Correspondence Clerks
43-4021.00Median wage $46,800/yr4,290 employed (US)Rank #11 of 923 scored · top 1% by substitution
Compose letters or electronic correspondence in reply to requests for merchandise, damage claims, credit and other information, delinquent accounts, incorrect billings, or unsatisfactory services. Duties may include gathering data to formulate reply and preparing correspondence.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution73
Exposure72
Augmentation74
Substitution — the headline: capability discounted by cost, barriers and adoption.
Exposure — technical capability alone, regardless of whether anyone deploys it.
Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.
Tasks on the substitution scale
17 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
65%
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%75
panel mean rating 4.0/5 → substitution pressure 75/100
Technical feasibility todayw 20%67
panel mean rating 3.7/5 → substitution pressure 67/100
Cost vs. human wagew 15%83
panel mean rating 4.3/5 → substitution pressure 83/100
Adoption barriersw 20%inverted — strong barriers lower the score73
panel mean rating 2.1/5 (barrier strength) → substitution pressure 73/100
Sector adoption velocityw 10%60
panel mean rating 3.4/5 → substitution pressure 60/100
Task breakdown (17 tasks)
Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.
Route correspondence to other departments for reply.
92CI 87–97 · exposure 95 · augmentation 75 · importance 3.5/5 · click for rater detail
Route correspondence to other departments for reply.
92| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Email and document management systems have broadly adopted auto-routing and classification features across finance, government, customer service, and professional services sectors. Adoption is mature and widespread, though some smaller or traditional organizations lag. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Office/administrative workflow automation is widely adopted in corporate settings, with routing and triage systems already standard in customer service and back-office operations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI routing systems assist human clerks by pre-categorizing, suggesting department assignments, and flagging uncertain cases for review, substantially accelerating their work while maintaining quality control through human verification where needed. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-assisted routing tools help remaining human clerks triage exceptions and ambiguous cases faster, improving throughput even where full automation isn't used. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | This task is highly automatable: AI can read incoming correspondence, classify it by topic/department, and route it with rule-based logic or ML classifiers. Modern email systems already implement similar routing; achieving 50% time savings with equal quality is straightforward for a task requiring no subjective judgment or specialized knowledge. |
| Task automatability | claude-sonnet-5 | 5/5 | Routing correspondence based on content classification is a straightforward text-classification and workflow-routing task that automated systems handle end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or licensing barriers exist for automated mail routing. Some organizations may prefer human oversight for sensitive correspondence, and legacy systems may require integration work, but no legal requirement mandates human involvement in routing alone. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, legal, or human-contact requirement governs internal mail/correspondence routing; it's a purely administrative function with minimal friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated routing costs pennies per message in inference and integration overhead, while a correspondence clerk's loaded wage is $30–50k annually. The cost ratio favors automation by at least an order of magnitude. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated routing via rules engines or ML classifiers costs a tiny fraction of a clerk's wage per item processed, especially at volume. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Production-grade email routing systems using ML have existed for years in enterprises (spam filters, auto-categorization, ticket routing). Tools like Zapier, UiPath, and native email rule engines demonstrate reliable performance at scale, with minor configuration needed per organization. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Email routing, ticketing systems, and workflow automation tools (e.g., Zendesk, ServiceNow, rule-based and ML classifiers) reliably route correspondence in production across many organizations, though edge cases still need occasional human correction. |
Type acknowledgment letters to persons sending correspondence.
88CI 79–97 · exposure 87 · augmentation 75 · importance 3.5/5 · click for rater detail
Type acknowledgment letters to persons sending correspondence.
88| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Email and document automation tools are widely adopted in information-sector and administrative contexts. Office/administrative processes have shown strong AI adoption momentum, with many organizations already using templated or AI-assisted correspondence systems. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Clerical/administrative correspondence functions in offices and services sectors have seen fast adoption of automated templating and AI drafting tools already embedded in common software. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can draft acknowledgment letters that clerks review and customize, substantially reducing typing time and allowing humans to focus on complex correspondence. This assistive use case is already common in email and document management tools. |
| Augmentation potential | claude-sonnet-5 | 4/5 | Even where a human stays involved for review or personalization, AI drastically speeds up drafting these routine letters, giving strong productivity gains. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Typing standardized acknowledgment letters is straightforward text generation that current AI systems can handle reliably. LLMs can generate professional acknowledgment letters from minimal context (sender info, date) with 50%+ time savings compared to manual typing, though some human review and customization may be needed for quality assurance. |
| Task automatability | claude-sonnet-5 | 5/5 | Drafting standardized acknowledgment letters is a templated text-generation task that current LLMs handle fully, easily exceeding 50% time savings at equal or better quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While some organizations may require human review or sign-off for outbound correspondence, there are no strict legal barriers preventing automated acknowledgment letter generation. Organizational friction and quality-control preferences create minor barriers, but most modern workflows can adopt this readily. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, liability, or human-contact requirement attaches to typing acknowledgment letters; it's a purely administrative task with no legal barrier to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | API inference costs for generating a short acknowledgment letter are negligible (often <$0.001), versus a correspondence clerk's hourly wage ($25-35k/year loaded). AI cost is orders of magnitude lower. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Generating a form letter via AI costs fractions of a cent per instance versus paying a clerk's wage for manual typing, an order-of-magnitude or greater cost advantage. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Email and document generation systems are deployed at scale across organizations. GPT, Claude, and similar tools integrated into mail handling workflows can generate acknowledgment letters reliably in production environments today. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature email/document automation and AI writing tools already generate and send acknowledgment correspondence in production CRM and office systems, though some integration setup is needed. |
Maintain files and control records to show correspondence activities.
