Customer Service Representatives
43-4051.00Median wage $44,770/yr2,595,750 employed (US)Rank #28 of 923 scored · top 3% by substitution
Interact with customers to provide basic or scripted information in response to routine inquiries about products and services. May handle and resolve general complaints. Excludes individuals whose duties are primarily installation, sales, repair, and technical support.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution64
Exposure61
Augmentation78
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
13 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
38%
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 3.4/5 → substitution pressure 61/100
Technical feasibility todayw 20%61
panel mean rating 3.4/5 → substitution pressure 61/100
Cost vs. human wagew 15%74
panel mean rating 4.0/5 → substitution pressure 74/100
Adoption barriersw 20%inverted — strong barriers lower the score67
panel mean rating 2.3/5 (barrier strength) → substitution pressure 67/100
Sector adoption velocityw 10%66
panel mean rating 3.6/5 → substitution pressure 66/100
Task breakdown (13 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.
Keep records of customer interactions or transactions, recording details of inquiries, complaints, or comments, as well as actions taken.
89CI 79–100 · exposure 92 · augmentation 100 · importance 4.5/5 · click for rater detail
Keep records of customer interactions or transactions, recording details of inquiries, complaints, or comments, as well as actions taken.
89| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Customer service operations are highly digitized and information-intensive; major CRM and contact-center platforms have already integrated AI-driven call logging, transcription, and summarization widely. Adoption is deep and accelerating across enterprise and mid-market segments. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service and CRM software sectors have rapidly integrated AI-based transcription and summarization tools, though full deployment varies by company size. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments human representatives by automatically generating drafts of interaction logs, flagging key issues, and populating CRM fields in real-time, freeing representatives to focus on empathy and problem-solving rather than manual documentation. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially speeds up and improves consistency of interaction logging, letting reps focus on resolving issues rather than documentation. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can reliably capture, summarize, and log customer interactions from voice, chat, or email with high accuracy and can structure data into standard record formats. The task of recording interaction details and actions taken is straightforward documentation that can achieve >50% time savings through automatic transcription, summarization, and database entry, though some human review may be needed for complex edge cases. |
| Task automatability | claude-sonnet-5 | 5/5 | AI can automatically log, summarize, and categorize customer interactions from chat, email, or call transcripts with minimal human input, meeting the time-saving threshold easily. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Some regulatory frameworks (e.g., HIPAA, GDPR, financial compliance) require documented oversight of customer records and may mandate human review or audit trails. Customer expectations and organizational policy may still require human touch-points for sensitive complaints, creating friction but not hard legal barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or regulatory requirement mandates human record-keeping for customer service logs; it's a low-liability clerical task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Inference cost for transcription and summarization of a single interaction is typically $0.01–0.10, with integration amortized across high-volume deployments, versus human wages of $15–25/hour to manually document interactions. AI is easily an order of magnitude cheaper per task. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated note-taking and CRM entry via AI costs a small fraction of a cent per interaction versus manual data entry time by a paid representative. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature products (e.g., Salesforce, Zendesk, Amazon Connect, Gong) demonstrably perform this task in production at scale across thousands of organizations, automatically logging calls, chats, emails, and generating summaries with minimal human intervention. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | CRM systems (Salesforce, Zendesk, etc.) already deploy AI-driven auto-logging, transcription, and summarization features widely in production at scale. |
Determine charges for services requested, collect deposits or payments, or arrange for billing.
79CI 79–79 · exposure 75 · augmentation 75 · importance 4.2/5 · click for rater detail
Determine charges for services requested, collect deposits or payments, or arrange for billing.
79| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Payment processing and billing automation are deeply embedded in finance, e-commerce, SaaS, and telecommunications sectors with high digitization; adoption is mature and widespread rather than nascent. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service and billing functions in telecom, utilities, and finance have seen fast, deep automation adoption via self-service portals and automated billing systems. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assists customer service representatives by instantly calculating charges, flagging payment issues, suggesting billing arrangements, and auto-populating invoices, significantly raising agent productivity and accuracy on routine transactions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools assist reps by auto-calculating charges, flagging discrepancies, and streamlining payment collection, significantly speeding up the human-in-the-loop process for complex or exception cases. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can reliably determine service charges based on predefined rules, collect payments through automated billing systems, and arrange billing schedules with minimal human intervention, though complex disputes or custom pricing may require oversight. This meets the ≥50% time-saving bar for the majority of routine transactions. |
| Task automatability | claude-sonnet-5 | 4/5 | Determining charges from rate tables and processing payments/billing is highly structured and already handled by chatbots, IVR systems, and billing platforms with minimal human input for most standard cases. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While PCI compliance and fraud prevention add some regulatory friction, no legal requirement mandates a human must personally collect payment or determine charges; most barriers are organizational preference and oversight practices rather than hard legal restrictions. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Some friction exists around payment security/compliance (PCI-DSS) and dispute handling, but no licensing requirement mandates a human to determine charges or collect payment. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated payment processing, billing systems, and rule-based charge determination cost a small fraction of a full-time customer service representative's loaded wage, especially for high-volume transactions. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated billing and payment systems cost a fraction of a cent per transaction compared to a human rep's loaded wage for the same task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed payment-processing platforms, automated billing systems, and chatbots handle charge calculation and payment collection at scale in production; however, edge cases involving negotiation or special arrangements still see material error rates requiring human review. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Production systems (billing software, automated payment portals, AI-driven customer service chatbots) reliably calculate charges and process payments today across telecom, utilities, and retail sectors. |
Complete contract forms, prepare change of address records, or issue service discontinuance orders, using computers.
79CI 70–87 · exposure 83 · augmentation 75 · importance 4.1/5 · click for rater detail
Complete contract forms, prepare change of address records, or issue service discontinuance orders, using computers.
79| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Utilities, telecom, and financial-services sectors are actively deploying RPA and intelligent automation for contract and order processing; adoption is measured and growing in these digitized, high-transaction-volume industries. Customer service centers handling routine administrative work show consistent adoption of form-automation tools. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service in telecom, utilities, and finance sectors has rapidly adopted self-service portals and automated account management tools, with high digitization and API integration already common. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assistants can draft forms, pre-fill fields from customer records, and flag potential errors or missing information, significantly accelerating human agents' review and approval workflows. This augmentation allows representatives to handle more cases or focus on exceptions, raising productivity without full replacement. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-assisted systems pre-fill forms, auto-populate change requests, and flag anomalies, letting remaining human reps process higher volumes and handle exceptions more efficiently. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | This task is highly structured and rule-based: extracting customer information, populating form fields, and generating change-of-address or discontinuance orders can be performed end-to-end by current AI systems with document processing and form-filling capabilities, achieving >50% time savings. However, edge cases requiring judgment (disputed accounts, unusual service situations) may still need human review. |
| Task automatability | claude-sonnet-5 | 5/5 | Form completion, address changes, and service discontinuance are structured, rules-based data entry tasks that current AI/RPA systems handle end-to-end with significant time savings and equal or better accuracy. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Moderate barriers exist: service discontinuance orders may require human authorization or customer contact verification for regulatory/compliance reasons; some sectors (utilities) have customer-protection rules mandating human sign-off. However, no strict legal requirement prevents AI from completing the form and documentation itself, so adoption is feasible with oversight. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Some contracts may require identity verification or specific consumer protection disclosures, but most of this work carries minimal licensing or liability requirements. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven form completion and order processing cost a fraction of human labor once deployed (per-transaction cost is typically <$0.10 for inference + overhead vs. $2–5 human labor equivalent), representing a 10–50× cost advantage. Integration and oversight costs are modest for high-volume, repeatable tasks. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated form processing and account updates cost a small fraction of a cent per transaction versus a loaded human wage handling the same volume of routine requests. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed RPA and intelligent document processing products (e.g., UiPath, Blue Prism, commercial OCR + form-filling APIs) reliably automate form completion and basic order generation in production environments across utilities and service sectors. Minor limitations remain in handling ambiguous input or exception cases, but the core task is demonstrably automated at scale. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Many companies already deploy chatbots, IVR systems, and RPA bots in production to process address changes and service cancellations, though some edge cases still route to humans. |
Contact customers to respond to inquiries or to notify them of claim investigation results or any planned adjustments.
