Credit Authorizers, Checkers, and Clerks
43-4041.00Median wage $50,080/yr12,030 employed (US)Rank #19 of 923 scored · top 2% by substitution
Authorize credit charges against customers' accounts. Investigate history and credit standing of individuals or business establishments applying for credit. May interview applicants to obtain personal and financial data, determine credit worthiness, process applications, and notify customers of acceptance or rejection of credit.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution68
Exposure68
Augmentation66
Substitution — the headline: capability discounted by cost, barriers and adoption.
Exposure — technical capability alone, regardless of whether anyone deploys it.
Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.
Tasks on the substitution scale
16 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
50%
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%71
panel mean rating 3.8/5 → substitution pressure 71/100
Technical feasibility todayw 20%69
panel mean rating 3.8/5 → substitution pressure 69/100
Cost vs. human wagew 15%81
panel mean rating 4.3/5 → substitution pressure 81/100
Adoption barriersw 20%inverted — strong barriers lower the score58
panel mean rating 2.7/5 (barrier strength) → substitution pressure 58/100
Sector adoption velocityw 10%69
panel mean rating 3.8/5 → substitution pressure 69/100
Task breakdown (16 tasks)
Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.
Mail charge statements to customers.
97CI 95–100 · exposure 100 · augmentation 25 · importance 3.8/5 · click for rater detail
Mail charge statements to customers.
97| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Financial services and utilities have been rapidly automating statement generation and mailing for over a decade; digital-first, automated statement delivery is now standard practice across the sector. |
| Sector adoption velocity | claude-sonnet-5 | 5/5 | Billing automation is essentially universal in finance, retail, and utilities sectors, with automated statement generation the long-established default rather than an emerging trend. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI offers minimal assistive value once the task is automated; any augmentation would be in initial design or exception handling, not in helping humans execute the core mailing workflow. |
| Augmentation potential | claude-sonnet-5 | 2/5 | Since the task is already fully automated in most workflows, there is little room for AI to 'augment' a human performing it manually today. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Mailing charge statements is a fully automatable end-to-end task: AI can generate statements, format them, address envelopes, integrate with mail-merge systems, and trigger postal delivery—all with substantial time and cost savings compared to manual preparation and dispatch. |
| Task automatability | claude-sonnet-5 | 5/5 | Generating and sending charge statements is a fully rule-based, templated process (data merge plus dispatch) that automated billing systems already handle end-to-end with no quality loss. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Minimal legal barriers exist for automating statement generation and delivery, though some institutions face legacy compliance requirements around statement retention and optional human review; customer preference for digital statements further reduces friction. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, liability, or human-judgment requirement attaches to sending routine charge statements; it's a purely administrative task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated mailing via digital statement systems and third-party fulfillment services costs a fraction of human labor for envelope preparation, printing, and postal processing, achieving an order-of-magnitude cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated statement generation and mail-merge/print-fulfillment services cost a small fraction of a cent per statement compared to manual clerical processing costs. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Automated statement generation and mailing systems are mature, deployed products used at scale by financial institutions and utilities today; such systems reliably handle high volumes with minimal error rates in production. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Mature billing/invoicing software and e-statement systems reliably automate statement generation and mailing (print vendors or email) at scale across finance and retail sectors today. |
File sales slips in customers' ledgers for billing purposes.
96CI 92–100 · exposure 100 · augmentation 63 · importance 4.4/5 · click for rater detail
File sales slips in customers' ledgers for billing purposes.
96| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Accounting automation and ledger posting are among the earliest and most deeply adopted AI/automation functions in finance and professional services. The vast majority of firms with any digital accounting infrastructure have already automated this task. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Billing/accounting functions across finance and retail sectors have widely adopted automated ledger and invoicing systems, though some smaller firms still use manual processes. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | While the task is largely fully automatable, AI augmentation remains valuable for exception handling, flagging anomalies, and assisting human reviewers in reconciliation workflows. Humans working alongside automated systems benefit from AI-assisted validation and anomaly detection. |
| Augmentation potential | claude-sonnet-5 | 3/5 | For any remaining manual filing, AI tools can assist with data extraction and entry, but the task itself is largely already automated rather than augmented. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Filing sales slips in customers' ledgers is a highly structured, repetitive data-entry task that modern OCR and accounting automation systems can perform end-to-end with significant time savings. The task involves reading transaction data and posting it to the correct ledger—a core function of accounting software and RPA systems deployed today. |
| Task automatability | claude-sonnet-5 | 5/5 | Filing/recording sales slip data into customer ledgers is a structured, repetitive data-entry task that automated systems (OCR + billing software integrations) can fully handle with substantial time savings. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Regulatory and audit requirements (GAAP, SOX compliance) do impose some oversight and internal-control requirements, but they do not legally mandate human performance of posting itself—only documentation and review. Most organizations have already automated this function, indicating low structural barriers. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, liability, or human-contact requirement blocks automating routine internal record-filing; it's already standard practice. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated ledger posting through accounting software costs a fraction of a cent per transaction after initial setup, whereas manual filing by a clerk costs $15–25 per hour loaded. The cost differential is at least one order of magnitude in favor of automation. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated ledger posting via software costs a tiny fraction of a clerk's wage per transaction, especially at volume. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature accounting and ERP systems (SAP, NetSuite, QuickBooks, etc.) deployed at scale in production reliably handle automated posting of sales transactions to customer ledgers with minimal error rates. This is a standard, battle-tested capability in modern financial software. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Automated billing and accounting systems (ERP, POS integrations) have long performed this exact function reliably in production at scale, well before generative AI. |
Keep records of customers' charges and payments.