88CI 84–92 · exposure 83 · augmentation 63 · importance 4.1/5 · click for rater detail
Maintain files and control records to show correspondence activities.
88| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Administrative and clerical automation has seen rapid adoption in information-intensive sectors (finance, legal, government, insurance). Document management and workflow automation are now standard practice, with many organizations already deploying RPA for similar back-office record-control tasks. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Administrative and clerical functions across office-based sectors have seen fast digitization and adoption of automated records/document systems over the past decade. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist clerks by suggesting file locations, flagging miscategorizations, and auto-populating metadata fields, improving their productivity and accuracy. However, the task itself is primarily data-wrangling rather than judgment-intensive, limiting the depth of augmentation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-powered document management tools significantly speed up filing, categorization, and retrieval, letting clerks focus on exceptions and oversight. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | File maintenance and record control for correspondence can be largely automated through document management systems, OCR, metadata tagging, and RPA workflows. Modern systems can extract information from emails/letters, organize by sender/date/subject, and flag activities—saving well over 50% of manual time at equivalent accuracy. |
| Task automatability | claude-sonnet-5 | 5/5 | Filing and control record maintenance for correspondence is a structured data-management task that off-the-shelf document management and workflow systems already handle with substantial time savings. |
| Adoption barriers | claude-haiku-4-5-20251001 | 1/5 | No legal licensing, liability constraint, or mandatory human sign-off applies to file maintenance itself. Most organizations freely adopt document management and RPA; the main friction is internal IT/change management, not regulatory or authorization barriers. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, liability, or human-contact requirements protect this administrative record-keeping task from automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Once integrated, automated systems operate at marginal inference cost (pennies per document) plus modest oversight, far cheaper than a clerk's fully loaded wage for the same volume of records maintained and filed. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated indexing, tagging, and tracking software costs a small fraction of a clerk's salary for equivalent volume of records processed. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed enterprise document management and RPA solutions (e.g., SharePoint, Salesforce, UiPath) reliably automate filing, indexing, and activity tracking in production across organizations. Error rates are low for structured metadata and categorization, though edge cases (ambiguous correspondence) may require human review. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature document management, CRM, and records-tracking software (e.g., SharePoint, DMS platforms with automated logging) are deployed at scale in production, though full end-to-end control-record accuracy still needs occasional human verification. |
Compose letters in reply to correspondence concerning such items as requests for merchandise, damage claims, credit information requests, delinquent accounts, incorrect billing, or unsatisfactory service.
83CI 79–87 · exposure 83 · augmentation 100 · importance 3.8/5 · click for rater detail
Compose letters in reply to correspondence concerning such items as requests for merchandise, damage claims, credit information requests, delinquent accounts, incorrect billing, or unsatisfactory service.
83| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large enterprises (retail, financial services, utilities, customer service centers) are actively deploying AI email/letter generation in production. Adoption is deep in information and service sectors; mid-market and SMBs are in active pilot phases with measurable displacement of data-entry and junior clerk roles. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service and back-office administrative functions across finance, retail, and insurance have seen fast, deep AI adoption for templated written communications. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI-assisted composition dramatically accelerates clerk productivity: systems draft replies, clerks review and personalize, and quality often exceeds manual composition alone. This human-in-the-loop augmentation is already in widespread use and measurably raises throughput and consistency. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI drafting tools dramatically speed up composing routine reply letters while clerks retain review and personalization control, a strong augmentation case. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI systems can generate routine reply letters to standard correspondence (merchandise requests, billing inquiries, account issues) with high quality and minimal human intervention. Templates and language models handle ~70-80% of typical volume, though complex disputes or sensitive claims may need human review, falling short of full end-to-end automation at the 50% time-saving threshold for all cases. |
| Task automatability | claude-sonnet-5 | 5/5 | LLMs can draft accurate, context-appropriate correspondence replies from templates and case details with minimal editing, easily exceeding 50% time savings.text generation for structured business correspondence is a core LLM strength. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist; most correspondence requires no licensed professional sign-off and organizations face minimal liability if an AI generates routine replies. Primary friction is organizational preference for human quality control and customer sentiment, not hard authorization requirements. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for drafting correspondence, though some organizations require human sign-off for legal/financial commitments like credit or billing disputes, creating mild oversight friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI inference cost for generating a letter is negligible (fractions of a cent), compared to a clerk's fully-loaded wage (typically $30–50k/year, or $15–25/hour). Even accounting for oversight and integration, the cost ratio is 10–50x in AI's favor. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Generating a drafted letter via LLM API costs fractions of a cent versus minutes of clerical wage time, an order-of-magnitude or greater cost advantage. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature AI writing tools (GPT-4, Claude, specialized email automation software) reliably generate professional correspondence at scale in production environments. Products like Salesforce Einstein and dedicated customer service AI already deploy this capability, though some organizations still require human sign-off on output, indicating near-full but not universal reliability. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed customer service and CRM platforms (e.g., Gmail Smart Reply, Zendesk AI, Salesforce Einstein) already generate reply drafts for billing, complaints, and claims at scale, though human review remains common for edge cases. |
Complete form letters in response to requests or problems identified by correspondence.