74CI 61–87 · exposure 70 · augmentation 75 · importance 4.2/5 · click for rater detail
Contact customers to respond to inquiries or to notify them of claim investigation results or any planned adjustments.
74| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Insurance and customer service sectors have rapidly deployed AI chatbots and automated notification systems over the past 3–5 years; major carriers now use these extensively in production, though smaller firms lag behind. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service and insurance/financial sectors have rapidly adopted AI-driven chat and voice agents for notifications and inquiries, with production deployment now common. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assists human representatives by drafting responses, summarizing claim investigations, surfacing relevant policies, and triaging inquiries, which substantially raises productivity and reduces call handling time while the human verifies tone and complex cases. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools draft responses, summarize claim files, and suggest next steps, significantly speeding up representatives' work while they retain final judgment on sensitive claim communications. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Modern AI systems can handle end-to-end customer inquiry response and notification delivery via email, SMS, or chatbots with well-defined information (claim status, adjustment details) at well over 50% time savings compared to human agents, especially for routine notifications and standard inquiry categories. |
| Task automatability | claude-sonnet-5 | 3/5 | Routine outbound notifications and simple inquiry responses can be handled by AI chatbots or voice agents, but claim-specific nuanced explanations and empathetic handling of disputes still require human judgment for a large share of cases. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While many organizations prefer human contact for sensitive claim decisions and some customers demand human interaction, there are few legal licensing requirements or strict regulatory mandates forcing human sign-off on routine notifications or standard inquiry responses; mostly organizational and customer-experience friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No strict licensing requirement for this communication task, but liability and customer relationship sensitivity around claim decisions create moderate friction favoring human involvement in contentious cases. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-driven customer contact (chatbots, IVR systems, automated email) costs orders of magnitude less than loaded human agent wages per interaction, particularly for high-volume notification and FAQ-type inquiries. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated notification and basic inquiry-response systems cost a small fraction of a human agent's loaded wage per interaction, though oversight and escalation paths add some cost back. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed chatbots and AI customer service platforms (e.g., from major vendors) reliably handle structured inquiry responses and automated notifications in production at scale, though performance degrades on complex edge cases requiring nuanced investigation summaries. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed contact-center AI (chatbots, voicebots, automated notification systems) reliably handle status updates and FAQs, but complex claim outcome communications still commonly escalate to humans due to error and complaint risk. |
Compare disputed merchandise with original requisitions and information from invoices and prepare invoices for returned goods.
72CI 67–76 · exposure 70 · augmentation 75 · importance 3.6/5 · click for rater detail
Compare disputed merchandise with original requisitions and information from invoices and prepare invoices for returned goods.
72| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Retail, e-commerce, and logistics sectors are piloting and deploying return-processing automation, but many organizations still rely on manual review for fraud prevention and dispute sensitivity. Adoption is accelerating but not yet dominant in production workflows. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Customer service and back-office finance functions are adopting automation and RPA at a moderate pace, with mature pilots in invoice processing but not universal deployment across all customer service teams. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can serve as a powerful assistant, pre-populating return invoices, flagging discrepancies, and highlighting suspicious patterns, allowing a human agent to focus on edge cases and customer communication rather than data entry and comparison. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can quickly flag discrepancies, auto-populate return invoices, and surface relevant historical data, significantly speeding up the rep's verification and documentation process while they retain final judgment on disputes. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can extract and compare product details from invoices, requisitions, and returned merchandise images/metadata with high accuracy, and generate return invoices automatically. While some edge cases (condition disputes, partial returns) may require human review, the bulk processing easily achieves >50% time savings at quality parity. |
| Task automatability | claude-sonnet-5 | 4/5 | This is largely a document comparison and data-processing task—matching invoice line items, requisitions, and returned goods information—which AI systems with document parsing and structured data extraction handle well, though physical inspection of merchandise still requires human input. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing requirement or legal mandate for human sign-off; the main friction is customer preference for human contact and organizational hesitation to fully automate customer-facing interactions. However, the task itself—data matching and form generation—has minimal regulatory or liability barriers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement typically governs this task, but organizational policies often require human sign-off on discrepancies to avoid financial errors or customer disputes, creating moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Document scanning, comparison, and invoice generation via cloud APIs cost pennies per transaction, while a human agent's loaded wage for the same task is $20–50+. The cost ratio is easily 10:1 or better in high-volume environments. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Once integrated, automated document matching and invoice generation systems process transactions far cheaper than manual comparison and paperwork, though initial integration with ERP/inventory systems adds cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature OCR and document processing systems (AWS Textract, Azure Document Intelligence) reliably extract invoice and requisition data at scale; invoice generation is commodity software. Deployed AI systems in retail/logistics handle similar document matching and return-invoice creation in production today. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products exist for automated invoice matching and reconciliation (RPA, OCR-based systems) deployed in accounts payable/returns workflows, but end-to-end handling of disputed merchandise cases still requires human judgment for exceptions and edge cases. |
Check to ensure that appropriate changes were made to resolve customers' problems.