94CI 92–95 · exposure 100 · augmentation 75 · importance 4.5/5 · click for rater detail
Keep records of customers' charges and payments.
94| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Financial services, credit institutions, and accounting firms have already deeply adopted automated payment and charge recording systems. This is a mature, widespread automation in information and finance sectors, representing some of the earliest and most complete displacement in white-collar work. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Financial services and credit administration functions have deeply adopted automated recordkeeping and accounting software for decades, with continued modernization via cloud and AI-enhanced platforms. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI augments human clerks by automatically flagging unusual transactions, suggesting corrections, and summarizing account activity, allowing humans to focus on exception handling and customer service rather than routine data entry. This transforms productivity while the human remains available for judgment calls. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-enhanced systems improve accuracy, flag anomalies, and speed reconciliation, substantially boosting productivity for remaining human oversight tasks. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Recording customer charges and payments is fundamentally a data entry and ledger-management task that AI systems can perform end-to-end with significant time savings. Current software and AI can capture transaction details, validate entries, categorize charges, and update customer records with minimal human intervention, easily meeting the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 5/5 | Recording charges and payments is a structured, rules-based data entry and reconciliation task that is fully handled by existing accounting and ledger software with automation/RPA layers, meeting the 50% time-saving bar easily. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While regulatory oversight applies to financial records generally, there are no strict legal requirements that a licensed human must personally execute this clerical task; organizations can and do automate it today. Some oversight and audit trails are required, but they do not mandate human performance of the recording itself. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Some oversight and audit-trail requirements exist for financial recordkeeping, but no licensing requirement mandates a human perform basic transaction recording. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated record-keeping via deployed systems costs orders of magnitude less than paying a human clerk to manually log and maintain customer accounts. A single system can serve thousands of accounts with minimal marginal cost per transaction. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated ledger/database systems process transactions at a fraction of a cent per record compared to hourly clerical wages, an order-of-magnitude cost advantage. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature products already perform this task reliably in production across banking, fintech, and accounting systems. Automated payment processing, accounting software, and transaction recording are standard deployments in financial institutions handling millions of records daily. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Mature ERP, billing, and accounting systems (e.g., QuickBooks, SAP, credit management platforms) already perform automated transaction recording and reconciliation reliably in production at scale. |
Evaluate customers' computerized credit records and payment histories to decide whether to approve new credit, based on predetermined standards.
88CI 85–90 · exposure 100 · augmentation 75 · importance 4.6/5 · click for rater detail
Evaluate customers' computerized credit records and payment histories to decide whether to approve new credit, based on predetermined standards.
88| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Financial services and lending sectors have rapidly and deeply adopted automated credit decisioning; major banks, credit card issuers, and fintech platforms use AI-driven underwriting as standard practice, with human review now the exception rather than the rule for routine applications. |
| Sector adoption velocity | claude-sonnet-5 | 5/5 | Consumer lending and financial services have deeply adopted automated underwriting and credit-decision algorithms for decades, representing one of the most mature AI/automation use cases in any industry. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assists credit officers by flagging high-risk or borderline cases, summarizing payment histories, and highlighting relevant factors, allowing humans to focus judgment on complex edge cases and exceptions rather than routine screening. |
| Augmentation potential | claude-sonnet-5 | 4/5 | Where human review persists (exceptions, disputes, non-standard cases), AI-driven scoring and record synthesis significantly speeds up and informs the authorizer's judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Credit evaluation is a data-driven task performed on structured digital records with predetermined, rule-based approval criteria. Current AI systems can assess payment histories, credit scores, and risk metrics at scale and speed, easily exceeding 50% time savings while maintaining or improving quality through consistent application of lending standards. |
| Task automatability | claude-sonnet-5 | 5/5 | This is a structured, data-driven decision task using computerized records and predetermined rules, which is exactly the type of task algorithmic and ML-based credit decision systems already perform at scale with substantial time savings. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Credit decisions face significant regulatory barriers: Fair Lending Act, Equal Credit Opportunity Act, and increasing regulatory scrutiny of algorithmic bias and explainability in credit decisioning. Lenders must maintain human oversight, audit trails, and compliance documentation, slowing full automation despite technical capability. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Fair lending, adverse action notice, and credit regulation (e.g., ECOA, FCRA) impose compliance and explainability requirements, and human review is often retained for edge cases or disputes, creating moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated credit decisioning costs pennies per application (inference + integration), while a human credit analyst costs $25–50+ per hour to review a single application. AI is orders of magnitude cheaper per decision when scaled. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated scoring engines process applications in seconds at near-zero marginal cost compared to a human clerk's wage and time per file review. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Credit decisioning AI and automated underwriting systems are mature, deployed products in use by major financial institutions, banks, and fintech platforms today. These systems reliably process millions of credit applications annually in production environments with well-established performance metrics. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Automated credit scoring and decisioning systems are mature, widely deployed products used by banks, card issuers, and fintechs for real-time approval decisions in production today. |
Relay credit report information to subscribers by mail or by telephone.
