often involves internal controls, dual-custody rules, and audit requirements that create moderate friction, though not strict professional licensure. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated systems cost orders of magnitude less per transaction than human labor; a typical general office clerk wage is $35k–$45k annually, while accounting automation via SaaS or integrated banking APIs costs pennies to dollars per transaction. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Software subscriptions for bookkeeping are cheap relative to clerk wages for that portion, but the physical cash-handling component still requires paid human labor, keeping blended cost roughly comparable. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature accounting software (QuickBooks, Xero), automated banking APIs, and cash-handling automation systems are deployed and reliable in production at scale across thousands of organizations. These products demonstrably perform bookkeeping, transaction recording, and fund transfer tasks reliably today. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like QuickBooks, Xero, and bank automation tools reliably handle bookkeeping entries and reconciliation, but physical money collection/disbursement is not something deployed AI products perform independently. |
Communicate with customers, employees, and other individuals to answer questions, disseminate or explain information, take orders, and address complaints.
60CI 59–61 · exposure 50 · augmentation 75 · importance 4.3/5 · click for rater detail
Communicate with customers, employees, and other individuals to answer questions, disseminate or explain information, take orders, and address complaints.
60| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Customer service and support teams across retail, finance, e-commerce, and tech are rapidly deploying chatbots and AI agents in production. Adoption is measurable and accelerating in digitized sectors, though physical and small-business service remains slower. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Customer service functions across many industries have adopted chatbots and virtual agents at a moderate pace, with pilots widespread but many general office clerk roles still relying primarily on human interaction. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assists human clerks by drafting responses, suggesting information, transcribing calls, and flagging priority complaints, significantly raising their throughput and accuracy. The human typically remains in the loop for judgment and final communication, making this a strong augmentation case. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools like canned response suggestions, chat co-pilots, and knowledge-base search meaningfully speed up how clerks answer questions and address complaints while the human retains control over final responses. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can handle routine inquiries, FAQs, and straightforward complaint categorization through chatbots and automated systems, but complex multi-turn conversations requiring nuanced judgment, empathy, or escalation logic still require human intervention. The task is partially automatable but not fully end-to-end at consistent quality. |
| Task automatability | claude-sonnet-5 | 3/5 | Chatbots and voice agents can handle routine FAQ, order-taking, and simple complaint intake, but escalations, emotional complaints, and ambiguous requests still require human judgment, so only partial time savings are realized without significant customization. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or licensing requirements mandate human communication for most general office clerk tasks, though some sectors (finance, healthcare) have compliance constraints. Customer preference for human contact and organizational concerns about brand risk create friction, but these are soft barriers rather than hard legal ones. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement blocks AI from answering questions or taking orders, though some customers prefer human contact and certain complaint handling may require accountable human sign-off in regulated contexts, creating moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference costs for handling routine inquiries are substantially lower than human labor (wages + benefits), especially at scale. Integration and oversight add cost, but the ratio still favors automation for high-volume, repetitive customer interactions. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI-driven chat/voice systems cost a fraction of a human clerk's loaded wage per interaction for routine queries, though ongoing integration, monitoring, and escalation handling add cost that keeps it below the top tier. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed chatbots and virtual assistants exist in production across many organizations (customer service, support lines), but they typically handle only a narrow subset of inquiries and transfer difficult cases to humans. Error rates and scope limitations prevent a 5-rating. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed customer service chatbots, IVR systems, and helpdesk AI are common in production, but they typically handle a subset of queries and route complex or sensitive matters to humans, indicating narrower reliable scope than full task coverage. |
Count, weigh, measure, or organize materials.
59CI 35–84 · exposure 58 · augmentation 38 · importance 3.2/5 · click for rater detail
Count, weigh, measure, or organize materials.
59| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Logistics, warehousing, and supply-chain sectors have rapidly deployed automated counting and weighing systems; adoption is visible and accelerating in high-volume operations. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Office clerk roles are in a low-digitization, physical-task-heavy sector where AI adoption for material handling remains slow compared to information-processing tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-assisted tools (real-time inventory dashboards, pick-and-pack guidance, anomaly detection) provide useful support to clerks when full automation is not deployed, raising accuracy and speed on portions of the workflow. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI-enabled tools like smart scales or inventory scanners can assist with counting accuracy, but the core physical organizing and handling still requires human effort with limited AI augmentation. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Counting, weighing, measuring, and organizing materials are routine, repetitive, physical-world operations that can be fully automated with computer vision, robotic handling systems, and sensor-based measurement—many of which already deliver >50% time savings at equal or better quality in warehouses and manufacturing. |
| Task automatability | claude-sonnet-5 | 2/5 | This is fundamentally a physical manipulation task requiring hands-on handling of materials, which current AI cannot perform without robotics; software alone cannot count/weigh/measure physical items., |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers prevent automation; the main friction is initial capital investment and organizational inertia, not licensing or liability asymmetry. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Few regulatory or licensing barriers exist for this task, though physical workspace constraints and equipment costs create moderate friction to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Automated systems for counting, weighing, and organizing materials typically cost much less per unit processed than human labor once amortized, especially at scale in warehouses and distribution centers. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Specialized automated counting/weighing equipment can be cost-effective at scale, but general-purpose AI systems are not the driver here; integration costs for physical automation often exceed simple human labor for small-scale tasks. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (e.g., automated counting systems, robotic arms with weighing sensors, barcode/RFID inventory systems) reliably perform these tasks in production across logistics and manufacturing, though the degree of automation depends on material type and context. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some sensor-based or vision-based counting/weighing systems exist (e.g., smart scales, barcode scanners) but they are narrow-purpose tools, not general AI products replacing this broad clerical task. |
Operate office machines, such as photocopiers and scanners, facsimile machines, voice mail systems, and personal computers.
55CI 35–75 · exposure 50 · augmentation 38 · importance 4.6/5 · click for rater detail
Operate office machines, such as photocopiers and scanners, facsimile machines, voice mail systems, and personal computers.
55| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Office automation has been widespread for over a decade; RPA and document management are in production across information, finance, and professional services sectors. Adoption is steadily increasing as systems mature and integration costs fall. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | General clerical work is in a low-digitization, slower-adopting segment; physical machine operation sees minimal AI-driven change in practice. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by suggesting which documents to scan, prioritizing print jobs, or managing voice mail queues, but the core task is largely routine procedural work where augmentation is less transformative than in judgment-heavy roles. |
| Augmentation potential | claude-sonnet-5 | 2/5 | AI can assist with digital file management or voicemail transcription but offers little enhancement to the core physical machine operation. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Most routine machine operations (photocopying, scanning, faxing, basic PC tasks) can be partially or fully automated via RPA, APIs, and document management systems, achieving significant time savings. However, physical device operation (loading paper, clearing jams) and context-dependent decisions about which documents to copy/scan may require human oversight, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 2/5 | Physical operation of copiers, fax machines, and scanners requires manual handling of paper and hardware that AI software cannot perform; only the digital/software portions (e.g., PC-based tasks) are automatable., so overall time savings is limited. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or licensing barriers exist for automating these routine office operations. Main friction comes from organizational change management, legacy system integration, and employee preferences, rather than legal or liability constraints. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal barriers, but physical presence and manual dexterity requirements create a practical barrier to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | RPA and document automation are cost-effective once deployed; inference and integration costs are low relative to the loaded wage of a general office clerk performing these routine tasks at scale. Cost per task is typically a small fraction of human labor. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Since the task is largely physical/manual, there's no AI substitute to compare cost against; a human is still required to operate hardware, so AI offers little cost advantage here. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (RPA platforms, document management systems, print-management software) reliably automate many of these operations in production environments, particularly for scanning, faxing, and voice mail integration. Physical machine interaction remains partially manual, and integration complexity varies by legacy infrastructure. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed AI product operates physical office machines; robotics for this narrow task is not commercially deployed in typical offices. |
Train other staff members to perform work activities, such as using computer applications.
34CI 30–39 · exposure 25 · augmentation 75 · importance 3.3/5 · click for rater detail
Train other staff members to perform work activities, such as using computer applications.
34| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While AI-generated training content is increasingly used, end-to-end AI replacement of trainer roles remains limited; most organizations are in pilot or supplementary phases rather than full adoption of AI as primary trainers. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | General office clerk roles are in a sector with slower AI adoption for interpersonal training tasks compared to fully digital knowledge-work tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI is highly effective at assisting human trainers by generating lesson plans, creating explanations, producing practice materials, and adapting content on demand, significantly boosting trainer productivity while the human retains responsibility for delivery and relationship. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly help trainers by generating instructional materials, quizzes, and step-by-step guides, improving efficiency while the human still delivers the training. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can generate training materials and explanations, but delivering effective training requires adapting to individual learner needs, answering unexpected questions, and providing personalized feedback—tasks requiring real-time human judgment and contextual understanding that current systems handle poorly at scale. |
| Task automatability | claude-sonnet-5 | 2/5 | Training others involves live demonstration, adapting to trainee questions, and interpersonal coaching that current AI cannot fully replicate end-to-end, though AI can generate training materials.of the task itself. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Organizations often prefer human trainers for accountability and relationship-building, and there is modest friction from need for oversight and verification that training was effective; however, no hard legal requirement mandates human delivery. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational preference for human trainers who can answer follow-up questions and model workflows creates moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Training materials generation is inexpensive via AI, but the total cost including integration into workflows, oversight, and remediation of poorly trained staff often exceeds the cost of a human trainer, especially for small cohorts. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-generated training content is cheap to produce, but the actual delivery and troubleshooting with trainees still requires human time, keeping overall cost roughly comparable. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI can produce training content and tutorials, deployed products for interactive staff training lack the reliability, responsiveness, and adaptive capability needed for consistent real-world training delivery; most organizations still rely on humans for this task in production. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products like AI-generated tutorials, chatbots, and documentation tools exist but are not deployed as substitutes for hands-on peer training in most offices. |
Troubleshoot problems involving office equipment, such as computer hardware and software.
34CI 30–38 · exposure 25 · augmentation 75 · importance 3.0/5 · click for rater detail
Troubleshoot problems involving office equipment, such as computer hardware and software.
34| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Many organizations have deployed AI chatbots and ticket-routing systems for help desk functions, but actual displacement of human troubleshooters remains limited; adoption is steady in large enterprises but slow in small and medium firms. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | General office/clerical settings adopt AI slowly compared to specialized IT or software sectors, and this task is often bundled with in-person duties reducing AI uptake. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments human troubleshooters by providing instant access to knowledge bases, suggesting diagnostic steps, and handling routine inquiries—allowing clerks to focus on complex or escalated issues and improving response time. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI chat assistants and diagnostic tools substantially help clerks look up solutions, generate troubleshooting steps, and resolve simple software issues faster, meaningfully boosting productivity while humans still handle physical aspects. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Current AI can assist with basic troubleshooting steps and documentation lookup, but most office equipment problems require hands-on diagnosis, hardware inspection, and context-dependent solutions that AI cannot reliably perform end-to-end. The task often requires physical interaction and judgment beyond 50% automation. |
| Task automatability | claude-sonnet-5 | 2/5 | Basic software troubleshooting can be guided by AI chat assistants, but hands-on hardware diagnosis, physical repairs, and unpredictable equipment issues require in-person action AI cannot perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Organizations often prefer human contact for technical support due to liability concerns (data access, system changes) and the need for accountability; however, no legal licensing requirement exists, creating moderate but not hard barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational reliance on physical access to equipment and preference for a known point-of-contact creates moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-powered help desk tools reduce some clerk labor but require human oversight, integration setup, and fallback to skilled technicians for complex issues. The all-in cost remains comparable to or higher than employing a general office clerk for mixed duties including troubleshooting. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI-assisted troubleshooting tools are cheap per query, the need for physical intervention and escalation to human technicians means overall cost savings versus a generalist office clerk are limited. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While chatbots and AI support tools exist for common software issues, they produce material error rates on non-standard problems and lack the ability to physically diagnose hardware faults or integrate with legacy systems many offices still use. Deployed products handle only narrow, routine scenarios. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI chatbots and IT helpdesk copilots exist and can triage common software issues, but reliable resolution of hardware and mixed office equipment problems still typically needs human IT support in most organizations. |
Monitor and direct the work of lower-level clerks.