83CI 79–87 · exposure 83 · augmentation 88 · importance 3.4/5 · click for rater detail
Complete form letters in response to requests or problems identified by correspondence.
83| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Financial services, insurance, and large administrative/public organizations have been automating correspondence for years. RPA and AI-based systems are moving from pilots to production at scale in digitized back-office environments. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service and back-office administrative functions across many industries have rapidly adopted AI-generated correspondence and templated response tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can draft form letters in real time as a clerk works, suggesting appropriate templates and fills in details, significantly boosting throughput. The human remains in the loop to review tone, accuracy, and fit before sending. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI drastically speeds up drafting of form responses, letting clerks review/edit rather than write from scratch, a clear productivity transformation while retaining human oversight. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Form letter completion is highly automatable. Current AI can reliably generate appropriate responses to standardized requests using templates and context from correspondence. LLMs can match request types to response templates and fill in variable fields, achieving >50% time savings with minimal setup. |
| Task automatability | claude-sonnet-5 | 5/5 | Filling in form letters based on identified requests/problems is a templated text-generation task that LLMs handle end-to-end with substantial time savings at equal or better quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist for automated form letters; oversight is primarily internal quality assurance. Some organizations maintain human sign-off for liability, but no licensing requirement mandates human involvement, making substitution straightforward. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for this clerical task, though some organizational review or approval before sending customer-facing letters creates minor friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI inference cost per letter is negligible (cents), while a correspondence clerk's loaded wage per completed letter is typically $5–15. Even including integration and light human oversight, AI cost is orders of magnitude lower. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated drafting/generation of routine correspondence costs a small fraction of a cent per letter versus a clerk's loaded wage, an order-of-magnitude or greater saving. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Production systems like enterprise AI writing assistants and template-based automation tools are deployed in customer service, banking, and administrative settings. Reliability is high for routine correspondence, though complex or unusual cases still show material error rates. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature products (customer service AI, email automation platforms, CRM-integrated generative tools) already draft and send such correspondence in production, though some scope remains for edge cases needing escalation. |
Process orders for goods requested in correspondence.
81CI 75–87 · exposure 83 · augmentation 63 · importance 3.9/5 · click for rater detail
Process orders for goods requested in correspondence.
81| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Retail, e-commerce, and logistics sectors have already deployed RPA and intelligent automation for order processing at significant scale. Adoption is measurable and accelerating, particularly among digitized, volume-driven enterprises where cost pressure is highest. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Order processing automation is deeply adopted in retail, logistics, and administrative support sectors via ERP/CRM-integrated AI tools, reflecting fast, broad uptake in transactional back-office functions. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI assists clerks by pre-filling order forms, flagging anomalies (missing data, out-of-range quantities), and suggesting product matches from ambiguous descriptions. This reduces manual effort and error-checking time, though the human typically remains in the loop for exceptions and quality control. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially speeds up correspondence review, data extraction, and order entry for clerks handling exceptions or verification, meaningfully raising throughput while humans remain for edge cases. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Order processing from correspondence is highly structured and repetitive: extracting product names, quantities, customer details, and payment information from emails or letters, then entering them into order systems. Current AI can parse unstructured text, extract key entities, and populate databases with 50%+ time savings compared to manual data entry, often with equal or better accuracy. |
| Task automatability | claude-sonnet-5 | 4/5 | Order processing from correspondence (email, forms) is highly structured data extraction and entry that current AI systems can handle end-to-end with substantial time savings, though edge cases and exceptions still need human review. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist; order processing is not a licensed profession and carries low liability risk. Main friction is organizational inertia, legacy system integration, and human preference for oversight—but these are surmountable without hard institutional blockers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement exists for this clerical task, though some organizational friction around data accuracy, system integration, and customer trust in automated confirmations exists. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | The cost of inference for text parsing and API calls to populate order systems is negligible per transaction—cents to fractions of a cent—compared to a correspondence clerk's fully loaded wage (typically $35k–$45k annually, or ~$20+/hour). AI cost is orders of magnitude lower. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated order processing systems cost a small fraction of a clerk's loaded wage per transaction once integrated, though initial integration and exception-handling oversight add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature RPA and AI-powered document processing products (e.g., intelligent document capture, OCR+extraction APIs) reliably handle order data extraction and entry in production environments. Some error rates persist with ambiguous or handwritten correspondence, but most structured digital orders are processed reliably at scale by deployed systems. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed order-management and RPA/AI systems with NLP extraction are widely used in production for processing purchase orders and customer requests, though full autonomy without human checks is less common for exceptions. |
Review correspondence for format and typographical accuracy, assemble the information into a prescribed form with the correct number of copies, and submit it to an authorized official for signature.
81CI 76–85 · exposure 83 · augmentation 75 · importance 3.7/5 · click for rater detail
Review correspondence for format and typographical accuracy, assemble the information into a prescribed form with the correct number of copies, and submit it to an authorized official for signature.
81| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Document processing automation is gaining traction in government, finance, and larger organizations, but adoption remains mixed; many legacy operations still rely on manual correspondence handling. Pilots are common in digitized sectors, but deep production deployment in this specific task is not yet widespread. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Clerical and administrative support functions are adopting AI tools like document assistants and grammar checkers at a moderate pace, with pilots and partial deployments more common than full-scale substitution. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist clerks by flagging suspected errors, suggesting corrections, and auto-populating form fields, meaningfully raising accuracy and speed. However, the task itself is largely mechanical and rule-bound, so augmentation is less transformative than for judgment-heavy tasks. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools substantially speed up proofreading and template assembly for clerks who remain responsible for final review and submission, making this a strong augmentation case. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Modern OCR and document processing systems can reliably detect format issues, typographical errors, and verify document structure. Assembly into prescribed forms can be automated with workflow logic. The main remaining friction is routing to an authorized official, which requires human discretion rather than automation of the clerk's core checking and assembly work. |
| Task automatability | claude-sonnet-5 | 5/5 | Reviewing text for format/typos and assembling it into a template is a well-bounded text task that current AI (grammar/formatting checkers, LLM drafting tools) can do end-to-end with substantial time savings at equal or better quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Submission to an authorized official is typically an authorization/approval gate rather than a legal requirement that the clerk must be human. No licensing requirement applies to the clerk role itself. Organizational friction exists (legacy systems, preference for human review) but is surmountable. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for formatting/proofreading, though the final submission to an authorized official for signature preserves a human-in-the-loop step that creates mild organizational friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Cloud-based document processing and RPA inference costs are fractional (cents per document), versus a correspondence clerk's fully loaded wage (~$40–50k/year). Even accounting for integration and oversight, the ratio strongly favors AI automation. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated proofreading and document assembly software costs a small fraction of a clerk's hourly wage for equivalent throughput, especially at scale. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products in document automation and RPA (e.g., UiPath, Blue Prism, and specialized compliance platforms) demonstrably perform format verification, copy assembly, and metadata validation at scale in production. Error rates are low for well-defined templates, though complex or novel formats may require oversight. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed products like Grammarly, Microsoft Editor, and LLM-based document assembly tools reliably handle proofreading and template population in production, though full workflow integration with signature routing varies by organization. |
Read incoming correspondence to ascertain nature of writers' concerns and to determine disposition of correspondence.
79CI 79–79 · exposure 75 · augmentation 88 · importance 4.0/5 · click for rater detail
Read incoming correspondence to ascertain nature of writers' concerns and to determine disposition of correspondence.
79| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Email routing and triage automation are already widespread in information-sector organizations, SaaS platforms, and customer service. Deployment is common in medium to large enterprises; early-stage but accelerating in smaller firms. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Administrative/clerical functions in finance, insurance, and customer service sectors have rapidly adopted AI-based email triage and classification tools, reflecting fast adoption patterns in white-collar back-office work. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assists human clerks by pre-sorting, tagging, and suggesting disposition, allowing humans to focus on edge cases and judgment calls. This augmentation significantly speeds classification and reduces cognitive load while humans retain final oversight. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI significantly speeds up reading, summarizing, and categorizing correspondence for human clerks, who can then focus on complex judgment calls or exceptions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can reliably classify incoming correspondence by topic/concern and route to appropriate departments with >50% time savings. LLMs excel at understanding context, intent, and categorical sorting of unstructured text; minimal human intervention needed for most correspondence. |
| Task automatability | claude-sonnet-5 | 4/5 | LLMs can read and classify correspondence, extract intent, and route/triage it with high accuracy for most standard business correspondence categories, meeting the time-saving bar for routine cases. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or licensing barriers exist; this task does not require signed authority or human accountability in most cases. Some organizations prefer human review for sensitive complaints or legal matters, but organizational friction is low and easy to override. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Some correspondence (legal, medical, regulatory complaints) may require human review for liability reasons, but most disposition/triage work has no licensing or legal requirement for human handling. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Inference cost for reading and classifying a document is negligible ($0.001–0.01 per email), while a correspondence clerk's loaded wage is ~$30–40/hour. AI is at least two orders of magnitude cheaper per task. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated text classification and intent extraction via API calls costs a small fraction of a cent per item versus a clerk's per-item loaded labor cost, an order-of-magnitude or greater savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed email filtering, routing, and triage systems (including those powered by LLMs) perform this task in production at scale for customer service, help desks, and back-office operations. Error rates are acceptable for non-critical routing, though some organizations still maintain human review for sensitive cases. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed products (customer service email triage, ticketing systems with AI classification, inbox routing tools) reliably perform this today across many industries, though edge cases and ambiguous requests still need human review. |
Compute costs of records furnished to requesters, and write letters to obtain payment.
77CI 67–87 · exposure 78 · augmentation 75 · importance 3.6/5 · click for rater detail
Compute costs of records furnished to requesters, and write letters to obtain payment.
77| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Financial services, government, and utilities—major users of correspondence clerks—are actively adopting RPA and AI-driven letter generation for billing and payment collection. Adoption in back-office automation is rapid and measurable. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Administrative/clerical functions in many organizations are adopting automation and AI-assisted document processing, but adoption is uneven and often mid-stage rather than fully deployed. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can draft payment request letters and pre-compute costs, allowing a clerk to review, edit, and handle exceptions far faster than manual composition. This assistance remains valuable even where human oversight is desired for complex or disputed cases. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can significantly speed up drafting of payment letters and assist with calculating costs from structured records, greatly boosting clerk productivity while they verify accuracy. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | This task is highly automatable: AI can retrieve record costs from databases or systems, compute totals, and generate appropriate payment request letters using templates. The entire workflow—cost calculation plus letter composition—meets the 50% time-saving threshold with current systems. |
| Task automatability | claude-sonnet-5 | 4/5 | Cost computation from records and drafting standard payment-request letters follow well-defined rules and templates, which AI can handle end-to-end with human review for edge cases. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Minimal legal or regulatory barriers exist for automating cost computation and payment request letters; no licensing requirement mandates human signature on routine billing correspondence. Primary friction is organizational (override for complex disputes, audit practices), not legal. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for this clerical task, though organizations may want human review to avoid billing errors or customer relations issues, creating mild friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI cost per letter (inference + template execution + database lookup) is a fraction of a cent per task, while correspondence clerk labor at typical loaded wages ($35–50k annually) amounts to several dollars per letter. AI is orders of magnitude cheaper. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automating routine cost calculation and letter drafting via AI/workflow tools is far cheaper than paying a clerk per-letter, though some integration and oversight costs remain. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products exist in business process automation, document generation, and financial letter-writing (e.g., AP automation, template-based correspondence systems). Some integration and data access setup is needed, but production systems are handling similar workflows at scale in organizations. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI drafting tools and workflow automation for billing correspondence exist and are used in some organizations, but full integration with records systems and payment computation varies in reliability across deployments. |
Prepare documents and correspondence, such as damage claims, credit and billing inquiries, invoices, and service complaints.
76CI 76–76 · exposure 75 · augmentation 88 · importance 4.0/5 · click for rater detail
Prepare documents and correspondence, such as damage claims, credit and billing inquiries, invoices, and service complaints.
76| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Finance, insurance, and large enterprises have rolled out document automation in pilots and some production deployments, but adoption remains spotty and slower than in knowledge-work roles like coding. Smaller firms and legacy-heavy organizations lag significantly. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Administrative/clerical functions in service and financial sectors are adopting AI drafting tools at a moderate pace, with pilots more common than full production deployment in many organizations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at assisting clerks by auto-populating fields, suggesting response templates, and flagging missing information, dramatically raising human throughput. A clerk working with AI assistance can process several times more correspondence per day while retaining oversight. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI drafting assistants substantially speed up composing these routine documents while clerks retain review and personalization control. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Document preparation for routine correspondence (damage claims, billing inquiries, invoices, complaints) is highly amenable to automation. Current AI systems can extract relevant information, populate templates, and generate complete documents; this task would easily meet the 50% time-saving threshold with minimal human setup, though final review for accuracy and compliance would still be advisable. |
| Task automatability | claude-sonnet-5 | 4/5 | Drafting standardized correspondence like claims, invoices, and complaint responses from templates and data inputs is well within current LLM capabilities, especially with structured input data.4Manual review remains needed but most drafting time can be saved. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers prevent automation of routine correspondence preparation. The main friction is organizational (preference for human review, internal policy requiring sign-off) and quality assurance needs, but these are light compared to regulated professions like law or medicine. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for this clerical task, though some organizations require human sign-off for financial/legal correspondence like damage claims and billing disputes, creating light oversight friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-driven document generation has marginal inference costs measured in cents per document, plus modest integration overhead, versus a clerk's fully-loaded wage (30–50k USD annually). Cost per document is at least an order of magnitude lower than human labor. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Generating routine correspondence via AI costs a fraction of a cent per document compared to clerk wages for the same drafting work. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed products (document automation platforms, RPA tools, LLM-based document generation) reliably perform this task in enterprise settings today. Error rates on routine correspondence are low, though edge cases and novel scenarios may require human review, limiting this from a perfect 5. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed products (e.g., customer service AI drafting tools, CRM-integrated generative AI) already produce this type of correspondence in production at many companies, though accuracy on complex claims still requires human review. |
Compile data from records to prepare periodic reports.
72CI 72–72 · exposure 75 · augmentation 88 · importance 3.8/5 · click for rater detail
Compile data from records to prepare periodic reports.
72| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Adoption is uneven: large enterprises with digitized records and BI teams are automating this widely, while small firms and those with legacy paper records lag. Overall adoption is steady but not as rapid as in pure software or finance sectors; pilots are common, full displacement pockets exist, but the typical firm is still in early/mid stages. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Clerical and administrative functions are adopting automation steadily but unevenly, with many small offices still relying on manual compilation despite broader digitization trends in office/professional services. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can assist clerks by auto-populating templates, flagging anomalies in source data, and drafting report sections, allowing humans to focus on validation and interpretation. This productivity multiplication is substantial even when human oversight remains required. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI and automation tools substantially speed up data aggregation, formatting, and drafting of periodic reports, letting clerks focus on verification and exception handling. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can reliably extract data from structured and semi-structured records, aggregate it, and format it into reports with minimal human oversight. This task involves mostly rule-based data compilation rather than complex judgment, making it highly automatable with standard tools like LLMs and RPA systems that can achieve well over 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 4/5 | Compiling structured data into periodic reports is a well-defined, repetitive task that current AI/automation tools (RPA, LLMs with data connectors, spreadsheet/BI tools) can handle with significant time savings, provided data sources are reasonably structured. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or regulatory barriers exist for automating data compilation itself; no license is required. Modest barriers include internal data governance, record standardization requirements, and organizational inertia, but these are friction rather than hard blockers preventing substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal requirement mandates human performance of this task, though some organizational inertia and need for data-quality oversight create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven data extraction and report generation costs (inference + integration overhead) are typically a fraction of the loaded wage for a correspondence clerk ($40–50k annually). A single automated pipeline can handle hundreds of records per month at near-zero marginal cost, achieving better than 10x cost advantage at scale. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated data compilation and report generation tools run at a fraction of the cost of a clerk's time once set up, though initial integration and maintenance costs reduce the savings somewhat. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (document processing APIs, RPA platforms, BI tools with AI assist) demonstrably perform data extraction and report compilation in production environments. Error rates on well-defined schemas are low, though performance varies with data quality and record heterogeneity; this is a solved problem for many organizations but not universally deployed. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature products (Excel/Power BI automation, RPA platforms like UiPath, LLM-based reporting agents) are deployed in production for report compilation, though edge cases with messy or unstructured records still require human review. |
Gather records pertinent to specific problems, review them for completeness and accuracy, and attach records to correspondence as necessary.
70CI 67–72 · exposure 70 · augmentation 75 · importance 4.0/5 · click for rater detail
Gather records pertinent to specific problems, review them for completeness and accuracy, and attach records to correspondence as necessary.
70| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Document automation and enterprise search tools are widely piloted in finance and insurance, but production AI agents for full end-to-end record gathering and correlation remain less common; adoption is accelerating but not yet deeply embedded in all sectors. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Clerical and administrative functions are being augmented by AI within back-office operations at a moderate pace, with pilots and partial deployments more common than full-scale replacement. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can dramatically assist by auto-populating search results, flagging missing or inconsistent records, and pre-attaching documents, leaving humans to verify edge cases and make final judgment calls, substantially raising throughput. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly speed up locating relevant records and flagging discrepancies, allowing human clerks to focus on final verification and judgment calls, substantially boosting productivity while retaining oversight. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI systems can reliably search, retrieve, and validate records against checklists with high accuracy, and attach documents programmatically, achieving >50% time savings. The task is largely structured data retrieval and validation with minimal judgment required. |
| Task automatability | claude-sonnet-5 | 4/5 | Gathering, reviewing, and attaching records to correspondence is largely a structured retrieval and verification task that AI document-processing systems can perform end-to-end with substantial time savings, though edge cases in accuracy review may need human checks. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While some industries (banking, insurance) have compliance review requirements, most record gathering and attachment is routine administrative work with no hard legal barrier to automation; oversight friction exists but is not prohibitive. |
| Adoption barriers | claude-sonnet-5 | 2/5 | There are minimal licensing or legal barriers to automating this clerical task, though some organizational friction exists around data privacy, system integration, and ensuring accuracy before customer-facing correspondence is sent. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Automated record gathering and validation costs (API calls, inference, document processing) are a small fraction of a correspondence clerk's loaded wage, especially when handling high volumes. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated document retrieval and attachment via existing enterprise software is far cheaper per transaction than manual clerical labor once integrated, though initial setup costs offset some savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Document management systems with AI search and validation capabilities are deployed in production across financial, legal, and healthcare sectors; OCR and record-matching tools reliably perform these subtasks at scale, though some edge cases may require human review. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Document management and RPA/AI systems integrated with CRM or case-management software can retrieve and attach records today, but full accuracy verification across varied record types is not yet universally reliable in production without human oversight. |
Compile data pertinent to manufacture of special products for customers.
66CI 56–75 · exposure 62 · augmentation 75 · importance 3.5/5 · click for rater detail
Compile data pertinent to manufacture of special products for customers.
66| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Manufacturing and supply chain sectors have been actively adopting document automation, data extraction, and ERP integration tools for years. Cloud-based systems and RPA platforms handling similar compilation tasks show strong production deployment across enterprise organizations. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Clerical roles in manufacturing-adjacent administrative functions show slower AI adoption than pure information-sector roles, with pilots more common than full production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can substantially amplify clerk productivity by auto-extracting and pre-populating data fields, organizing information hierarchically, and flagging missing or inconsistent specifications, enabling the human clerk to focus on exception handling and quality validation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can significantly speed up gathering, organizing, and formatting customer and product data for clerks, who then verify and finalize the compiled information. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Compiling data about special product manufacturing is largely structured work: gathering specifications, requirements, materials, and process parameters from various sources and organizing them. AI systems can extract, aggregate, and format this information efficiently, achieving substantial time savings, though some domain-specific validation may require human input. |
| Task automatability | claude-sonnet-5 | 3/5 | Compiling structured data from orders, specs, and customer requests into a usable format is largely automatable with AI/RPA tools, but requires integration with company-specific systems and some judgment on ambiguous or incomplete data.: |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few barriers exist beyond organizational inertia and preference for human verification. No legal requirement mandates human participation, though companies may require a clerk's final sign-off for quality assurance, which is easily accommodated. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal requirement mandates a human for this compilation task, though internal accuracy/liability concerns for custom manufacturing specs create moderate organizational caution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Inference cost for automated data compilation is low per task. Integration into existing data systems is straightforward, and oversight is minimal once validated. This is substantially cheaper than human clerk labor, particularly for routine compilation of standard data fields. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Once set up, automated data extraction and compilation costs a fraction of a clerk's wage per unit of output, though integration and oversight add some cost relative to a fully manual comparison. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed document processing, data extraction, and database management systems already perform similar tasks reliably in production environments. Tools using OCR, NLP, and structured data pipelines can compile manufacturing specifications and customer requirements at scale with high accuracy. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like RPA plus LLM-based data extraction tools are deployed for document/data compilation in production, but reliability on messy, unstructured manufacturing specs varies and often needs human review. |
Prepare records for shipment by certified mail.
63CI 47–79 · exposure 58 · augmentation 50 · importance 3.2/5 · click for rater detail
Prepare records for shipment by certified mail.
63| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Logistics, finance, and professional services sectors—where correspondence clerks concentrate—have rapidly adopted RPA and document automation systems for mail and records management, with many organizations reporting production deployments. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Clerical/administrative mailroom functions are gradually adopting shipping software but are not fast-moving AI-adoption sectors; largely rules-based tools rather than AI-driven automation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by auto-populating metadata, flagging records needing special handling, and generating shipping documentation in real time, raising clerk productivity on oversight and exception-handling tasks, though full end-to-end automation is the primary value. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-adjacent tools (address validation, automated form-filling, tracking systems) can meaningfully speed up preparing certified mail paperwork, though the physical mailing step remains manual. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI systems can automate most of the mechanical steps: identifying records requiring certified mail, preparing documentation, generating labels, and arranging pickup—with minimal human intervention. Some domain knowledge of mailing protocols and record standards is required, but these are deterministic tasks that current automation handles well. |
| Task automatability | claude-sonnet-5 | 3/5 | Generating and printing certified mail records/labels and tracking documentation can be templated and automated with software, but physical preparation and handoff to postal carriers still require human/physical action.5 The digital portion (address verification, form generation) can be automated, but end-to-end includes physical steps. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few hard regulatory or licensing barriers exist for automating record preparation and mail coordination itself, though certified mail tracking requires carrier compliance (a mature, standard integration). Organizational friction around human oversight and customer expectations poses modest friction but does not legally block automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but certified mail requires physical presence at a postal facility and signature-based custody chain, creating minor procedural barriers beyond just software automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | End-to-end cost of AI-driven document preparation, label generation, and logistics coordination is typically an order of magnitude cheaper than the loaded wage of a correspondence clerk performing these mechanical tasks. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Software-based label/record generation is cheap, but physical mail prep still requires human labor for stuffing, sealing, and dropping off, keeping overall cost comparable to human-driven clerical work. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature document processing and logistics platforms (including RPA systems and document management software) demonstrably perform these functions in production at scale; integration with certified mail carriers is well-established. Minor exceptions arise in edge cases requiring human judgment about record eligibility. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Shipping software and mail-merge/label systems exist and are widely deployed, but full 'certified mail record preparation' as an integrated AI-driven task isn't a mature standalone AI product; it's mostly conventional automation, not AI-driven judgment. |
Ensure that money collected is properly recorded and secured.
60CI 25–95 · exposure 62 · augmentation 50 · importance 4.4/5 · click for rater detail
Ensure that money collected is properly recorded and secured.
60| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Financial services, accounting, retail, and most digitized sectors have already adopted automated payment processing and accounting systems at massive scale; displacement of manual recording roles is well underway and deeply embedded in production. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Clerical and cash-handling functions in many organizations remain manual or semi-digitized, with slower AI adoption compared to digital-native financial roles. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI assists human clerks by auto-categorizing transactions, flagging discrepancies, and offering reconciliation suggestions, raising productivity on exception-handling and verification tasks, though the core recording and securing functions are largely automated rather than augmented. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI-powered accounting and reconciliation tools can help clerks track and verify recorded amounts more efficiently, though the physical security component still requires human diligence. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Recording and securing financial transactions is highly automatable: automated accounting systems, payment processing software, and bank reconciliation tools already handle money tracking, ledger entry, and secure storage with minimal human intervention. Current AI and payment infrastructure can perform end-to-end cash/payment recording with well over 50% time savings. |
| Task automatability | claude-sonnet-5 | 2/5 | The reconciliation and recording aspects could be aided by software, but 'securing' money physically and verifying accountability involves physical custody and judgment that AI cannot perform end-to-end.: |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While some regulatory oversight (SOX, PCI-DSS) applies to financial controls, automation of recording and securing money does not legally require a licensed human sign-off in most jurisdictions; internal audit trails and system logs substitute for clerk oversight, with minimal organizational or liability friction. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Handling and securing money typically requires accountability, internal controls, and often bonded/trusted personnel, creating strong organizational and liability-driven barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated accounting and payment systems cost a small fraction of a human correspondence clerk's loaded wage per transaction processed; enterprise systems amortize to near-zero marginal cost per additional transaction, making AI vastly cheaper all-in. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While bookkeeping software is cheap, human oversight for physical security and accountability of cash remains necessary, keeping combined costs relatively high compared to a fully automated solution. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature, production-grade systems (accounting software like QuickBooks, SAP; payment processors; bank APIs; encryption and secure storage infrastructure) reliably perform financial recording and security at scale across millions of organizations worldwide. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Accounting software and cash management systems exist that record transactions, but no deployed AI product independently ensures secure handling and custody of collected funds in production. |
Present clear and concise explanations of governing rules and regulations.