81CI 74–87 · exposure 83 · augmentation 38 · importance 3.3/5 · click for rater detail
Relay credit report information to subscribers by mail or by telephone.
81| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Credit agencies, banks, and fintech platforms have rapidly adopted automated notification systems for credit reports over the past decade; this is standard practice in the highly digitized financial services sector. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Financial services and credit reporting are high-digitization sectors with strong existing automation of customer communications and self-service portals, indicating fast, deep adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | Automation handles routine delivery; limited augmentation benefit remains since AI can already fully perform the core task and human judgment adds little value to standardized credit report transmission. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can draft responses, pull data, and support phone agents in real time, improving speed and consistency, though human clerks still often handle escalations and verification. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Relaying credit report information to subscribers can be substantially automated through email systems, automated telephone systems with text-to-speech, or API delivery, achieving significant time savings. However, handling complex inquiries or exceptions may require human intervention, preventing a full 5. |
| Task automatability | claude-sonnet-5 | 5/5 | Relaying structured credit report data via mail or phone is a templated information-transfer task that AI systems (chatbots, IVR, automated letter/email generation) can fully handle with equal or better accuracy and major time savings. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Fair Credit Reporting Act (FCRA) compliance and data security requirements apply, but do not prohibit automation; no licensing mandate requires a human to deliver the information. Organizations must implement proper controls, but automation itself is legally permissible. |
| Adoption barriers | claude-sonnet-5 | 3/5 | FCRA and related regulations impose disclosure and accuracy requirements and privacy handling obligations, creating moderate compliance friction, though no law requires a human to personally relay the information. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated email, SMS, and IVR systems cost a fraction of a cent per delivery, far below the loaded wage of a human clerk handling even a modest volume of such notifications. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated data relay via generated letters, emails, or voice/chat systems costs a small fraction of a human clerk's wage per interaction, especially at volume. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Automated systems for credit report delivery are mature and deployed at scale by credit reporting agencies and fintech firms; voice and email notification systems reliably deliver standardized information. Minor friction remains in handling personalized or complex requests, but the core task is well-established in production. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Automated credit report delivery systems, IVR phone systems, and chatbot-based inquiry handling are already deployed at scale by credit bureaus and financial institutions, though some phone interactions still route to humans for complex questions. |
Review individual or commercial customer files to identify and select delinquent accounts for collection.
74CI 70–79 · exposure 75 · augmentation 75 · click for rater detail
Review individual or commercial customer files to identify and select delinquent accounts for collection.
74| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Financial services and credit operations are highly digitized sectors with rapid AI adoption; many lenders and collection agencies already deploy automated delinquency flagging systems in production, though not universally. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Financial services is a fast-adopting sector for AI-driven credit risk and collections analytics, with mature production deployment of automated account flagging systems. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially assists human clerks and authorizers by pre-filtering accounts, prioritizing high-risk delinquencies, and surfacing relevant data, allowing staff to focus on complex or disputed cases while maintaining human oversight on final selection decisions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly enhance clerks' ability to prioritize and identify delinquent accounts by surfacing risk scores and patterns, even where human review of borderline cases remains. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can reliably identify delinquent accounts by reviewing transaction history, payment dates, and account status data with high accuracy, meeting the 50% time-saving threshold. However, edge cases involving legitimate disputes or special circumstances may still require human judgment, preventing a perfect 5. |
| Task automatability | claude-sonnet-5 | 4/5 | Identifying and flagging delinquent accounts from structured customer data is a well-defined, rule-based/scoring task that current AI and standard analytics systems can perform largely automatically, saving significant time over manual file review.4/5 reflects that some edge cases and judgment calls remain. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Regulatory oversight (Fair Debt Collection Practices Act, FCRA, banking compliance) and internal approval workflows require human sign-off or review of AI-selected accounts before collection action, creating moderate friction. Liability concerns over false positives add oversight requirements. |
| Adoption barriers | claude-sonnet-5 | 2/5 | There are some regulatory requirements around fair lending and collections practices, but the identification/selection step itself is largely administrative and not subject to strict licensing requirements, unlike final collection actions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated delinquency detection via AI agents costs orders of magnitude less than human review—a single API call or batch processing run replaces hours of manual file review per account, with minimal ongoing overhead. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated scoring and flagging of accounts is computationally cheap compared to manual clerk review of individual files, especially at scale across large portfolios. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed financial software and AI systems in banks and fintech firms regularly automate delinquency identification and account flagging in production at scale. Some material error rates remain in complex cases, but the core task is reliably performed by existing products. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Collections software and credit management systems widely deployed in banks and lenders already use automated rules and predictive models to flag delinquent accounts for collection action, though human review still occurs for exceptions. |
Compile and analyze credit information gathered by investigation.