67CI 55–79 · exposure 62 · augmentation 75 · importance 4.4/5 · click for rater detail
Check to ensure that appropriate changes were made to resolve customers' problems.
67| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Customer service is a digitized, high-volume sector with strong incentives for cost reduction; major platforms already embed AI verification tools and many mid-to-large enterprises are deploying them. Adoption is already visible in production across contact centers and e-commerce. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Customer service is a fast-adopting sector for AI tools generally, but this specific sub-task (verification of resolution) is less commonly fully automated compared to initial response drafting. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-assisted verification significantly boosts human agent productivity by surfacing problematic or incomplete tickets in real time, allowing agents to focus on exceptions and follow-up rather than manual spot-checking, while keeping the human in final approval. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can flag inconsistencies, pull relevant records, and suggest whether resolution criteria were met, significantly speeding up the rep's verification process while they retain final judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can systematically verify that problem-resolution steps were completed (cross-referencing tickets, logs, and system records) and flag discrepancies, achieving significant time savings over manual spot-checking. However, complex or ambiguous cases requiring judgment about whether a solution truly satisfies the customer's intent may require human oversight, preventing a full end-to-end rating of 5. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can cross-reference ticket status, account records, and resolution logs to verify changes were applied, but ambiguous or judgment-heavy verification (e.g., confirming customer satisfaction) still needs human review. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few hard regulatory or licensing barriers exist for automated verification of ticket resolution; most barriers are organizational (preference for human QA, customer trust concerns) rather than legal. Data privacy and audit-trail requirements create minor friction but do not prevent adoption. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but customer trust, error-cost asymmetry (wrongly closing unresolved issues), and org policies create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Once integrated into existing CRM infrastructure, automated verification incurs minimal incremental cost (API calls, compute) compared to the loaded wage of a human agent performing the same spot-check or verification work, achieving clear order-of-magnitude savings. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated verification against structured records is cheap, but integration with diverse backend systems and edge-case escalation still requires human oversight, keeping costs roughly comparable in many setups. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature ticket-management and CRM systems with built-in AI modules (ServiceNow, Zendesk, Salesforce) now include automated verification workflows that check task completion against predefined criteria and flagging inconsistencies in production. Reliability is high for structured, rule-based verification but can struggle with novel or context-dependent resolution scenarios. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | CRM and support platforms with AI-driven QA and ticket-verification features exist and are deployed, but they still have notable error rates and often require human confirmation for closure. |
Confer with customers by telephone or in person to provide information about products or services, take or enter orders, cancel accounts, or obtain details of complaints.
66CI 56–75 · exposure 62 · augmentation 88 · importance 4.7/5 · click for rater detail
Confer with customers by telephone or in person to provide information about products or services, take or enter orders, cancel accounts, or obtain details of complaints.
66| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large contact centers and e-commerce organizations are rapidly deploying chatbots, IVR upgrades, and AI-assisted routing; Fortune 500 adoption is widespread. However, smaller firms and specialized service sectors remain largely human-staffed, tempering the overall velocity. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service is among the fastest and most deeply AI-adopting functions, with chatbots and voice agents in widespread production across telecom, retail, and finance. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists customer service representatives by auto-suggesting responses, surfacing relevant product information, draft-writing emails, and flagging high-risk interactions. These tools measurably boost agent productivity and first-contact resolution while keeping humans in control of sensitive decisions. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools like call summarization, real-time suggested responses, and CRM auto-entry substantially boost representative productivity even when full automation doesn't occur. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | Current AI can handle routine inquiries (product information, basic order entry, simple complaints) but struggles with complex troubleshooting, emotional de-escalation, and nuanced account decisions. A hybrid approach—AI handling initial triage and data entry, humans managing exceptions—could achieve partial time savings, but end-to-end automation for the full scope falls short of the 50% threshold at equal quality. |
| Task automatability | claude-sonnet-5 | 4/5 | Voice and chat AI agents already handle order-taking, cancellations, and basic complaint intake for many routine queries, meeting time-saving thresholds for a large share of interactions, though complex complaints and edge cases still need humans. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | No strict licensing requirement, but organizational friction is real: customer preference for human interaction, contractual service-level expectations, liability concerns around order accuracy, and internal change management delays adoption. Regulatory requirements vary by jurisdiction (e.g., financial services) but are not uniformly prohibitive. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement typically applies, but customer preference for human contact on complaints/cancellations and retention risk creates moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven contact-center automation (inference + integration + oversight) is substantially cheaper than full-time customer service staff, particularly for high-volume, repetitive interactions. Cost per interaction can be an order of magnitude lower for routine tasks, though human oversight costs add up. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI-driven voice/chat handling costs a fraction of a per-call human agent, especially at scale, though integration and oversight still add nontrivial cost for complex complaint handling. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Chatbots and IVR systems are deployed in production handling straightforward queries and order placement, but error rates remain material on complex issues and natural-language variance. Real-world systems still require human escalation for ~20–40% of interactions, limiting reliable end-to-end performance. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Production IVR/chat systems (e.g., large telecom, retail, and bank deployments) reliably handle standard order and account requests today, though escalation to humans remains common for nuanced complaints. |
Refer unresolved customer grievances to designated departments for further investigation.
64CI 52–75 · exposure 55 · augmentation 75 · importance 4.1/5 · click for rater detail
Refer unresolved customer grievances to designated departments for further investigation.
64| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Customer service organizations, particularly large information-sector companies, have rapidly adopted automated ticketing and AI-driven routing systems over the past 3–5 years, with many now using agents or rule engines to triage and escalate complaints at scale. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service is a fast-adopting sector for AI tooling, with widespread deployment of automated triage and routing in call centers and support platforms. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists customer service representatives by suggesting appropriate departments, summarizing grievances, and highlighting priority escalations, allowing reps to make faster, more informed routing decisions while maintaining control over the final referral. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially speeds up identifying the right department and drafting escalation notes, letting reps focus on customer communication rather than manual routing. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can classify customer issues and route some to departments, the task requires judgment to identify which grievances are genuinely unresolved and warrant escalation—a determination that often depends on context and nuance that current systems struggle with reliably. Most implementations still require human review before escalation. |
| Task automatability | claude-sonnet-5 | 4/5 | Routing unresolved grievances to correct departments based on categorization is a well-structured task that current AI ticketing/triage systems handle well, though edge cases still require judgment. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | There are minimal legal or regulatory barriers to automating referral routing; organizations face mainly internal friction around customer satisfaction preference for human contact and concerns about misrouting complaints that could escalate issues. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Little regulatory or licensing barrier exists for internal routing decisions, though some organizations require human review for sensitive complaints or legal/compliance escalations. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Automating the referral and routing of grievances via rule-based systems or basic AI is significantly cheaper than paying a representative's fully loaded wage to make these routing decisions, especially at scale. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated routing via AI classifiers costs a small fraction of a human rep's time per ticket, though some oversight and integration costs remain. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Ticketing systems and basic chatbots can perform automated routing to departments, but many organizations still rely on human representatives to make the escalation decision, and AI systems often misroute complex or edge-case grievances. Reliable end-to-end performance is limited and varies by complaint complexity. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | CRM and helpdesk platforms (Zendesk, Salesforce Service Cloud, Freshdesk) already deploy AI-based ticket routing and escalation in production at scale, though occasional misclassification occurs. |
Solicit sales of new or additional services or products.