46CI 25–67 · exposure 45 · augmentation 75 · importance 3.7/5 · click for rater detail
Present clear and concise explanations of governing rules and regulations.
46| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption is slow in most correspondence-clerk-heavy sectors (government, law firms, insurance) due to compliance sensitivity and risk aversion. Pilots are common but production deployment of fully autonomous AI explanation-generation is rare; human review remains the norm. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Clerical and administrative functions in insurance, government, and finance are adopting AI drafting tools steadily, but full deployment for regulatory correspondence remains uneven. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist clerks by drafting initial explanations or summarizing complex regulations, which the clerk then refines and verifies. This provides meaningful productivity lift on research and drafting, though the human remains responsible for accuracy and final presentation. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially speeds up drafting and clarifying regulatory explanations while clerks retain responsibility for accuracy and final sign-off. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can generate explanations of rules and regulations from source documents, but current systems struggle with nuance, jurisdiction-specific application, and ensuring legal accuracy—critical for this task. Meaningful automation would require significant guardrails and expert review, falling short of the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 4/5 | LLMs can draft clear, accurate explanations of rules and regulations from source documents with substantial time savings, especially when the governing text is provided or retrievable via RAG. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Legal and regulatory accuracy creates liability exposure; errors in explaining rules can expose organizations to compliance risk. Many organizations require human sign-off on regulatory communications, and correspondence involving legal interpretation often has implicit human-contact or authorization requirements. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement typically attaches to explaining rules, though organizations may require review for legal accuracy in regulated correspondence. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference costs are low, but the oversight, fact-checking, and legal review required to ensure accuracy add significant labor costs. The all-in cost approaches or exceeds that of a junior clerk producing the original explanation, offsetting computational savings. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating explanatory text via AI costs a small fraction of a clerk's time-equivalent, though some oversight cost is needed to ensure compliance accuracy. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While LLMs can draft rule explanations and some products offer legal summarization, deployed systems lack the reliability needed for correspondence clerks to rely on them end-to-end without substantial human verification. Error rates in regulatory interpretation remain material in production settings. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like AI writing assistants and customer-service chatbots do this today, but accuracy on nuanced or updated regulations still requires human review, limiting reliability at scale. |
Confer with company personnel regarding feasibility of complying with writers' requests.
28CI 25–30 · exposure 20 · augmentation 50 · importance 3.3/5 · click for rater detail
Confer with company personnel regarding feasibility of complying with writers' requests.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Correspondence and clerical work remains heavily human-driven in most organizations; adoption of AI agents for inter-departmental conferencing is still rare and experimental, with most firms preferring human intermediaries for stakeholder communication. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Correspondence clerk roles are in a shrinking, low-digitization niche with limited enterprise AI agent deployment for internal cross-functional consultation tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by summarizing writers' requests, drafting preliminary feasibility assessments, and surfacing relevant constraints, helping the clerk prepare for or streamline the conference—but the human typically must own the dialogue itself. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help summarize requests, draft talking points, or provide feasibility research to inform the human-led conferral, offering moderate productivity support. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | This task requires nuanced judgment about organizational constraints, feasibility assessment, and interpersonal communication with multiple stakeholders. While AI could draft feasibility analyses or compile relevant information, the core negotiation and conferencing element—understanding context-specific business logic and building consensus—remains difficult for current systems to automate end-to-end with equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires interactive human deliberation and internal negotiation across departments, which AI cannot fully substitute for since it depends on organizational authority and judgment calls that aren't yet delegated to AI systems.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Organizational norms and internal governance often require a human agent to formally confer on request feasibility, and clients may expect to dialogue with named personnel. However, no strict legal barrier prevents AI assistance or partial automation of analysis stages. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but organizational friction is high since internal decision-making authority and interpersonal negotiation are not easily replaced by automated systems. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | While AI inference is cheap, the task requires integration with organizational workflows, human verification of feasibility judgments, and potential re-engagement with stakeholders, making total delivered-value cost comparable to or higher than a clerk's hourly rate for this specific function. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI could support drafting or summarizing but the core conferring/negotiation activity still requires human staff time, so cost savings are limited to peripheral support functions. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably conducts independent negotiations or conferences with company personnel on feasibility questions. LLMs can draft responses or analyze requests, but deployed systems lack the real-time interaction, contextual authority, and accountability needed for genuine conferencing at scale. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously confers with internal personnel to assess feasibility of external requests; this remains a human coordination task not addressed by current AI products. |
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.