72CI 70–74 · exposure 75 · augmentation 100 · importance 4.4/5 · click for rater detail
Compile and analyze credit information gathered by investigation.
72| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Financial services and credit risk sectors have aggressively adopted automated credit analysis and decisioning tools over the past decade; many large lenders now process volumes automatically. Smaller lenders and niche credit products still rely on manual review, but the trend and deployment depth are substantial. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Financial services is among the fastest AI-adopting sectors, with automated credit scoring and underwriting tools widely deployed in production across banks, fintechs, and credit bureaus. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI-powered credit analysis platforms assist human authorizers by flagging inconsistencies, summarizing files, and highlighting risk factors, significantly accelerating review time and reducing error while keeping the human in final decision-making loops. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools significantly speed up data compilation, flag anomalies, and summarize investigative findings, letting human clerks focus on judgment calls and exceptions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can perform document review, data extraction, and pattern recognition across credit files with high reliability and >50% time savings. However, some subjective judgment calls on ambiguous applicant information may still require human oversight, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 4/5 | Compiling structured credit data and generating summary analysis is largely pattern-based work that current AI/ML credit-scoring and document-analysis systems already handle well, though edge cases and ambiguous investigative findings still require human judgment. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Credit decisions are subject to Fair Lending laws and regulatory scrutiny (FCRA, ECOA), creating compliance and audit requirements that slow automation but do not legally forbid it. Many institutions still require human sign-off on final decisions, introducing friction without hard legal barriers to task automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Credit decisions are subject to fair-lending regulations (ECOA, FCRA) requiring explainability and audit trails, and some human review is often mandated for adverse actions, creating moderate regulatory friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated credit analysis via APIs and intelligent document processing costs pennies per application versus the $30–50 loaded cost of a human clerk performing similar compilation and basic analysis, yielding >10x cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated data aggregation and scoring systems process large volumes of credit information at a fraction of the marginal cost of a human clerk once integrated, though initial integration and compliance costs are non-trivial. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed document intelligence and credit analysis platforms (e.g., intelligent document processing, rule-based credit decisioning engines) reliably extract and flag credit data in production. Minor gaps remain in handling novel or unusual file formats, but mainstream credit bureau data flows are well-handled. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Automated underwriting and credit decision-support systems (e.g., FICO-based automated scoring, AI-driven document extraction tools) are deployed at scale in banks and lenders today, though fully autonomous compilation of investigative findings from disparate sources is less mature. |
Obtain information about potential creditors from banks, credit bureaus, and other credit services, and provide reciprocal information if requested.
72CI 70–74 · exposure 75 · augmentation 75 · importance 4.0/5 · click for rater detail
Obtain information about potential creditors from banks, credit bureaus, and other credit services, and provide reciprocal information if requested.
72| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Large financial institutions and credit bureaus have extensively automated information retrieval and reciprocal sharing workflows; adoption is rapid and deep in the formal credit and banking sector, though smaller lenders may lag. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Banking and financial services are among the fastest adopters of automation and API-driven data exchange, with credit bureau integrations being industry-standard practice for years. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems can accelerate data gathering, validation, and matching of creditor records, allowing human reviewers to focus on exception handling and compliance verification; this significantly raises throughput and accuracy for clerks managing high-volume inquiries. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI and automated systems significantly speed up data retrieval and cross-referencing tasks for clerks, letting them focus on exception handling and judgment calls rather than manual lookups. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | API-driven data retrieval from banks and credit bureaus is largely automatable today; information gathering and cross-referencing can be performed by AI agents with minimal human intervention, achieving well over 50% time savings. However, nuanced judgment about what reciprocal information to share in borderline cases may still require human oversight. |
| Task automatability | claude-sonnet-5 | 4/5 | Retrieving and compiling credit information from standardized databases and bureaus is largely structured data retrieval, which AI/automation systems can execute with substantial time savings via API integrations and RPA.dio. Some reciprocal reporting still requires judgment on data-sharing policies. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Regulatory oversight (FCRA, banking secrecy rules) and data-sharing agreements impose compliance checks and audit trails that require human review and legal sign-off in many jurisdictions; however, the task itself is not restricted to licensed practitioners, only subject to procedural controls. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Financial data-sharing is regulated (FCRA, GLBA) requiring compliance safeguards and audit trails, and some reciprocal information exchanges may require authorized personnel sign-off, creating moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated API calls and database queries incur minimal inference costs compared to the loaded wage of a human clerk performing repeated lookups and data transfers across systems; the cost advantage is at least one order of magnitude. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | API-based credit data retrieval and automated reciprocal reporting cost a fraction of a clerk's hourly wage once integration is built, though setup and compliance overhead reduce the ratio somewhat. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Credit data integration platforms and automated information-sharing systems are deployed in production by financial institutions and credit bureaus; APIs reliably pull and validate creditor information at scale. Reciprocal information sharing is governed by protocol and can be automated, though some organizations still use semi-manual workflows. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Automated credit bureau pulls (via Experian, Equifax, TransUnion APIs) and RPA-based information exchange are already standard in banking and lending operations today, functioning reliably in production. |
Examine city directories and public records to verify residence property ownership, bankruptcies, liens, arrest record, or unpaid taxes of applicants.