60CI 59–61 · exposure 50 · augmentation 75 · importance 3.9/5 · click for rater detail
Solicit sales of new or additional services or products.
60| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large contact centers, e-commerce, and financial services are rapidly deploying chatbots, predictive dialers, and AI recommendation engines for upsell and cross-sell. Adoption is visible in earnings calls and industry reports, with significant production use in call centers and online platforms. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Contact centers across telecom, retail, and finance are piloting and partially deploying AI-assisted upselling, but most production use is augmentative rather than fully autonomous solicitation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly boosts rep productivity by screening leads, suggesting relevant products in real time, and drafting pitches; human reps then close and handle objections. Sales teams widely adopt these tools to increase conversion and call efficiency while reps remain central to closure. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools effectively surface next-best-offer recommendations and prompts in real time, meaningfully boosting a rep's ability to identify and pitch relevant products during interactions. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can identify upsell/cross-sell opportunities and generate personalized product recommendations with reasonable accuracy, but requires human intervention to close the sale and handle objections effectively. Current chatbots achieve partial automation of the pitch phase but lack the nuanced persuasion and relationship-building that drive conversion at scale. |
| Task automatability | claude-sonnet-5 | 3/5 | AI chatbots and voice agents can pitch upsells and cross-sells using scripts and customer data, but nuanced persuasion, objection handling, and relationship-based selling still benefit from human judgment, limiting full end-to-end substitution. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Telemarketing and email solicitation face regulatory barriers (TCPA, CAN-SPAM, state attorney general rules), but these constrain the activity itself, not the automation. No license requirement for the task itself, and liability for bad recommendations is shared broadly; companies already substitute partial automation with modest friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement typically applies, but some regulated industries (insurance, financial services) require disclosures or human sign-off on sales, and customer trust/preference can favor human interaction for upsells. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven outbound solicitation (automated calls, email campaigns, chatbot upsells) is substantially cheaper per touch than human reps once integrated, though setup and compliance oversight add costs. At scale, AI solicitation costs a fraction of loaded human wages for the same volume. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI-driven upsell prompts and chatbot-based solicitation cost a fraction of a human agent's time per interaction, though oversight and CRM integration add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed systems (chatbots, recommendation engines, outbound dialing assistants) can solicit new services in production, but with elevated error rates—false product recommendations, tone mismatches, regulatory missteps—compared to trained humans. Success is highly variable by use case and sector. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Conversational AI and recommendation-driven upsell prompts are deployed in call centers and chat support today, but conversion rates and reliability for actual sales solicitation still lag skilled human reps, especially for complex or high-value offers. |
Review insurance policy terms to determine whether a particular loss is covered by insurance.
51CI 49–54 · exposure 50 · augmentation 75 · importance 4.0/5 · click for rater detail
Review insurance policy terms to determine whether a particular loss is covered by insurance.
51| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Insurance and financial services sectors show active pilots and adoption of document-review automation, but full displacement remains limited by regulatory constraints and liability concerns. Real production use tends to focus on low-stakes, routine determinations rather than complex or contested claims. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Insurance is a data-rich, moderately digitized sector adopting AI in claims and underwriting, but many carriers still pilot rather than fully deploy AI for binding coverage decisions. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems can rapidly surface relevant policy sections, highlight applicable clauses, and flag potential coverage matches, substantially accelerating a representative's review process. This assistive role—keeping the human in the loop—is already demonstrable and widely deployed in customer service and claims management software. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can quickly surface relevant policy clauses, precedent determinations, and flag potential coverage issues, substantially speeding up a rep's review process while the rep confirms the final call. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can extract and match policy language against loss descriptions with high accuracy, automating substantial portions of the coverage determination workflow. However, edge cases involving ambiguous policy wording, competing coverage clauses, or borderline fact patterns still require human judgment, preventing full end-to-end automation at the ≥50% time-saving threshold for all scenarios. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can parse policy language and flag likely coverage determinations quickly, but complex claims with ambiguous facts or edge cases still require human judgment and liability oversight, limiting full end-to-end automation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Insurance regulation typically requires human judgment and sign-off on coverage determinations, especially for disputes or high-value claims; state insurance commissioners and policy language statutes often mandate or imply human accountability. Liability and error-cost asymmetry are high, since incorrect coverage denials create direct financial and reputational harm. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement typically applies to CSRs making these determinations, but insurers face regulatory and liability exposure for wrongful denials, creating institutional caution around full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference on document review is inexpensive relative to the loaded wage of a customer service representative or claims adjuster. API costs for LLMs and OCR-based systems are substantially lower than human labor, though integration and oversight infrastructure add moderate overhead. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated document review and NLP-based policy matching is far cheaper per query than a human rep's time, though human review costs remain for complex or contested cases. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed NLP and document-extraction systems can reliably parse policy documents and identify relevant coverage clauses in production environments. Mature products exist in the insurance tech space, but they typically require human review of complex or disputed claims and have material error rates on genuinely ambiguous policies. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Insurers deploy AI-assisted claims triage and policy-review tools in production, but these typically flag or draft determinations rather than issue final binding coverage decisions independently. |
Resolve customers' service or billing complaints by performing activities such as exchanging merchandise, refunding money, or adjusting bills.