28CI 25–30 · exposure 25 · augmentation 50 · importance 3.7/5 · click for rater detail
Monitor and direct the work of lower-level clerks.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While workflow management and task-tracking tools have spread, actual delegation of supervisory decision-making to AI remains rare in practice. Most organizations maintain human supervisors; adoption of AI-directed clerk work is nascent and not yet embedded in mainstream administrative operations. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | General clerical and administrative sectors show slower AI adoption for supervisory functions compared to fast-moving finance or tech sectors; pilots for workflow automation exist but managerial oversight lags. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist supervisors by highlighting work bottlenecks, tracking task completion, and surfacing performance metrics, helping them direct efforts more efficiently. However, the augmentation is partial—the human supervisor remains central to judgment, motivation, and interpersonal direction. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help by tracking task completion, flagging errors, and summarizing performance data, aiding supervisors but not replacing their judgment or interpersonal direction. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Monitoring work and directing lower-level staff requires real-time judgment about human performance, motivation, and task prioritization—tasks where current AI lacks reliable situational awareness and authority to direct humans. Some components like flag-checking completed tasks or suggesting workload distribution could be partially automated, but the full supervisory and interpersonal direction function remains primarily human. |
| Task automatability | claude-sonnet-5 | 2/5 | Supervising and directing junior staff requires interpersonal judgment, motivation, coaching, and accountability that current AI cannot fully replicate end-to-end. AI can support scheduling and task tracking but not the core supervisory function. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Organizational hierarchy, employment law, and accountability norms strongly prefer a human manager to sign off on personnel decisions, work direction, and performance feedback. Liability and organizational culture create meaningful friction against full automation of supervisory functions. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but organizational structure typically vests supervisory authority and accountability in a human role, creating moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI monitoring and dashboarding tools are available but typically require human interpretation and decision-making to remain valuable. The all-in cost of adequate AI oversight plus human supervisory review often approaches or exceeds the loaded wage of a junior supervisor performing the task directly. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While software for task assignment is cheap, real supervisory functions (feedback, correction, motivation) still require a human, so AI alone isn't a comparable substitute yielding major cost savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While analytics dashboards can track work progress and some workflow systems offer task-assignment features, no deployed product reliably performs the core function of directing human subordinates with the contextual judgment and accountability expected of supervision. Existing tools assist but do not replace the supervisor. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some workflow/task-management tools exist that assign and track work, but no deployed product independently 'directs' human clerks with judgment-based oversight in production at scale. |
Deliver messages and run errands.
14CI 5–24 · exposure 8 · augmentation 13 · importance 3.5/5 · click for rater detail
Deliver messages and run errands.
14| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Adoption of automation for general office messenger tasks remains minimal; this is a low-digitization, human-contact role in traditionally conservative office environments with few pilot deployments. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Physical errand-running is not being displaced by AI in any sector; this remains entirely human-performed with no meaningful automation trend. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI can assist with route optimization and message scheduling, but the core task of physical delivery requires human presence, so augmentation value is limited to planning and logistics layers rather than the delivery act itself. |
| Augmentation potential | claude-sonnet-5 | 1/5 | AI offers no meaningful assistance for the physical acts of walking, carrying, or delivering; there's no software augmentation pathway for this specific physical task. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Physical delivery and errands require navigation, interaction, and manipulation in uncontrolled environments. Current AI systems cannot reliably perform door-to-door delivery, locate recipients, handle unexpected obstacles, or manage the physicality of errands at human speed and safety standards today. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires physical presence and movement in the real world (delivering items, walking to locations), which current AI systems cannot perform; software agents have no embodiment to execute errands.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Physical presence in real-world spaces, liability for lost or mishandled messages, customer expectation of human interaction, and organizational preference for human judgment on sensitive deliveries create substantial barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No regulatory or licensing barrier exists, but the physical nature of the task itself is the barrier—it's a practical/technical limitation rather than legal or organizational friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Automation of general message delivery and errands would require mobile robots or autonomous agents with significant capital cost, maintenance, and oversight, far exceeding the loaded wage of a part-time or entry-level office clerk. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute for this physical task, so any comparison to human cost is moot—AI cannot perform it at any price today. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed commercial product reliably automates the full cycle of message delivery and errand-running as a general office task. Specialized robotics and autonomous systems exist in narrow contexts (warehouse delivery, fixed routes) but not for general office messenger work. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product physically delivers messages or runs errands in an office setting; this remains outside the capability of software-based AI entirely. |
Related occupations — Office & Administrative Support
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.