68CI 62–74 · exposure 70 · augmentation 63 · importance 3.4/5 · click for rater detail
Examine city directories and public records to verify residence property ownership, bankruptcies, liens, arrest record, or unpaid taxes of applicants.
68| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Credit bureaus, financial services, and background check vendors have been automating this task for decades. Modern loan origination and identity verification platforms increasingly run these checks automatically; adoption in finance and FinTech is deep and rapid. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Financial services and credit-related back-office functions have moderate AI adoption with many pilots and some production tools, but full deployment across all credit authorization workflows remains uneven. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-powered public record lookup assists human reviewers by instantly surfacing contradictions, flagging discrepancies, and organizing findings—useful augmentation for a human clerk who must interpret nuanced conflict scenarios or make final adjudication calls. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools substantially speed up cross-referencing public records and flagging discrepancies, greatly aiding human clerks who still validate and make final determinations. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can reliably extract and verify residence, property ownership, liens, and tax records from public databases and online records with high accuracy. This is a structured data lookup and cross-reference task where AI excels, likely achieving >50% time savings when integrated with existing public record APIs. |
| Task automatability | claude-sonnet-5 | 4/5 | This is largely a structured information-retrieval and verification task against public records and directories, which AI-driven data aggregation and lookup tools can perform quickly with high time savings, though some edge-case judgment remains. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Moderate barriers exist: fair-use restrictions on scraped public records, compliance with FCRA and similar regulations for credit/background checks, and organizational reliance on authorized data sources. However, no law requires a human signature on record lookups themselves, and many firms already license automated verification. |
| Adoption barriers | claude-sonnet-5 | 3/5 | There are moderate regulatory constraints (FCRA compliance, data accuracy liability) around using such records for credit decisions, requiring oversight, though no strict licensing mandate requires a human to perform the lookup itself. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated public record lookup and verification costs pennies per query after initial integration, whereas a human clerk examining directories and records manually incurs full labor cost. AI is orders of magnitude cheaper per completed verification. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated public record and database queries cost a small fraction of a clerk's hourly wage once integrated, though data licensing and verification overhead keep it from being a full order-of-magnitude cheaper in all cases. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed products perform public record searches and verification at scale (e.g., credit bureaus, background check services, property title search platforms). While error rates exist for edge cases (name ambiguity, record conflicts), production systems handle the core task reliably across large volumes. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed background-check and credit verification platforms (e.g., LexisNexis, identity verification APIs) already automate much of this record-pulling, but accuracy across fragmented, non-standardized public record systems still requires human review in production settings. |
Prepare reports of findings and recommendations.
66CI 62–70 · exposure 70 · augmentation 100 · click for rater detail
Prepare reports of findings and recommendations.
66| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Financial services and credit operations are early adopters of AI and automation; many institutions are actively piloting and deploying AI-assisted reporting and decision support. Report generation and summarization rank among the highest-velocity automation use cases in fintech and banking. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Financial services broadly adopt AI for document drafting and analytics, but credit clerk roles specifically show slower, more cautious integration due to compliance concerns. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI powerfully augments credit analysts and authorizers by drafting reports, organizing findings, and proposing recommendations, freeing humans to focus on judgment, exception handling, and validation. This is a leading use case for LLM-based assistants in credit and compliance roles. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI strongly assists by synthesizing data, generating draft narratives, and flagging risk factors, letting the human clerk focus on verification and judgment calls. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can generate structured reports with findings and recommendations from data, documentation, and analysis with minimal human intervention. Modern LLMs and business intelligence tools can synthesize information, identify patterns, and draft professional recommendations, achieving >50% time savings on report writing and organization. |
| Task automatability | claude-sonnet-5 | 4/5 | AI can draft structured reports summarizing credit findings and recommendations from underlying data with significant time savings, though final review/edits are typically needed for accuracy and compliance.ed |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Credit decisions and findings often require human sign-off and may have regulatory oversight requirements (compliance, audit trails), and final recommendations typically remain a human responsibility. However, the report-preparation component itself faces no hard legal barriers to automation, only organizational preference for human review. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Credit decisions may require documented human accountability and compliance with fair lending regulations, creating moderate oversight requirements even if drafting is automated. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference and integration costs for report generation are a fraction of the loaded salary of a credit clerk or authorizer, especially when handling volume. One model invocation can replace 1–2 hours of human report drafting labor, yielding a significant cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Once integrated with data sources, AI-generated drafts cost a fraction of the loaded wage for clerical report-writing time, though some oversight cost remains. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (LLMs, report automation platforms, BI tools) routinely generate business reports with findings and recommendations in production environments. Minor gaps remain in domain-specific nuance and context-dependent judgment, but the core task is reliably performed at scale in financial and credit operations. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Generative AI and template-based reporting tools are deployed in financial services for drafting summaries and recommendations, but full end-to-end autonomous credit report generation with reliable accuracy is not yet standard practice at scale. |
Receive charge slips or credit applications by mail, or receive information from salespeople or merchants by telephone.