47CI 32–61 · exposure 38 · augmentation 75 · importance 4.2/5 · click for rater detail
Resolve customers' service or billing complaints by performing activities such as exchanging merchandise, refunding money, or adjusting bills.
47| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Many organizations have deployed basic chatbots for complaint triage and simple resolutions, yet meaningful automation remains concentrated in high-volume, low-complexity scenarios (password resets, order status). Broader adoption of AI for judgment-heavy complaint resolution is still in pilot and early-production phases. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Customer service is one of the fastest-adopting sectors for AI agents, with major deployments in telecom, retail, and finance for billing and complaint resolution. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assistants meaningfully augment human reps by drafting responses, suggesting resolution options, summarizing customer history, and flagging policy exceptions. These tools visibly improve productivity and consistency while keeping human judgment and accountability in the loop. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly speed up complaint triage, suggest resolutions, and draft responses, letting human reps handle higher volumes and more complex cases. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can handle some complaint categorization and routing, the task requires judgment about refunds, adjustments, and exchanges that depend on context-specific policies, customer history, and discretionary decisions. End-to-end automation with 50% time savings and equal quality remains difficult without significant human oversight. |
| Task automatability | claude-sonnet-5 | 3/5 | AI chatbots can handle straightforward complaints, refunds, and billing adjustments via integrated systems, but complex or emotionally charged disputes still require human judgment and authority to override policy. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Financial and legal exposure is meaningful—incorrect refunds or billing adjustments carry direct cost and compliance risk—creating organizational caution. However, no strict licensing requirement exists; companies can delegate authority to AI systems if they accept liability and implement oversight. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Few legal requirements mandate human involvement, though some jurisdictions have consumer protection rules on billing disputes and companies retain human oversight for high-value refunds or fraud risk. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Basic chatbot inference is cheap, but the integration cost, error correction, human review for non-trivial cases, and compliance oversight are substantial. For the majority of complaints requiring judgment, the all-in cost approaches or exceeds a loaded customer service rep wage. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated resolution via chatbots and self-service portals is dramatically cheaper per interaction than human agent time, though oversight and escalation paths add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Chatbots and automated systems exist for simple, rule-based complaints (e.g., straightforward refund requests), but deployed products struggle with nuanced cases, exceptions, and high-stakes adjustments. Most real-world production systems handle only the most routine inquiries, requiring human escalation for complex complaints. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Many companies deploy AI-driven support (e.g., automated refund approvals, billing bots) but these systems often escalate ambiguous or high-value cases to humans, limiting full reliability. |
Obtain and examine all relevant information to assess validity of complaints and to determine possible causes, such as extreme weather conditions that could increase utility bills.
46CI 37–55 · exposure 42 · augmentation 75 · importance 3.6/5 · click for rater detail
Obtain and examine all relevant information to assess validity of complaints and to determine possible causes, such as extreme weather conditions that could increase utility bills.
46| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Customer service centers are actively piloting AI chatbots and automated intake systems, but actual displacement of assessment work is still limited; most deployments augment rather than replace. Adoption is accelerating but production-level automation remains piecemeal across the sector. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Utility and customer service sectors are adopting AI for complaint triage and data lookups at a middling pace, with pilots more common than full production deployment. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can significantly assist representatives by auto-populating complaint summaries, retrieving relevant historical data, flagging weather or service-disruption context, and suggesting common causes, substantially raising a human agent's speed and decision quality while they retain final judgment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can quickly surface relevant account history, usage patterns, and weather data to help reps assess complaints faster, significantly aiding investigation even if final judgment remains human. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can retrieve and parse structured complaint data and cross-reference weather records automatically, the task requires contextual judgment about complaint validity and causal reasoning that often involves nuance, customer rapport, and domain-specific knowledge. Current systems can accelerate data gathering but cannot reliably complete the full assessment end-to-end at quality parity. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can gather account data, weather records, and usage history to assess complaint validity, but nuanced judgment about causation and edge cases still requires human oversight for full task completion. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Customer service roles have moderate friction: many organizations retain human complaint handlers for liability and trust reasons, and some regulatory or contractual obligations require human sign-off on complaint resolution. However, no hard legal barrier prevents AI-driven initial assessment and triage. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but customer trust, potential billing disputes, and liability for wrong determinations create moderate organizational friction favoring human review. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference for data retrieval and basic triage is cheap, but integration into legacy CRM systems and the need for human oversight of assessments means per-task cost approaches or exceeds the loaded wage of a junior representative handling the same task. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted data retrieval and analysis is cheaper than manual research, but integration with billing/weather systems and human review keeps costs closer to comparable rather than order-of-magnitude cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed chatbots and knowledge-management systems can extract complaint details and retrieve reference information (weather data, billing histories), but they struggle with novel or ambiguous cases and typically require human handoff for validation. Production systems exist but with material error rates and narrow scope limits. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed customer service AI systems can pull data and flag likely causes, but reliable end-to-end validation of complaints with contextual reasoning is still narrow and error-prone in production. |
Recommend improvements in products, packaging, shipping, service, or billing methods and procedures to prevent future problems.
31CI 25–38 · exposure 25 · augmentation 63 · importance 3.2/5 · click for rater detail
Recommend improvements in products, packaging, shipping, service, or billing methods and procedures to prevent future problems.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Customer service teams have adopted AI for routine routing and FAQ responses, but AI-driven improvement recommendation systems remain niche pilots. Most organizations still rely on human representatives, supervisors, and feedback loops to surface and validate improvements. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Customer service functions are adopting AI analytics tools moderately, with pilots for sentiment/trend analysis but limited use for formal process improvement recommendations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by summarizing frequent complaint themes, clustering issues by type, or surfacing statistical patterns that a human representative can then synthesize into prioritized recommendations. This augmentation is real but incremental rather than transformative. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can effectively summarize complaint trends, flag recurring issues, and draft improvement suggestions, significantly aiding a human who finalizes and contextualizes recommendations. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can identify patterns in complaint data and suggest generic improvements, the task requires contextual judgment about business trade-offs, customer needs, and feasibility—factors that vary significantly across organizations. Current AI lacks the domain expertise and accountability to generate reliable, actionable improvements that reduce problems at the level a human representative would. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires synthesizing patterns across customer interactions and business context to generate actionable, prioritized recommendations, which current AI can partially support but not fully replace end-to-end."}, |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Organizational friction is substantial: improvements typically require sign-off from product, operations, or executive teams; legal/compliance review of procedural changes; and strong preference for recommendations grounded in direct customer interaction rather than automated systems. Liability concerns for flawed recommendations also create hesitation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational buy-in and accountability for process changes create moderate friction against pure AI-driven recommendations. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted analysis tools are affordable, but the total cost including integration, validation, expert review of recommendations, and fallback to human judgment remains comparable to or exceeds the cost of a thoughtful human representative generating the same quality insight. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Generating credible operational recommendations still requires significant human review and business judgment, so AI-only cost savings are limited relative to human analyst time. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs this task end-to-end. LLMs can generate template suggestions from complaint summaries, but quality is inconsistent, often generic, and requires substantial human validation and refinement before implementation. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some analytics/AI tools surface trends in complaints, but deployed products rarely generate reliable, business-specific improvement recommendations without heavy human curation. |
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.