59CI 50–67 · exposure 50 · augmentation 75 · importance 3.9/5 · click for rater detail
Receive charge slips or credit applications by mail, or receive information from salespeople or merchants by telephone.
59| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Credit and lending organizations have adopted some mail scanning and transcription automation, but adoption is piecemeal and often supplementary rather than wholesale replacement. Many smaller merchant networks and credit departments still use manual intake, placing this in the pilot-and-mixed-deployment stage rather than deep, sector-wide production deployment. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Financial services and credit processing are moderately fast adopters of automation for intake tasks, but full replacement of phone-based interactions lags behind document-based automation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems (OCR, speech-to-text, data entry assistance) materially augment intake staff by automatically capturing and pre-populating forms, allowing humans to focus on verification and exception handling. These tools demonstrably raise throughput and reduce transcription burden. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-assisted intake tools (auto-transcription, form auto-fill, OCR extraction) meaningfully speed up the clerk's initial data capture step while a human still verifies and processes the information. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Receiving and logging charge slips or applications can be partially automated through mail scanning and OCR, and phone data can be captured via call recording/transcription systems. However, capturing information accurately from phone conversations still requires significant human intervention to resolve ambiguities, verify details, and handle edge cases, so the task does not meet the ≥50% time-saving threshold end-to-end. |
| Task automatability | claude-sonnet-5 | 4/5 | Receiving and logging charge slips, applications, or phone-relayed information is largely structured data intake that can be automated via OCR, document ingestion pipelines, and call-handling/IVR or voice-AI systems with human fallback for edge cases. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing or regulatory requirement mandates a human receive applications or charge slips; they are information intake steps. However, organizations may retain human receivers to ensure accuracy and handle exception cases, creating modest organizational and quality-assurance friction rather than hard legal barriers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for receiving intake information, but some organizational preference for human contact with merchants/salespeople and internal verification protocols create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Automated mail scanning and call transcription are now commoditized services (OCR, speech-to-text APIs cost pennies per item). The all-in cost of automation per received application is substantially lower than the loaded wage of a full-time data entry clerk. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated document capture and voice-to-text/intake systems are cheap per transaction relative to a human clerk's wage, especially at scale, though initial integration costs exist. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Documented systems exist for mail receipt automation (scanning, optical character recognition) and call transcription/logging (available in modern telephony platforms), but these products typically require human review and correction due to error rates, especially for handwritten applications or poor audio quality. Deployments are real but material error rates persist. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed document ingestion and intelligent intake systems (OCR, e-forms, API submissions) and AI call agents exist and are used in financial services, but many organizations still route phone-based merchant/salesperson information through human staff due to variability and verification needs. |
Prepare credit cards or charge account plates.
51CI 23–79 · exposure 45 · augmentation 13 · importance 3.0/5 · click for rater detail
Prepare credit cards or charge account plates.
51| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Credit card and charge account plate preparation is a mature, physical manufacturing process handled by specialized third-party vendors and legacy equipment. The sector shows little active adoption of novel AI approaches, and digitization has already plateaued. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Financial services and retail card issuance are high-digitization sectors with mature automated production systems already deployed at scale. |
| Augmentation potential | claude-haiku-4-5-20251001 | 1/5 | AI offers minimal assistance to workers preparing physical cards or plates, as the task is already either manual or handled by existing machinery with limited cognitive components to augment. |
| Augmentation potential | claude-sonnet-5 | 2/5 | Since the task is largely mechanical, AI's role is more about full automation of the process rather than augmenting a human performing it manually. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Preparing physical credit cards or charge account plates involves physical handling, embossing, cutting, and assembly steps that require robotic manipulation or manual labor. While data entry and batching could be automated, the physical production component remains difficult to fully automate without specialized hardware, limiting time savings to data-related portions only. |
| Task automatability | claude-sonnet-5 | 4/5 | This is a highly structured, rules-based data entry and card-issuance task that maps well to automated systems already used by card processors and issuers. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Security and compliance requirements around card production (fraud prevention, PCI standards) create moderate friction, though they are not absolute legal barriers. Organizations prefer established, certified card production vendors over novel AI-based alternatives. |
| Adoption barriers | claude-sonnet-5 | 2/5 | Some regulatory and security compliance (PCI, identity verification) applies to card issuance, but the physical preparation task itself carries no licensing requirement and is already largely automated in industry. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Existing card production equipment requires significant capital investment, and current labor costs for manual preparation or semi-automated processes are relatively low. AI-driven solutions would need to compete with established, amortized manufacturing infrastructure. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated card production and account plate systems are vastly cheaper per unit than manual clerical processing once integrated into existing issuance pipelines. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI systems or production automation solutions reliably handle the full task of physically preparing credit cards or charge account plates end-to-end. This remains primarily a manual or legacy machinery task in real organizations. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Card issuance platforms and instant-issuance systems at banks/retailers already automate embossing, encoding, and account setup with minimal human involvement in production today. |
Call customers to collect payment on delinquent accounts.
44CI 35–54 · exposure 38 · augmentation 63 · click for rater detail
Call customers to collect payment on delinquent accounts.
44| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Collection agencies and financial institutions have piloted AI-driven outbound calling, but adoption remains moderate due to regulatory uncertainty, customer complaint backlash, and need for human oversight. The sector is cautious rather than aggressively deploying automation at scale. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Financial services and collections agencies are moderately fast adopters of voice AI and chatbots for early-stage delinquency, but human-heavy processes persist for higher-risk accounts. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist human collectors by preparing account summaries, suggesting payment arrangements, and pre-filtering accounts by likelihood-to-pay, raising their efficiency. However, the core negotiation and relationship-building still benefits from human judgment, so augmentation is partial rather than transformative. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can pre-screen accounts, draft call scripts, predict best contact times, and handle first-touch reminders, meaningfully boosting human collector productivity. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate outbound calls and basic payment prompts, the task requires nuanced negotiation, customer relationship management, and real-time adaptation to diverse objections—skills current systems handle poorly at scale. Meaningful automation would need to handle payment arrangement complexity and error recovery that today's voice agents struggle with reliably. |
| Task automatability | claude-sonnet-5 | 3/5 | AI voice agents can conduct routine collections calls with scripted negotiation, but complex disputes, emotional escalation, and legal compliance nuances still require human handling for a meaningful share of calls.atta |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | This task faces substantial regulatory barriers including FDCPA (Fair Debt Collection Practices Act) restrictions, state-level debt collection licensing requirements, and consumer protection rules that limit automated calling frequency and methods. Many jurisdictions require a licensed debt collector or explicit authorization, creating legal friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Debt collection is subject to FDCPA and similar regulations requiring specific disclosures and prohibiting harassment, creating compliance friction, though not requiring a licensed human per se. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven outbound calling infrastructure is significantly cheaper per attempted contact than human labor, with minimal per-call marginal cost once deployed. The cost advantage is substantial despite integration and compliance overhead. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated dialers and voice AI cost a fraction of a human agent's loaded wage per call, though compliance monitoring and escalation paths add some overhead. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Deployed robocall and voice AI systems exist but suffer from high failure rates, customer friction, and regulatory compliance issues; they cannot independently navigate the full payment collection conversation at production quality. Most real-world deployments still require human escalation for substantive resolution. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Voice AI collections agents are deployed by some fintechs and BPOs, but adoption is narrower than claims, with material error/complaint rates and human backup still common. |
Consult with customers to resolve complaints or verify financial or credit transactions.
44CI 41–46 · exposure 34 · augmentation 75 · click for rater detail
Consult with customers to resolve complaints or verify financial or credit transactions.
44| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Large financial institutions are piloting AI-driven complaint resolution and transaction verification, but deployment remains inconsistent. Smaller credit operations and community banks lag significantly, and many organizations still rely on mixed human-AI workflows. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Financial services is a digitized sector with growing chatbot and AI adoption for customer service, but complaint resolution specifically still sees mixed pilot-to-production deployment given liability concerns. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI tools that present transaction history, flag anomalies, and suggest resolutions significantly boost human clerk productivity in handling complaint triage and verification. Humans remain the final decision-maker, but AI-assisted workflows materially accelerate resolution. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools significantly speed up verification lookups, draft responses, and flag likely fraud or discrepancies, letting human clerks focus judgment on genuine complaint resolution. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can verify financial transactions with high accuracy, resolving customer complaints requires nuanced judgment, empathy, and handling of edge cases. Customer interaction and complaint resolution are highly variable and context-dependent, preventing end-to-end automation that meets the 50% time-saving threshold at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Resolving complaints and verifying transactions often requires nuanced judgment, empathy, and handling of edge cases that current AI struggles to fully replicate end-to-end.rt Chatbots can handle simple verification but complaint resolution still needs human escalation frequently. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Regulatory oversight and potential liability for incorrect credit decisions or unresolved customer complaints create modest friction, but no legal requirement mandates human sign-off. Financial institutions face reputational and compliance risks that slow substitution but do not prevent it. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Financial services carry regulatory compliance requirements (e.g., dispute resolution rules, KYC) and customer preference for human contact when money and credit are involved, creating moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI verification systems have lower unit costs than human review, but integration with existing credit systems and ongoing oversight needs offset savings. The total cost per resolved complaint is roughly comparable to human clerk labor for complex cases. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-driven verification bots are cheap to run, but complaint resolution requiring human oversight and escalation adds cost, keeping overall cost roughly comparable rather than a clear order-of-magnitude savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed products exist for transaction verification and basic complaint triage, but few systems handle full complaint resolution reliably in production. Current chatbots and rule-based systems can handle routine inquiries but often fail on complex complaints requiring human judgment. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed chatbots and IVR systems handle basic verification and FAQ-style complaints in banks and credit unions, but complex disputes routinely escalate to human agents, showing narrow scope and material error rates. |
Interview credit applicants by telephone or in person to obtain personal and financial data needed to complete credit report.
41CI 25–56 · exposure 38 · augmentation 63 · importance 4.0/5 · click for rater detail
Interview credit applicants by telephone or in person to obtain personal and financial data needed to complete credit report.
41| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Financial services has adopted AI for some data processing and decisioning, but interview automation in credit remains limited; most banks and lenders still employ staff for this task or rely on human-supervised hybrid workflows with low measured displacement. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Financial services is a fast-adopting sector for AI-driven customer intake and chatbots, with many banks and lenders already using automated systems for initial data collection. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by suggesting follow-up questions, auto-populating forms from spoken data, and flagging inconsistencies, raising clerk productivity moderately, but the human must remain the primary interviewer to ensure legal compliance and relationship rapport. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can pre-fill forms, transcribe calls, flag missing data, and prompt interviewers with next questions, meaningfully speeding up the interview and reducing errors while a human remains for final review. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Conducting a full credit interview requires empathetic engagement, complex follow-up questioning, and judgment calls about sensitive financial disclosures that current AI struggles with at production quality. While AI can handle some scripted information gathering, the nuanced interpersonal dynamics and need to probe inconsistencies make full automation well below the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 3/5 | Data collection via structured interview can largely be automated with conversational AI/chatbots and voice agents, but edge cases, applicant hesitancy, and nuanced follow-up questions still require human judgment for full end-to-end quality parity. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Credit authorization is subject to fair lending laws (ECOA, FHA), anti-discrimination rules, and Know Your Customer (KYC) requirements; using AI without human sign-off creates legal and compliance risk, effectively requiring a human to review and validate the interview and credit decision. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Financial data collection is subject to consumer protection and fair lending regulations (e.g., ECOA, TILA) requiring accurate disclosures and non-discriminatory practices, creating moderate compliance friction though not strictly requiring a licensed human. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Integration and oversight costs for AI interview systems (including regulatory compliance, error checking, and human escalation) currently approach or exceed the cost of a credit clerk conducting the interview, especially when accounting for liability and rework. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated voice/chat intake systems cost a fraction of a human interviewer's loaded wage per interaction, though integration with core banking and compliance oversight adds some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably conducts end-to-end credit interviews via phone or in-person with the accuracy and legal compliance required. Some companies use chatbots for preliminary screening, but human review remains mandatory and error rates are material in the financial context. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Voice bots and chat-based intake forms are deployed in banking and fintech for credit applications, but many institutions still route to human agents for complex or high-value applications, indicating narrower production reliability. |
Contact former employers and other acquaintances to verify applicants' references, employment, health history, or social behavior.
31CI 25–36 · exposure 25 · augmentation 50 · click for rater detail
Contact former employers and other acquaintances to verify applicants' references, employment, health history, or social behavior.
31| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for reference verification remains very limited. Most lending, credit, and HR operations still use manual calls, form letters, and human verification specialists. While automation of adjacent tasks (data entry, response parsing) is increasing, the contact-and-verify function itself has seen minimal production AI displacement due to regulatory friction and liability concerns. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Background/reference-check industry is adopting automation (e.g., automated employment verification databases) at a moderate pace, but many verifications still rely on manual phone outreach. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by generating contact lists, drafting outreach templates, organizing and flagging response data for human review, and flagging inconsistencies—moderately raising human productivity. However, the assistant role is constrained because humans must retain full responsibility for contact authenticity and credibility judgment, limiting the transformative potential. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can draft verification scripts, transcribe calls, flag inconsistencies, and pre-fill forms, meaningfully speeding up the clerk's workflow while humans still make final judgment calls. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can draft reference verification requests and parse responses, the task requires interactive dialogue with external parties who expect legitimate human contact, voice authenticity verification, and judgment about whether responses are credible. Current systems cannot reliably handle this end-to-end with 50% time savings because they cannot actually place calls, authenticate themselves to third parties, or navigate social resistance to speaking with bots. |
| Task automatability | claude-sonnet-5 | 2/5 | Contacting third parties and eliciting candid verbal/written verification involves live phone calls and judgment about ambiguous responses, which AI can partially script but not fully replace end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: regulatory requirements (Fair Credit Reporting Act, lending rules) often require documented verification trails signed by a licensed human; liability asymmetry (bad hires from falsified references create employment law exposure); and third-party contacts resist AI callers. Many jurisdictions and organizations explicitly require human-conducted reference checks for compliance. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Privacy laws (FCRA, GDPR) and consent requirements around health history and background checks impose meaningful compliance friction, though not an outright licensing barrier. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI contact systems (voice agents, email outreach) still require significant human oversight to verify contact success and credibility of responses, plus integration with HR systems. The loaded cost of human verification agents remains lower when you account for legal liability, failed contact recovery, and mandatory human review of findings. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated calling/verification systems can be cheaper per contact than a human clerk, but exception handling, callbacks, and disputes still require human labor, keeping overall cost roughly comparable. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs live reference verification calls at scale. Voice AI can simulate calls but organizations resist AI impersonation for legal and ethical reasons. Existing systems perform narrow steps (data lookup, response parsing) but not the full contact-and-verify workflow that satisfies compliance and due-diligence requirements. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some voice-bot and automated verification services exist (e.g., employment verification platforms) but they mostly handle structured data lookups, not open-ended interviews about health history or social behavior. |
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.